Discourse Particles: Formal Approaches to their Syntax and Semantics 9783110497151, 9783110488821

Citation preview

Discourse Particles

Linguistische Arbeiten

Edited by Klaus von Heusinger, Gereon Müller, Ingo Plag, Beatrice Primus, Elisabeth Stark and Richard Wiese

Volume 564

Discourse Particles Formal Approaches to their Syntax and Semantics Edited by Josef Bayer and Volker Struckmeier

DE GRUYTER

ISBN 978-3-11-048882-1 e-ISBN [PDF] 978-3-11-049715-1 e-ISBN [EPUB] 978-3-11-049348-1 ISSN 0344-6727 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. © 2017 Walter de Gruyter GmbH, Berlin/Boston Typesetting: Compuscript Ltd., Shannon, Ireland Printing: CPI books GmbH, Leck ♾ Printed on acid-free paper Printed in Germany www.degruyter.com

Contents Josef Bayer and Volker Struckmeier The status quo of research on discourse particles in syntax and semantics 1 Markus Egg and Johannes Mursell The syntax and semantics of discourse particles

15

Waltraud Paul and Victor Junnan Pan What you see is what you get: Chinese sentence-final particles as head-final complementizers 49 Nathalie Scherf The syntax of Swedish modal particles

78

Jóhannes Gísli Jónsson Discourse particles and hvað-exclamatives

100

Hans-Martin Gärtner Root infinitivals and modal particles. An interim report Daniel Gutzmann Modal particles ≠ modal particles (= modal particles)

115

144

Yvonne Viesel Discourse particles “embedded”: German ja in adjectival phrases Sonja Müller Combining ja and doch: A case of discourse structural iconicity Werner Abraham Discourse marker = discourse particle = thetical = modal particle? A futile comparison 241 Rosemarie Lühr Stressed and unstressed particles in Old Indic

281

173

203

vi 

 Contents

Svetlana Petrova On the status and the interpretation of the left-peripheral sentence particles inu and ia in Old High German 304 Index

332

Josef Bayer and Volker Struckmeier

The status quo of research on discourse particles in syntax and semantics What are discourse particles (DiPs), also known as modal particles (German Modalpartikeln) or downtoners (German Abtönungspartikeln)? Morphosyntactically, they are simply non-inflecting parts of speech. In this sense, they pattern with prepositions, complementizers, adverbs, focus particles and various other categories. However, DiPs are noticeably different from most of those elements: They show some similarities with the class of higher adverbs and with the class of focus particles. Their grammar is, nevertheless, quite different, as has often been noted: DiPs depend syntactically on sentence types and/or semantically on the speech act that a sentence type represents. DiPs are, however, extra-propositional themselves, in many ways. To give an example, consider rhetorical questions. In English, the question in (1) can be interpreted as a rhetorical question: (1) Who likes to be criticized? The expected answer Nobody! is already suggested by the question itself. The reason for this expectation is probably just world knowledge: Most people simply do not like to be criticized. However, there is no linguistic device in (1) which would enforce this interpretation. As a consequence, the more far-fetched interpretation as an information-seeking question is – at least technically – still possible. Compare (2): (2) Who wants to go shopping today? Here, it is not obvious at all that the question should be interpreted as a rhetorical question. It could be, technically, but such an interpretation seems to be rather far-fetched. Again, the interpretation depends on how we see the world: Since many people like to shop, the rhetorical question interpretation remains in the background.

Josef Bayer: Fachbereich Sprachwissenschaft, Universität Konstanz, Konstanz, e-mail: [email protected] Volker Struckmeier: Institut für Deutsche Sprache und Literatur I & Englisches Seminar I, Universität zu Köln, Köln, e-mail: [email protected]

DOI 10.1515/9783110497151-001

2 

 Josef Bayer and Volker Struckmeier

Consider now the following contrasts in German in which the b-examples involve the particle schon (lit. ‘already’). Wer wird gerne kritisiert? (3) a. who becomes gladly criticized b. Wer wird schon gerne kritisiert? who becomes SCHON gladly criticized (4) a. Wer will heute einkaufen? who wants today shop b. Wer will heute schon einkaufen? who wants today SCHON shop (3a) compares to the English example in (1): World knowledge may easily lead the listener to interpret it as a rhetorical question. (3b), however, only allows for the rhetorical interpretation. The DiP’s role becomes even more obvious in (4): A rhetorical interpretation of (4a) appears to be even more far-fetched than for the English example (2). (4a) – Who will go shopping today? – is quite clearly an information-seeking question. When schon is used, however, as in (4b), it turns unambiguously into a rhetorical question.1 This demonstrates that the DiP schon is a formal device in the formation of a special interrogative sentence type with a highly specific semantic interpretation. It seems, then, that DiPs do make contributions to the interpretation of sentences, e.g. in relation to sentence types – even though their semanto-pragmatic contributions are not always as easily described. It could be suggested that expressions like on earth or the fuck are English equivalents of certain DiPs in German. Although there are some interesting similarities, the two kinds of expressions do not seem to be completely equivalent. Consider questions as in (5). These can still be considered as information-seeking questions – albeit very emotional (or rude) ones: (5) a. Who on earth will go shopping today? b. Who the fuck will go shopping today? In the history of linguistics, DiPs have received relatively little attention. In some research traditions, they seem to have remained unidentified or ignored altogether. One reason for this may be that languages differ with respect to the occurrence of DiPs. German is known for its rich inventory of DiPs, whereas English

1 König and Siemund (2007) go as far as calling schon an “assertive particle”, obviously because it is thought to turn a question into a declarative.



The status quo of research on discourse particles in syntax and semantics  

 3

seems to lack them entirely, despite being a closely related language.2 Another reason may be that linguistics is still more in the grip of the grammar of written language than many of its practitioners would like to admit. Since DiPs are dependent on speech acts, they tend to be dominant in spoken language more than in certain written styles. Note, however, that DiPs are clearly attested in certain genres and styles of written language, such as drama, modern prose and quite generally texts that use direct speech. In this sense, they are certainly not to be discarded as some sort of performance phenomenon (such as an overuse of tags, hesitation markers, etc.). A third (and maybe more serious) reason may be that they have so far presented researchers with distinctions that do not always fit neatly into those categories we have come to employ in established frameworks in formal syntax and semantics. From this point of view, DiPs either seem to be “too complex” to deal with, or researchers believe they can ignore them because they belong to “performative” aspects of communication – rather than to language as a formal system (core grammar). As the present volume attempts to demonstrate, both of these views are probably untenable. More recently, linguists have developed theories, methods and techniques to describe and explain DiPs’ functions and formal properties in such a way as to integrate them into formal grammatical and compositional semantic frameworks more than ever before. The study of German DiPs has yielded an impressive body of work in a relatively short time. However in its earliest beginnings, this linguistic endeavor concentrated almost exclusively on the pragmatic or discourse functions of DiPs. Scientifically valid descriptions begin with the (quite insightful) remarks by von der Gabelentz ([1891] 1972). His observations later on became central to authors like Weydt (1969, 1977) who had a strong influence on the pragmatic turn of the 1970s. Strikingly, syntax and formal semantics were almost completely neglected in these early research contributions. While the choice to exclude syntactic and semantic properties was probably, to put it politely, very much in line with the overall aims of this line of research, it also must have seemed difficult, if not altogether impossible to the field in those times to actually get to grips with DiPs from a formal, syntactico-semantic point of view. According to Altmann’s (1980) review of Weydt (1977), there was a lack of satisfying semantic analysis which would be able to pick up on speech act theory and Gricean principles of conversation. Simultaneously, there was an increasingly critical evaluation of transformational

2 König and Siemund (2007) argue that basic sentence types are similarly modified across languages by intonation, while other devices are language-specific. The example they mention are modal particles in German.

4 

 Josef Bayer and Volker Struckmeier

grammar. One point of criticism was that it was supposed to be inadequate for the description of DiPs. Lacking the notion of functional categories, the syntax of the 1970s had to treat German DiPs as adverbs with peculiar properties, floating around in the sentence but refusing to move to peripheral positions (certainly leftwards, but in many cases also rightwards) the way that proper adverbs should. The DiPs’ rather limited word order options had to be captured by optional syntactic movements. The semantic effects of these alleged movement transformation was not an issue at this point in the development of syntactic theory, given that (post-base) transformations had to be meaning-preserving. Some of these shortcomings could be avoided, however, when syntactic theories became more restrictive and started to ask for possible reasons of movement operations – e.g., in later developments in Government and Binding (GB) Theory. According to the GB framework (and even more prominently in subsequent versions of the theory), syntactic objects move into new positions for a reason. Possible reasons now included both purely formal and possibly also semantic considerations: For example, theta-marked arguments move for reasons of Case, which in turn is responsible for theta-visibility. Wh-phrases move into a left-peripheral A’-position because they are semantic operators which require propositional scope – which is encoded syntactically, via movement. Head movement operations can be argued to serve the extension of a local domain, allowing for certain formal operations to stretch over larger domains than would otherwise be possible. However, why would DiPs want to move around in the clause? The syntax of the 1980s and 1990s had no good answers. In semantic research, the situation was not much better: The formal semantics of the 1970s was almost completely under the regime of model-theoretic truthconditional semantics. Since DiPs do not contribute to truth conditions, they were not in the focus of contemporary research questions as a matter of cause. Although there was a substantial interest in speech acts, it was generally unclear how utterance meaning could be composed. An interesting development in the syntactic research of these days was the performative analysis suggested by Ross (1970) and Sadock (1974).3 None of these works was connected to DiPs, however. The contribution of DiPs to sentence meaning was felt to be rather vague and therefore hard to grasp. As there was no leading idea about the mapping from syntax to speech acts, it appeared impossible to incorporate DiPs into

3 In the course of increasing research on the left clausal periphery, this line of research has recently seen a revival, cf. Speas and Tenny (2003), Haegeman and Hill (2013), Miyagawa (2010), among others.



The status quo of research on discourse particles in syntax and semantics  

 5

compositional meaning computations. This left the field of DiP usage to (sociology-oriented) discourse analysis and efforts of applied linguistics, such as attempts to help along the study of German as a second language. Since the actual formal and functional contributions of DiPs were, however, unclear at the time, these efforts were not met with great success: Learners of German as a second language must get an idea of how to handle DiPs, since they will encounter them in large numbers in the productions of native speakers. Sadly, however, the only “solution” which was occasionally offered was to hope that intuitions about the proper use of DiPs would develop “automatically”, i.e. without explicit training, with increasing German L2 proficiency. However, there also were attempts to tackle the problem in a more organized fashion: Useful contributions came especially from the work of Gerhard Helbig (cf. the useful overview in Helbig [1988]). Some of these descriptions laid the ground work for proper empirical investigations of DiPs, as we know now – but only in hindsight. Important research stimuli also came from the study of the diachronic development of DiPs (cf. Hentschel 1986; Abraham 1991; Wauchope 1991, among others). Virtually any part of speech which qualifies as a DiP has a counterpart from which it has developed historically, and which in many cases continue to coexist with the DiP. Examples would be English just (originally ‘morally upright’, ‘conforming to rules’) or German eben (lit. ‘flat’, ‘smooth’), nur (lit. ‘only’), schon (lit. ‘already’). DiPs are prime examples for processes of grammaticalization. More often than not, tracing their history leads to valuable insights about their synchronic characteristics – and can help to avoid the pitfall of mistaking the coexisting “doubles” for the DiPs. Interest in formal investigations of DiPs was inspired by early work on focus particles (e.g. Jackendoff 1972; Altmann 1976; Jacobs 1983). Doherty (1987) and Jacobs (1991) can be seen as early formal attempts to account for the semantics of DiPs. DiP syntax was not a big issue at first. However, one of the most influential and comprehensive works on German DiPs and their combinatorial properties is Thurmair (1989). To this day, much of the research on German DiPs relies heavily on the observations first compiled in her work. Inspired by Cinque’s (1999) exploration of the hierarchy of adverbs and his proposal that adverbs are generated in the specifiers of functional heads, Coniglio (2011) represented Thurmair’s generalizations in a cartographically organized hierarchy for German DiPs. Semantically, the syntactic hierarchy translates into scope relations. For many cases, however, it is not really clear how the order of DiPs can be explained in terms of scope. The reason for this state of affairs lies in the fact that the lexical meaning of DiPs is quite abstract and occasionally hard to pin down in a sufficiently precise way. It may also turn out that general points of criticism about strictly cartographic analyses of the adverb hierarchy (cf., e.g.,

6 

 Josef Bayer and Volker Struckmeier

Frey and Pittner 1998) carries over to the cartography of DiPs (cf., e.g., Struckmeier 2014a). Doherty (1987), Jacobs (1991), and Meibauer (1994) pointed out ways of accounting for discourse particles compositionally. Leading ideas suggest that DiPs share a core aspect of meaning with (occasionally truth-conditionally relevant) focus particles – but DiPs differ from focus particles fundamentally, too, in that they influence utterance-level properties rather than propositional, sentence-level meanings. A discourse-semantic approach helps address two important properties of DiPs: (i) their dependence on certain sentence types (cf. Abraham 1995) and (ii) their predominant occurrence in root-clauses. A valid rule of thumb seems to be that only root clauses are utterances because only they can encode the speech act carried out by the actual speaker (cf. modifications in Hooper and Thompson [1973], Heycock [2006], Krifka [2014]). Recent work by, e.g. Gutzmann (2015) and others, shows ways in which DiPs affect the expressive (i.e. “not-at-issue” or “use-conditional”) dimension in a multi-dimensional model of meaning. These developments extend earlier developments in formal semantics, aiming for a simultaneous representation of propositional meaning and implicatures (cf. Karttunen and Peters 1979). They provide a framework for the study of DiPs on a par with various other aspects of epistemic and emotional states of the speaker. It is easily overlooked, however, that beyond their semantic properties, DiPs also have a special syntactic status. Much work in formal semantics is still carried out under the (often tacit) assumption that DiPs are operators on a par with higher adverbs. In many cases, this belief leads authors to skip the discussion of DiP syntax altogether. For instance, in Karagjosova’s (2004) dissertation on the meaning of German DiPs, their constitutive role in clause structure is not even mentioned. While such accounts may yield proper descriptions of meaning, they fall short of explaining how exactly DiPs actually come to contribute to the observed meaning effects, in a compositionally computable way. To point out but one problem that is generally overlooked: DiPs very often do not occur in positions in the clause that one would be tempted to associate with the “left periphery”. In German and in various other languages. Rather, they occur in what is arguably a lower, potentially vP-related position.4 Given the standard assumption that clause type and illocutionary force belong to the upper left clausal periphery (cf. Rizzi [1997] and subsequent work on split CP), this appears to hamper a compositional analysis. Reconsider for concreteness (3) and (4). The question here is, by which (non-stipulative) process the DiP schon actually manages, from its structural position in the middle field, to modify the interrogative mood of

4 For the interaction of DiPs with information structure see Grosz (2016).



The status quo of research on discourse particles in syntax and semantics  

 7

the clause in such a way that the CP is interpreted as a rhetorical and not as a straightforward information-seeking question. Problems of this type have motivated logic-oriented approaches to argue for movement to such a position in terms of LF movement (cf. Abraham 1995, 2012; Zimmermann 2004, 2008). Such a move seems promising in some respects but also introduces new problems, especially with regard to new advances in syntactic theorizing. Firstly, the development of syntax within the Minimalist Program and especially the admission of more functional projections has provided new ways to describe how DiPs contribute to utterance meanings. Rizzi’s (1997) work on the fine structure of the left periphery can be seen as an important inspiration for proposals that attempt to integrate DiPs into a split C scenario. However, on the other hand, the nature of the mechanism these proposals seek to employ is far from clear. As Bayer, Häussler, and Bader (2016) show, LF movement of DiPs leads to wrong predictions about their scope; they also show that DiPs in complement clauses would have to raise across CP – an unattested LF-operation, as far as we know. As already pointed out above, the kinds of semantic accounts of DiPs which do not pay special attention to the syntax (and compositional semantics) of DiPs take them to be adverbs. Various properties of DiPs, however, raise doubts about the appropriateness of this assumption. In fact, contemporary models of syntax cast doubt on the structural status of DiPs: If DiPs are adverbs, they are maximal structural objects, i.e. phrases. Phrases, in German, however, can be moved to the sentence-initial position, or they can occasionally also be extraposed. Both topicalization and extraposition sharply differentiate between adverb phrases and DiPs: These peripheral movements can be observed with adverbs – but never with DiPs.5 DiPs, it seems, are completely “frozen” in place: They almost never leave their middle field-internal position. This fact, of course, follows immediately if DiPs are heads, not phrases: Functional heads, such as T or C, show exactly the same kind of behavior, even in free word order languages like German. There is additional evidence for the head status of DiPs (cf. Meibauer 1994;

5  (i)  vielleicht as an adverb a. Renate ist vielleicht in die Oper gegangen Renate is perhaps in the opera gone   ‘Perhaps, Renate went to the opera.’ b. Vielleicht ist Renate in die Oper gegangen. / Renate ist in die Oper gegangen vielleicht. (ii) vielleicht as a DiP a. DIE ist vielleicht eingebildet! she is VIELLEICHT arrogant ‘Boy, is SHE arrogant!’ b. *Vielleicht ist DIE eingebildet! / *DIE ist eingebildet vielleicht!

8 

 Josef Bayer and Volker Struckmeier

Coniglio 2005, 2011; Bayer and Obenauer 2011; Bayer 2012; Struckmeier 2014a; Bayer and Trotzke 2015, among others). Still, the issue has remained controversial so far. Certain DiPs (similarly to focus particles and expressions of negation) are syncategorematic and therefore seem to lend themselves more to a (phrasal) modifier than to a head analysis. With respect to German verb movement to C (yielding the well-known V2 order), it has repeatedly been argued that DiPs cannot be heads: If DiPs were heads, the argument goes, they would be predicted to act as interveners when finite verbs move across the DiP (head) position, on their way to C. V2-clauses and V1-clauses are, however, often root clauses – and thus, exactly the kinds of clauses that DiPs are attested in most often. This would seem to lend support to the idea that DiPs are phrasal elements, as e.g. proposed by Cardinaletti (2007, 2011). However, note that the idea that V2 movement would be barred by intervening heads can be called into question itself. There definitely exist alternative analyses that could derive the V2 property without problems with intervention by DiPs or negation.6 In strictly head final languages, we observe that DiPs are often clitic-like elements in clause-final positions or also phrase-final elements which attach to major constituents. We can refer here to work on Japanese (cf. Endo 2012; Saito & Haraguchi 2012), Bangla (a.k.a. Bengali) (cf. Dasgupta 1980, 1984, 1987, 2005; Bayer et al. 2014), and Hindi (cf. Shapiro 1999; Montaut 2002, 2012). In these languages, DiPs are generally on a par with “regular” functional heads such as interrogative particles. The evidence for their head status is therefore straightforward. Work on sentence-final DiPs in Chinese, which is typologically not head-final, and their relation to the clausal periphery can be found in Li (2006). One of the contributions to the present volume, the chapter by Pan and Paul, takes up this topic again. This edited volume aims to provide contributions that add to the newly developing field: Some of the chapters collected here provide analyses of languages, diachronic stages of languages, or particle properties, which, we hope, will help

6 Since in V2 phenomena, only finite verbs move, the process could be analyzed as feature movement. Given that features are not categories, an X° DiP or an X° element of negation would not be an intervener. Bayer and Obenauer (2011) propose a distinction in terms of Relativized Minimality. Alternatively, it has been argued that the V2 property is actually caused by remnant vP-movement (Müller 2003). Similarly, Struckmeier (2014b, 2016) proposes that vP as a whole, a phrasal constituent, can move across DiPs in the German clause. Verbs contained in vP could thus be “smuggled” past an intervening DiP head in the clause. Head movement, such as in V2, has also been argued to be an entirely extra-syntactic displacement operation (Chomsky 1999). Thus, it might well turn out that the argument from intervention against the head status of DiPs lose their force.



The status quo of research on discourse particles in syntax and semantics  

 9

avoid a view of DiPs that is based too narrowly on, say, modern-day German DiPs. Since roughly a decade, there is increasing work on DiPs in Romance languages, in particular in Northern Italian dialects, mainly Alto Adige (South Tyrol) and the Veneto (cf. Poletto 2000; Munaro and Poletto 2003, 2004, 2005; Coniglio 2008; Hack 2009; Cardinaletti 2015; Hinterhölzl and Munaro 2015; Manzini 2015). So far, however, there has been limited work in formal linguistics on particles in languages related to German (Dutch, cf. de Vriendt, Vandeweghe, and Van de Craen [1991], Vismans [1994], Foolen [1995]; Scandinavian, cf. Aijmer [1996] for Swedish, Davidsen-Nielsen [1996] for Danish). The English tradition speaks mainly of “discourse markers”, (cf. Fraser 1999 and Blakemore 2004) or “pragmatic markers” (Aijmer, Foolen, and Simon-Vandenbergen 2006). These include elements with occasionally rather heterogeneous properties. For comparative work on English/German see Schubiger (1965), for Dutch/German see Foolen (2006). Jónsson, in his contribution to this volume, provides an interesting analysis of Icelandic particles in special exclamative sentence types, contributing to the cross-linguistic description of particles. Scherf, in this volume, analyses Swedish modal particles and proposes that the set of DiPs in this related language subdivides into classes that warrant a different syntactic treatment. Petrova, in her article to the current publication, analyses particles in Old High German, and reaches the (perhaps surprising) conclusion that despite many superficial similarities, there is no continuous development even of those DiPs that seem to have straightforward equivalents in later diachronic stages of German. Abraham, in his contribution to this volume, compares German DiPs with comparable elements from other languages, and proposes a syntacto-semantic analysis for the different cases. Lühr, in this volume, provides a diachronic description of particles in Old Indic, also arriving at different subclasses of elements that need to be carefully kept apart. Paul and Pan, in their contribution to this volume, present convincing evidence that “what you see is what you get” in Mandarin Chinese: Sentence-final particles in this language constitute the high functional heads syntactically that their semanto-pragmatic function requires, leading to an interesting conflict with word order constraints, such as the Final-over-Final constraint. It is, we believe, a healthy development to see research on DiPs cover languages other than German more and more: This will greatly broaden our views of the possible phenomena and their dependency on typological and parametric properties. The field of linguistics is now in a much better position to study DiPs than it has been in the first wave of intensive work in the 1970s and 1980s. Both in the syntax and in the semantics of DiPs, we are also in a position to ask much more meaningful and detailed questions, and provide clearer answers than previously possible. In semantics, e.g., it is the inclusion of the expressive dimension of meaning in formal research and the acknowledgment of

10 

 Josef Bayer and Volker Struckmeier

multi-dimensionality of meaning. In syntax, it is the more finely calibrated clause structure, and the more sophisticated view of movement, agreement and related dependencies. In both fields, DiPs seem to be able to take a more central and more core-grammatical role than previously assumed. At least to some degree, we now have strategies, techniques and theoretical models that allow, at the very least, new investigations to be carried out. The volume at hand seeks to support this field of research: Gärtner, in his article, provides a syntactico-semantic analysis that rests on observations regarding the behavior of certain DiPs in root infinitivals. Gutzmann, in this volume, points out that uniform treatment of various DiPs in terms of their semantics may not be warranted, given the data he discusses. Müller, in her contribution to the volume, argues that the relative order of DiPs may change in German – contrary to many claims to the contrary. She argues that changes in DiP order reflect different contributions of DiPs to the utterances they occur in, in their discursive context. Egg and Mursell, in their article in this volume, take a more syntactic and compositional-semantic view: They argue that DiPs may be involved in more syntactic operations than previously assumed, which may help us understand how their semantic contributions to sentence types may come to pass. Last, but not least, Viesel contributes an argument to the volume that investigates attributive structures, which can also contain DiPs. From these special syntactic contexts, she argues, we can gain new insights about the compositional contribution of DiPs in clause structure. We must stress here that research on DiPs in formal syntax and semantics has barely just begun – and uncontroversially so, since research on DiPs is connected to formal theoretical and comparative work in linguistics, which is only just know coming into reach. We hope, however, that from the new beginnings the authors in this volume present, this exciting young field will be able to draw some inspiration for the work that will, without a shadow of doubt, be necessary in the future. We would like to thank the Deutsche Forschungsgemeinschaft / German Research Foundation for grant BA 1178/9-1 Clause Structure and Utterance Meaning – Word Order, Particles and Emphasis, which helped us to organize the workshop The Syntax and Semantics of Particles at the 36th annual conference of the Deutsche Gesellschaft für Sprachwissenschaft / German Linguistic Society (Marburg, March 2014). Thanks also to Klaus von Heusinger for his offer to publish selected presentations from this workshop in the Linguistische Arbeiten series. Klaus was also helpful in organizing the review process. Many initial ideas for the workshop and the book benefitted very much from Andreas Trotzke’s thoughts and his great help, which is gratefully acknowledged here. In its final stage, the completion of this volume benefitted greatly from the careful and professional copy-editing of Susanne Trissler and her interaction with all the contributors. We thank you all very much!



The status quo of research on discourse particles in syntax and semantics  

 11

References Abraham, Werner. 1991. The grammaticization of the German modal particles. In Elizabeth Closs Traugott & Bernd Heine (eds.), Approaches to grammaticalization, Vol. II: Types of grammatical markers, 331–380. Amsterdam: Benjamins. Abraham, Werner. 1995. Wieso stehen nicht alle Modalpartikel in allen Satzformen? Zeitschrift für deutsche Sprache 23(2). 124–146. Abraham, Werner. 2012. Illocutive force is speaker and information source concern. What type of syntax does the representation of speaker deixis require? Templates vs. derivational structure? In Werner Abraham & Elisabeth Leiss (eds.), Modality and theory of mind elements across languages, 67–108. Berlin: de Gruyter Mouton. Aijmer, Karin. 1996. Swedish modal particles in a contrastive perspective. Language Sciences 18(1–2). 393–427. Aijmer, Karin, Ad Foolen & Anne-Marie Simon-Vandenbergen. 2006. Pragmatic markers in translation: A methodological proposal. In Kerstin Fischer (ed.), Approaches to discourse particles, 101–114. Leiden: Brill. Altmann, Hans. 1976. Die Gradpartikeln des Deutschen. Tübingen: Niemeyer. Altmann, Hans. 1980. Review of Haraldt Weydt, Aspekte der Modalpartikeln. Studien zur deutschen Abtönung (1977). Beiträge zur Geschichte der deutschen Sprache und Literatur (PBB) 102. 84–90. Bayer, Josef. 2012. From modal particle to interrogative marker. A study of German denn. In Laura Brugé, Anna Cardinaletti, Giuliana Giusti & Nicola Munaro (eds.), Functional heads, 13–36. Oxford: Oxford University Press. Bayer, Josef & Hans-Georg Obenauer. 2011. Discourse particles, clause structure, and question types. The Linguistic Review 28(4). 449–491. Bayer, Josef & Andreas Trotzke. 2015. The derivation and interpretation of left peripheral discourse particles. In Josef Bayer, Roland Hinterhölzl & Andreas Trotzke (eds.), Discourseoriented syntax, 13–40. Amsterdam: Benjamins. Bayer, Josef, Jana Häussler & Markus Bader. 2016. A new diagnostic for cyclic wh-movement. Discourse particles in German questions. Linguistic Inquiry 47(4). 591–629. Bayer, Josef, Probal Dasgupta, Sibansu Mukhopadhyay & Rajat Ghosh. 2014. Functional structure and the Bangla discourse particle to. Paper presented at the 30th South Asian Languages Analysis Roundtable (SALA 30), Hyderabad, February 6–8, 2014. http://ling.uni-konstanz. de/pages/StructureUtterance/web/Events_files/Bayer_Dasgupta_MukhopadhyayGhosh_ SALA.pdf (accessed 8 June 2016). Blakemore, Diane. 2004. Discourse markers. In Laurence R. Horn & Gregory Ward (eds.), The handbook of pragmatics, 221–240. Oxford: Blackwell Publishing. Chomsky, Noam. 1999. Derivation by phase (MIT Occasional Papers in Linguistics 18). Cambridge, MA: MITWPL. [Reprinted in Michael Kenstowicz (ed.), Ken Hale: A life in language, 1–52. Cambridge, MA: The MIT Press, 2001]. Cardinaletti, Anna. 2007. Für eine syntaktische Analyse von Modalpartikeln. In Eva-Maria Thüne & Franca Ortu (eds.), Gesprochene Sprache – Partikeln. Beiträge der Arbeitsgruppen der 2. Tagung Deutsche Sprachwissenschaft in Italien, Rom 2006, 89–101. Frankfurt/M.: Peter Lang. Cardinaletti, Anna. 2011. German and Italian modal particles and clause structure. The Linguistic Review 28(4). 493–531. Cardinaletti, Anna. 2015. Italian verb-based discourse particles in a comparative perspective. In Josef Bayer, Roland Hinterhölzl & Andreas Trotzke (eds.), Discourse-oriented syntax, 71–91. Amsterdam: Benjamins.

12 

 Josef Bayer and Volker Struckmeier

Cinque, Guglielmo. 1999. Adverbs and functional heads. Oxford: Oxford University Press. Coniglio, Marco. 2005. Deutsche Modalpartikeln: Eine syntaktische Analyse. Venezia: Università Ca’ Foscari MA thesis. Coniglio, Marco. 2008. Modal particles in Italian. University of Venice Working Papers in Linguistics 18, 91–129. Coniglio, Marco. 2011. Die Syntax der deutschen Modalpartikeln: Ihre Distribution und Lizenzierung in Haupt- und Nebensätzen. Berlin: Akademie Verlag. Dasgupta, Probal. 1980. Questions and relative and complement clauses in a Bangla grammar. New York: NYU dissertation. Dasgupta, Probal. 1984. Bangla emphasizers and anchors. Indian Linguistics 45. 102–117. Dasgupta, Probal. 1987. Sentence particles in Bangla. In Elena Bashir, Madhav M. Deshpande & Peter E. Hook (eds.), Selected papers from SALA 7, 49–75. Bloomington, IN: Indiana University Linguistics Club. Dasgupta, Probal. 2005. Q-baa and Bangla clause structure. In Rajendra Singh (ed.), The yearbook of South Asian languages and linguistics 2005, 45–81. Berlin: de Gruyter. Davidsen-Nielsen, Nils. 1996. Discourse particles in Danish. In Elisabeth Engberg-Pedersen, Michael Fortescue, Peter Harder, Lars Heltoft & Lisbeth Falster Jakobsen (eds.), Content, expression and structure – studies in Danish Functional Grammar, 283–314. Amsterdam: Benjamins. de Vriendt, Sera, Willy Vandeweghe & Piet Van de Craen. 1991. Combinatorial aspects of modal particles in Dutch. Multilingua 10(1/2). 43–59. Doherty, Monika. 1987. Epistemic meaning. Berlin: Springer. [Engl. version of Epistemische Bedeutung. Berlin: Akademie Verlag, 1985]. Endo, Yoshio. 2012. The syntax-discourse interface in adverbial clauses. In Lobke Aelbrecht, Liliane Haegeman & Rachel Nye (eds.), Main clause phenomena. New horizons, 365–384. Amsterdam: Benjamins. Foolen, Ad. 1995. Dutch modal particles: The relevance of grammaticalized elements. In Thomas F. Shannon & Johan P. Snapper (eds.), The Berkeley Conference on Dutch Linguistics 1993, 57–70. Lanham, MD: University of America Press. Foolen, Ad. 2006. Polysemy patterns in contrast: The case of Dutch toch and German doch. In Karin Aijmer & Anne-Marie Simon-Vandenbergen (eds.), Pragmatic markers in contrast, 59–72. Amsterdam: Elsevier. Fraser, Bruce. 1999. What are discourse markers? Journal of Pragmatics 31(7). 931–952. Frey, Werner & Karin Pittner. 1998. Zur Positionierung der Adverbiale im deutschen Mittelfeld. Linguistische Berichte 176. 489–534. von der Gabelentz, Georg. 1972 [1891]. Die Sprachwissenschaft, ihre Aufgaben, Methoden und bisherige Ergebnisse. Zweite, vermehrte und verbesserte Auflage, herausgegeben von Albrecht Graf von der Schulenburg. Leipzig: Tauchnitz, 1901. Reprint Tübingen: Narr, 1972. Grosz, Patrick. 2016. Information structure and discourse particles. In Caroline Féry & Shinichiro Ishihara (eds.), The Oxford handbook of information structure. Oxford: Oxford University Press. Gutzmann, Daniel. 2015. Use-conditional meaning. Studies in multidimensional semantics. Oxford: Oxford University Press. Hack, Franziska. 2009. Syntactic variation in interrogatives in Romance languages and dialects. Konstanz: University of Konstanz MA thesis. Haegeman, Liliane & Virginia Hill. 2013. The syntactization of discourse. In Raffaella Folli, Christina Sevdali & Robert Truswell (eds.), Syntax and its limits, 370–390. Oxford: Oxford University Press.



The status quo of research on discourse particles in syntax and semantics  

 13

Helbig, Gerhard. 1988. Lexikon deutscher Partikeln. 1st edn. Leipzig: Verlag Enzyklopädie. Hentschel, Elke. 1986. Funktion und Geschichte deutscher Partikeln. Ja, doch, halt und eben. Tübingen: Niemeyer. Heycock, Caroline. 2006. Embedded root phenomena. In Martin Everaert & Henk van Riemsdijk (eds.), The Blackwell companion to syntax. Vol. II, 174–209. Malden, MA: Blackwell. DOI: http://dx.doi.org/10.1002/9780470996591.ch23 (accessed 8 June 2016). Hinterhölzl, Roland & Nicola Munaro. 2015. On the interpretation of modal particles in non-assertive speech acts in German and Bellunese. In Josef Bayer, Roland Hinterhölzl & Andreas Trotzke (eds.), Discourse-oriented syntax, 41–70. Amsterdam: Benjamins. Hooper, Joan & Sandra Thompson. 1973. On the applicability of root transformations. Linguistic Inquiry 4(4). 465–497. Jackendoff, Ray. 1972. Semantic interpretation in Generative Grammar. Cambridge, MA: The MIT Press. Jacobs, Joachim. 1983. Fokus und Skalen. Zur Syntax und Semantik der Gradpartikeln im Deutschen. Tübingen: Niemeyer. Jacobs, Joachim. 1991. On the semantics of modal particles. In Werner Abraham (ed.), Discourse particles, 141–162. Amsterdam: Benjamins. Karagjosova, Elena. 2004. The meaning and function of German modal particles. Saarbrücken: Saarland University dissertation. Karttunen, Lauri & Stanley Peters. 1979. Conventional implicature. In Choon-Kyu Oh & David A. Dinneen (eds.), Syntax and semantics 11: Presupposition, 1–56. New York: Academic Press. König, Ekkehard & Peter Siemund. 2007. Speech act distinctions in grammar. In Timothy Shopen (ed.), Language typology and syntactic description, 276–324. Cambridge: Cambridge University Press. Krifka, Manfred. 2014. Embedding illocutionary acts. In Margaret Speas & Thomas Roeper (eds.), Recursion: Complexity in cognition, 59–87. Berlin: Springer. Li, Boya. 2006. Chinese final particles and the syntax of the periphery. Leiden: University of Leiden dissertation. Manzini, M. Rita. 2015. Italian adverbs and discourse particles: Between recategorization and ambiguity. In Josef Bayer, Roland Hinterhölzl & Andreas Trotzke (eds.), Discourse-oriented syntax, 71–92. Amsterdam: Benjamins. Meibauer, Jörg. 1994. Modaler Kontrast und konzeptuelle Verschiebung: Studien zur Syntax und Semantik deutscher Modalpartikeln. Tübingen: Niemeyer. Miyagawa, Shigeru. 2010. Why agree? Why move?: Unifying agreement-based and discourseconfigurational languages. Cambridge, MA: The MIT Press. Montaut, Annie. 2002. La particule énonciative to en Hindi/Ourdou. Cahiers de Linguistique de l’INALCO 4. 111–134. Montaut, Annie. 2014. The discourse particle to and word ordering in Hindi: From grammar to discourse. In M. Jocelyne Fernandez-Vest & Robert D. Van Valin (eds.), Information structuring of spoken language from a cross-linguistic perspective, 263–282. Berlin: de Gruyter. Müller, Gereon. 2003. Verb-second as vP-first. In Anoop Kumar Mahajan (ed.), Syntax at sunset 3. Head movement and syntactic theory (UCLA Working Papers in Linguistics 10). 116–160. Los Angeles, CA: University of California Press. Munaro, Nicola & Cecilia Poletto. 2003. Ways of clausal typing. Rivista di Grammatica Generativa 27. 87–105. Munaro, Nicola & Cecilia Poletto. 2004. Sentential particles and clausal typing in the Veneto dialects. In Benjamin Shaer, Werner Frey, & Claudia Maienborn (eds.), Proceedings of the

14 

 Josef Bayer and Volker Struckmeier

Dislocated Elements Workshop, ZAS Berlin, November 2003 (ZAS Papers in Linguistics 35, vol. 2), 375–397. Berlin: Zentrum für Allgemeine Sprachwissenschaft (ZAS). Munaro, Nicola & Cecilia Poletto. 2005. On the diachronic origin of sentential particles in ­North-Eastern Italian dialects. Nordic Journal of Linguistics 28(2). 247–267. Poletto, Cecilia. 2000. The higher functional field. Oxford: Oxford University Press. Rizzi, Luigi. 1997. The fine structure of the left periphery. In Liliane Haegeman (ed.), Elements of grammar, 281–337. Dordrecht: Kluwer. Ross, John R. 1970. On declarative sentences. In Roderick A. Jacobs & Peter S. Rosenbaum (eds.), Readings in English Transformational Grammar, 222–272. Waltham, MA: Ginn & Co. Sadock, Jerrold M. 1974. Towards a linguistic theory of speech acts. New York: Academic Press. Saito, Mamoru & Tomoko Haraguchi. 2012. Deriving the cartography of the Japanese right periphery: The case of sentence-final discourse particles. IBERIA: An International Journal of Theoretical Linguistics 4(2). 104–123. Schubiger, Maria. 1965. English intonation and German modal particles – a comparative study. Phonetica 12(2). 65–84. Shapiro, Michael. 1999. Hindi to as discourse marker. In Purushottam J. Mistry & Bharati Modi (eds.), Vidyopaasanaa: Studies in honor of Harivallabh C. Bhayani, 179–189. Mumbai & Ahmedabad: Image Publications Pvt. Ltd. Speas, Margaret & Carol Tenny. 2003. Configurational properties of point of view roles. In Anna-Maria Di Sciullo (ed.), Asymmetry in grammar, 315–344. Amsterdam: Benjamins. Struckmeier, Volker. 2014a. Ja doch wohl C? Modal particles as C-related elements. Studia Linguistica 68(1). 16–48. Struckmeier, Volker. 2014b. Scrambling ohne Informationsstruktur: Zur komplexen Architektur grammatischer Beschreibungen. Berlin: de Gruyter. Struckmeier, Volker. 2016. Scrambling in German is driven by prosody and semantics. In Kyeong-min Kim, Pocholo Umbal, Trevor Block, Queenie Chan, Tanie Cheng, Kelli Finney, Mara Katz, Sophie Nickel-Thompson & Lisa Shorten (eds.), Proceedings of the 33rd West Coast Conference on Formal Linguistics, 381–389. Somerville, MA: Cascadilla Proceedings Project.  Thurmair, Maria. 1989. Modalpartikeln und ihre Kombinationen. Tübingen: Niemeyer. Vismans, Roel. 1994. Modal particles in Dutch directives: A study in Functional Grammar. Amsterdam: IFOTT. Wauchope, Mary Michele. 1991. The grammar of the Old High German modal particles thoh, ia, and thanne. Frankfurt/M.: Peter Lang. Weydt, Harald. 1969. Abtönungspartikel. Die deutschen Modalwörter und ihre französischen Entsprechungen. Bad Homburg: Gehlen. Weydt, Harald (ed.). 1977. Aspekte der Modalpartikeln. Studien zur deutschen Abtönung. Tübingen: Niemeyer. Zimmermann, Malte. 2004. Zum Wohl: Diskurspartikeln als Satztypmodifikatoren. Linguistische Berichte 199. 1–35. Zimmermann, Malte. 2008. Discourse particles in the left periphery. In Benjamin Shaer, Philippa Cook, Werner Frey & Claudia Maienborn (eds.), Dislocated elements in discourse, 200–231. Oxford: Routledge.

Markus Egg and Johannes Mursell

The syntax and semantics of discourse particles We offer a compositional analysis of the way in which the semantic contribution of discourse particles is integrated into the interpretation of their host sentences at the syntax-semantics interface. The particles interact syntactically both with the left periphery of the sentence, thus modifying its illocutionary force by changing its felicity conditions, but also with its vP via a focus feature.

1 Introduction Among the many interesting particles that can be found in natural ­languages, discourse particles, sometimes also called modal particles (in German, ­ ­Diskurspartikel or Modalpartikel; DiPs, for short1) still belong to the most puzzling ones, even though they have received considerable attention in the literature. Much recent work on discourse particles tries to account for the contribution of DiPs to their host utterances in terms of a modification of the illocutionary type of the utterance (Jacobs 1991) or of its illocutionary force (Zimmermann 2004b). The syntactic ramifications of this idea were worked out in the framework of Minimalism (Chomsky 1995), most recently, in Bayer and Obenauer (2011) and Coniglio and Zegrean (2012). These approaches both assume that the particles enter several complex agree relations with the left periphery of

1 We prefer the former term because these particles do not introduce “modality” in the standard sense (which could be modelled in terms of possible-world semantics). Acknowledgements: The authors would like to thank the participants of the workshop on particles at the annual DGfS meeting in 2014 for helpful discussions, in addition, to Nathalie Scherf, Sophia Döring, Beate Bergmann, Josef Bayer, Volker Struckmeier, and two anonymous reviewers. Any remaining errors are of course our own. Markus Egg: Institut für Anglistik und Amerikanistik, Humboldt-Universität zu Berlin, Berlin, e-mail: [email protected] Johannes Mursell: Institut für Linguistik, Goethe-Universität Frankfurt, Frankfurt/M., e-mail: [email protected] DOI 10.1515/9783110497151-002

16 

 Markus Egg and Johannes Mursell

the clause, more precisely with the illocutionary force component in a split CP ­following Rizzi (1997). In this paper, we combine the two strands of research to provide a comprehensive compositional analysis of how the contribution of discourse particles is integrated into their host utterance. On the one hand, we analyse DiPs as modifiers of illocutionary force following much recent and older work (Jacobs 1991; Zimmermann 2004a, among others). To make illocutionary force syntactically accessible, we assume that it heads the highest projections in the left periphery and can therefore be modified by syntax-internal processes (Bayer and Obenauer 2011; Coniglio and Zegrean 2012). On the other hand, we propose that discourse particles interact with the information structure of the clause in a far more complex way than previously assumed (Egg and Zimmermann 2012). In older approaches, the interaction between information structure and DiPs has often been reduced to simply marking the boundary between old and new information, between theme and rheme, at the edge of the vP (cf. Hentschel 1986 and Diesing 1992). We will show that this is not the only way in which DiPs are connected to information structure, since they also interact with a focussed constituent in their c-command domain. This interaction between discourse particles and information structure will be modelled via an agree mechanism based on a focus feature and provides a way to connect the DiP to the left periphery. In this way, discourse particles can be connected syntactically to illocutionary force and the actual modification of the illocutionary force is modelled as the negation of specific felicity conditions. The paper is structured as follows. We will first discuss the way in which DiPs interact with the illocutionary force in the left periphery of the clause. Building on Vanderveken (1990), this interaction will be analysed as a negation of felicity conditions. Then we will set the theoretical syntactic background for the ensuing discussion, introducing the split ForceP, multiple agree, and two earlier syntactic approaches. Based on these mechanisms, an analysis is developed in the following section connecting vP, DiP and illocutionary force via agree based on focus features. Before we set out to present our theory however, we will briefly introduce properties of DiPs which are important for the ­discussion.

2 Properties of discourse particles In this section, we will briefly discuss some general properties of discourse particles that bear upon the analysis to be developed, viz. their distribution and their contribution to the clause.

The syntax and semantics of discourse particles  



 17

2.1 Syntactic distribution DiPs are typically restricted to specific clause types and even though all clause types can host DiPs, none can host all, while, conversely, no DiP can occur in all clause types. Tab. 1: Distribution of three DiPs across clause types (Thurmair 1989: 49)

Declarative y/n-Question wh-Question Imperative Optative Exclamative wh-Exlamative

ja

wohl

doch

+ – – – – – –

+ + + – – – –

+ – + + + – +

The examples in (1) illustrate the restrictions indicated in Tab. 1 together with a paraphrase for the intended interpretation. (1a) illustrates that wohl cannot occur in imperatives, (1b)–(1d) show the ban on ja in yes-no questions, ­wh-questions and imperatives, respectively, while (1e) exemplifies the observation that doch cannot occur in yes-no questions. (1) a. *Gib mir wohl bitte den Stift! give me wohl please the pen int.: ‘Give me the pen please!’ b. *Trinkt Christoph ja gerne Bier? drinks C. ja like.to beer int.: ‘Does C. like to drink beer?’ c. *Was trinkt Christoph ja gerne? what drinks C. ja like.to int.: What does C. like to drink?’ d. *Gib mir ja bitte den Stift! give me ja please the pen int.: ‘Give me the pen, please!’ e. *Ist doch Berlin die Hauptstadt von Deutschland? is doch Berlin the capital of Germany int.: ‘Is Berlin the capital of Germany?’ However, clause type is not the only factor that restricts the occurrence of DiPs. E.g., in (2), denn is incompatible with the most prominent reading of the sentence

18 

 Markus Egg and Johannes Mursell

as a request (i.e., as an indirect speech act) and is only possible in a literal question reading: (2) #Kannst du denn das Fenster schließen? can you denn the window close int.:‘Could you close the window?’ Concerning the level of intrasentential distribution, it has been claimed that DiPs are restricted to the middle field (German Mittelfeld) of the clause.2 At the same time, it has been observed that DiPs, under most circumstances, take scope over the whole proposition (cf. Jacobs 1991; Ormelius-Sandblom 1997). However, this statement needs further qualification, since, as Bayer and Obenauer (2011) show, specific topical elements (definites, generic indefinites, etc.) can move around the immobile DiP, which marks the border of the vP/VP (Diesing 1992). In the examples in (3), the DiP is nur: (3) a. Wann könnte nur Otto den Brief gestern ins Büro when could NUR Otto the letter yesterday to office mitgenommen haben? along-taken have ‘When could Otto have yesterday taken the letter to the office? (I’ve found no answer so far.)’ b. Wann könnte Otto nur Otto den Brief gestern ins Büro mitgenommen haben? c. Wann könnte Otto den Brief nur Otto den Brief gestern ins Büro mitgenommen haben? d. Wann könnte Otto den Brief gestern nur Otto den Brief gestern ins Büro mitgenommen haben? e. Wann könnte Otto den Brief gestern ins Büro nur Otto den Brief gestern ins Büro mitgenommen haben?

2 The more general claim that DiPs are restricted to languages with a dedicated middle field Abraham (1991a, 1991b), however, is in conflict with Italian and French, which both have ­discourse particles but no middle field: (i) Deve aver pur letto il libro. 3sg.must have pur read the book ‘He must have read the book, I think.’ (Coniglio 2011: 37) (ii) Je sais pas quand même, les client, ce I know not quand même the clients that ‘I don’t actually know what the clients want.’

que what

ils they

veulent. want



The syntax and semantics of discourse particles  

 19

Information structure generally seems to play an important role with respect to the ordering of DiPs and other elements in the clause. As Bayer and Obenauer (2011) point out, DiPs require information focus to their right, and focussed ­elements to the left of the DiP have to be interpreted contrastively. (4) A: ‘What did you eat yesterday?’ B: Ich hatte doch [PASTA]F. I had DOCH pasta ‘I had pasta. Don’t you remember?’ (normal answer) B’: [PASTA]F hatte ich doch (und nicht den ekligen Fisch). Pasta had I DOCH and not the disgusting fish ‘It was pasta I had, and not the disgusting fish. Don’t you remember?’ (contrastive interpretation) However, there are counterexamples to this generalisation, mostly concerning occurrences of the particle wohl in examples like (5), in which the focus on the DP preceding wohl is not necessarily contrastive.3 (5) [PETER]F trinkt wohl den Wein. Peter drinks WOHL the wine ‘Someone is drinking the wine, and I think it’s Peter.’ However, this syntactic restriction of the position of discourse particles is unexpected from a semantic point of view if discourse particles are analysed as modifiers of illocutionary force (cf. Jacobs 1991; Waltereit 2001; Zimmermann 2004b;

3 The focussed element is actually merged below the particle and then moves to a higher position due to independent processes. Concerning the argument made below that DiPs associate with a focussed constituent, it needs to be pointed out that not all focus sensitive particles require the focussed element to be in their scope on the surface. As observed by Erlewine (2014), only focus sensitive particles that have a truth conditional impact need to have surface scope over the associated focus. Thus, while only requires the focussed element in its c-command domain, this is not the case for even, because the latter only introduces a presupposition and does not affect the truth conditions. (i) a. [JOHN]F , Mary even met at the party. b. *[JOHN]F , Mary only met at the party. (Erlewine 2014: 11) Consequently, since, as pointed out above, the contribution of DiPs is also not truth-conditional, they are not expected to necessarily take surface scope over the associated focussed element, similar to focus sensitive particles like even. A question which we will not try to answer in this paper is whether this analysis can be extended to sentence final focus sensitive or discourse particles.

20 

 Markus Egg and Johannes Mursell

Bayer and Obenauer 2011; Coniglio and Zegrean 2012; Egg 2013; among others): since illocutionary force, following Rizzi (1997), is encoded in the highest functional projection of the clause, ForceP, discourse particles are expected to occur in the periphery of the clause and not in its middle field (Zimmermann 2004a). Cross-linguistic data show that this expectation is borne out in languages like Japanese, which has sentence final discourse particles (consistent with the assumption that Japanese is strictly head final).4 (6) Are-o mi-ro-yo5 that-ACC look-IMP-PAR ‘(You should really) Look at that!’ (Endo 2012: 2) And even in German, DiPs can be found outside the middle field in certain exceptional cases. In three constructions, it is possible for the DiP to move to the prefield (German Vorfeld) as part of a larger constituent, viz., as part of DPs (7a) or of wh-phrases (7b), or even together with an adverb (7c).6 (7) a. Dieser ja leider viel zu früh verstorbene this prt unfortunately much too early died Komponist hat uns eine Reihe von großartigen composer has us a series of great Werken hinterlassen. works left.behind ‘This composer, who unfortunately died much too early (according to common opinion), has left behind many amazing works for us.’ (Thurmair 1989: 27) b. Wer denn soll befehlen? who PRT should command ‘Who should command (please tell me)?’ (Bayer and Obenauer 2011: 471) c. Vielleicht ja werden späterhin nur mehr Satzteile maybe prt will later only more sentence.parts und Wortfetzen über den Äther schweben. and word.chunks across the ether float ‘Maybe, later on, there will only be sentence parts and word chunks floating across the ether.’

4 For sentence peripheral discourse particles in older stages of German, see Petrova (2017). 5 The following abbreviations are used in glossing examples: acc accusative; imp imperative mood; prt particle; 3sg third person singular. 6 Thanks to a reviewer for this example from Die Zeit, 1988-10-28.



The syntax and semantics of discourse particles  

 21

A further exception that can be found in German is more problematic since it appears to be even more complex. A DiP can also occur in the left periphery together with a focus particle (FP) and a DP.7 (8) Wohl nur FRITZ hat er getroffen. prt only Fritz has he met ‘Fritz was probably the only one he met.’ The previous discussion has shown that even though generalisations can be made concerning the syntactic distribution of DiPs, exceptions to most of these generalisations are not difficult to find. Turning from their distribution to their contribution to the clause, in the next section, we briefly summarize how the impact of discourse particles on their host clause has been analysed in the literature.

2.2 Contribution to the interpretation of the clause This section summarises previous analyses of the semantic contribution of discourse particles to set the appropriate background for our analysis. As early as Helbig (1988: 33), who in turn summarizes older literature, we find the claim that DiPs qualify the position the speaker takes towards the expressed proposition. In other words, they modify the attitude of the speaker towards his utterance. In Thurmair’s (1989) terminology, if the clause is divided into a propositional or locutionary part, which hosts the content, the proposition, of the utterance, and an illocutionary part, which provides information of how the utterance is to be integrated into the general discourse, it is on this second level that the DiPs contribute to the clause. This general view on the contribution of discourse particles is a widely accepted approach towards their meaning up to now, but there are dissenting opinions, e.g., Gutzmann (2008, 2017). Nevertheless, the actual details of how this modification takes place or how DiPs impact the speaker’s attitude towards the proposition have often been subject to discussion, not in small part due to the apparently rather different meanings one single DiP seems to be able to contribute to the clause. Taking the particle ja as an example for how the contribution of DiPs can be analysed in various ways, it has been argued that the particle is used felicitously if

7 A discussion of this phenomenon is beyond the scope of this paper. However, it needs to be pointed out that in this kind of examples, the DiP cannot be argued to be inside the moved DP, and DiPs generally do not adjoin to DPs. (i) Wohl nur der schnellste ICE schafft es in zwei Stunden nach  Hannover prt only the fastest ICE manages it in two hours to Hannover ‘Probably only the fastest ICE can manage to be in Hannover in two hours.’

22 

 Markus Egg and Johannes Mursell

the proposition p of the host sentence is already part of the common ground. This, however, is not the only environment in which the DiP can occur in, as is shown by Kratzer and Matthewson (2009) and the discourse initial utterance in (9). (9) Du hast ja ein Loch im Ärmel. you have JA a hole in.the sleeve ‘Oh, you have a hole in your sleeve.’ To capture both apparent uses, Grosz (2010) moved away from the idea that p needs to be part of the common ground and claimed that the status of these host sentences can be understood as “privileged information”. The notion of “uncontroversiality”, introduced by Kaufmann and Kaufmann (2012), seems to spell out the same intuition. Jacobs (1991), in contrast, analyses the contribution of the DiP ja as a modification of the illocutionary type of the sentence the particle occurs in. Set in the framework of illocutionary semantics (Zaefferer 1979, 1984), Jacobs (1991) claims that a simple assertion like (10a) can be represented as (10b), where ass stands for assertion, and sp and ad for speaker and addressee, respectively. (10) a. Ruth verehrt Helmuth. Ruth adores Helmuth ‘Ruth adores Helmuth.’ b. ass(sp,ad,VEREHR(RUTH,HELMUTH)) (10b) basically assigns (10a) the assertive type that is directed towards an addressee and conveys the information specified in the semantic paraphrase of (10a). Adding a discourse particle like ja to (10a) maps the illocutionary type ASSERT onto a different type, JA-ASSERT. More generally, modifying an illocutionary type I with a discourse particle maps it onto a new, more specific type I′. However, this analysis does not yet spell out the exact details of the modification and its syntactic reflex, since no syntactic position is available to host the illocutionary type operator inside the clause (cf. Ross 1970; Speas and Tenny 2003). Rizzi’s (1997) split-CP analysis seems to provide such a syntactic position, as it reserves the highest projection in the split CP for the illocutionary force of the clause8 and thus enables an interaction between discourse particle and illocutionary force of the utterance already in the syntax. Consequently, this has been the most common way to analyse the contribution of DiPs to the clause from a

8 Even though the projection is called ForceP, it actually hosts clause type information. Thus, another projection hosting illocutionary force proper is necessary, a problem which will be ­addressed below.



The syntax and semantics of discourse particles  

 23

syntactic perspective, e.g., in Grosz (2005), Coniglio (2007, 2011), or Zimmermann (2004b). Evidence for the fact that DiPs do indeed modify the illocutionary force of the clause they occur in comes from combining this idea with assumptions about the left periphery of main and embedded clauses in Haegeman (2002, 2004): She claims that root clauses differ from non-root clauses in that only the former have a fully-fledged left periphery, represented in (11a), while the latter lack certain projections, most importantly ForceP (11b). (11) a. Root clauses: [ (Sub) [ForceP [TopP* [FocP [ModP* [FinP ... ]]]]]] b. Non-root clauses: [ (Sub) [ModP* [FinP ... ]]] Coniglio (2007) shows that DiPs can occur exactly in those environments that are considered to be root clauses, providing strong evidence for the assumption that DiPs modify illocutionary force. Thus, they can occur embedded under verba dicendi (12a), factive attitude predicates (12b), and those adverbial clauses that contain an independent force projection according to Haegeman (2004) (12c), but not in complements of other factive verbs (12d).9 (12) a. Peter hat gesagt, dass die NPD wohl verboten wird. Peter has said that the NPD WOHL banned will ‘Peter said that the NPD will (probably) be banned.’ b. Franziska ist eingefallen, dass Klaus ja allergisch auf Nüsse ist. Franziska is remembered that Klaus JA allergic on nuts is ‘Franziska remembered that Klaus is allergic to nuts (and that’s old information).’ c. Der hat die Prüfung nicht bestanden, trotzdem er the has the exam not passed even.though he ja recht intelligent ist. JA rather intelligent is ‘He didn’t pass the exam, even though (we both know) that he is rather intelligent.’ Thurmair (1989: 78) d. Er leugnete, dass er die Zeugin *ja unter he denied that he the witness JA under Druck gesetzt habe. pressure put had ‘He denied that he put pressure on the witness.’

9 Cf. also Döring (2013) who argues for a similar distribution of ja and wohl, against ­Zimmermann (2004a).

24 

 Markus Egg and Johannes Mursell

The sentences in (12) differ with respect to the question of whether the DiP in the embedded clause refers to the embedded clause (and with it, to the holder of the attitude expressed in this clause), or to the matrix clause and its speaker. E.g., in (12a), the assumption is Peter’s, not of the speaker of (12a). In contrast, Klaus’ allergy in (12b) is old information for the speaker, not for Franziska. As predicted by Coniglio’s (2007) argumentation, DiPs are also acceptable in non-restrictive but not in restrictive relative clauses. This distinction is often not marked in German, e.g., in (13a), but as soon as there is a DiP in the clause, it is disambiguated to the non-restrictive reading only (13b). (13) a. Die Dänen, die gerne Fußball gucken, trinken viel Bier. the danish who gladly football watch drin a.lot.of beer ‘The Danish who like to watch football drink a lot of beer.’ b. Die Dänen, die ja gerne Fußball gucken, trinken viel Bier. the danish who JA gladly football watch drink a.lot.of beer ‘The Danish, who (as it is known) like to watch football, drink a lot of beer.’ Consequently, analysing the contribution of DiPs along these lines seems to be on the right track. Even more evidence for this hypothesis is presented in Zimmermann’s (2004b) detailed discussion of wohl. He shows that wohl cannot be interpreted as part of the proposition, but neither as an expressive (which would be analysed on a completely different level of meaning). Therefore, he interprets wohl clause internally, but above the level of the proposition, and connects it to illocutionary force.10 In sum, many of the fundamental properties of DiPs are still being debated. The two most important questions we will try to answer concern the actual interaction of the particles with the illocutionary force of their host utterance, i.e., their semantic contribution, and their syntactic connection to the illocutionary force.

3 Discourse particles and felicity conditions In this section, we outline our proposal of how discourse particles and illocutionary force interact. In a nutshell, we propose that, following Egg (2013) and ­Waltereit (2001), DiPs modify a certain sub-component of illocutionary force, viz., felicity conditions. We will first introduce the notion of felicity conditions before outlining our proposal for the DiPs ja and wohl.

10 See Gutzmann (2015) for a modification of Potts’ (2007) approach which can analyse DiPs as expressives.



The syntax and semantics of discourse particles  

 25

3.1 Felicity conditions Linguistic utterances are subject to certain conditions to be felicitous (Austin 1962). They not only transmit information (in a locutionary act) but also perform a specific action (in an illocutionary act) such as asking a question or making a statement. These actions are called speech acts. Consequently, utterances can not only be true or false in the traditional truth conditional sense, they can succeed or be felicitous (Austin 1962: 14), if they comply with the necessary conditions for their successful use. Different illocutionary acts can have the same propositional content (Searle 1969: 22).11 (14) a. Sam smokes habitually. Act: assertive, Content: that Sam smokes habitually b. Sam, smoke habitually! Act: directive, Content: that Sam smokes habitually c. Does Sam smoke habitually? Act: question, Content: that Sam smokes habitually Searle (1969) works out the conditions that safeguard the felicitous performance of specific speech acts. They can be formulated as a set of propositions that describe what it means to perform a speech act successfully. He analyses in detail five illocutionary acts, viz., assertions, promises, requests, warnings, and yes-no questions. We will focus on the occurrences of ja, wohl, and doch in assertions, questions, and requests, whose felicity conditions are given in the table below. Here, S, H, and A stand for speaker, hearer, and described action, respectively: The notion of illocutionary force is further developed in Searle and Vanderveken (1985) and Vanderveken (1990). They model speech acts by F(  p); an illocutionary force F applied to a proposition p. In contrast to former approaches, however, force is not seen as atomic but decomposed into six more basic components: 1. Illocutionary Point Π 2. Mode of Achievement M 3. Propositional Content Condition C 4. Preparatory Conditions P 5. Sincerity Conditions S 6. Degrees of Strength D

11 Searle (1969: 29) actually claims that all three speech acts in (14) express the same proposition. We find this use of the term proposition rather problematic, especially with respect to questions. Therefore, we use the notion of propositional content.

26 

 Markus Egg and Johannes Mursell

Tab. 2: Felicity conditions for speech acts – Searle (1969: 66) Assert

Question

Request

Propositional Content Preparatory

Any proposition p.

Any proposition or ­propositional function. 1. S does not know the answer. 2. It’s not obvious to S and H that H will produce the answer without asking. 3. S has reasons to assume H knows the answer.12

Future act A of H.

Sincerity Essential

S believes p. Counts as an ­undertaking to the effect that p ­represents an actual state of affairs.

1. S has sufficient evidence for p. 2. It is not obvious to S and H that H knows p.

S wants information. Counts as an attempt to elicit this information from H.

1. H is able to do A and S believes H is able to do A. 2. It is not obvious to both S and H that H will do A in the normal course of events of his own accord. S wants H to do A. Counts as an attempt to get H to do A.

A speech act of the form F(p) can be represented in more detail like in (15) (with Φ representing the proposition). (15) F(Φ) = [M,C,P,S,D,Π](Φ) The components of the illocutionary force can then be modified individually. The adverb please, for example, makes the mode of achievement of the illocutionary point more polite. For a directive, symbolised by the illocutionary point !, this can be represented as follows. (16) F(Φ ⋅ please)=[M,C,P,S,D,!](Φ ⋅ please) = [please][M,C,P,S,D,!](Φ) = [M ⋅ please,C,P,S,D,!](Φ)

12 The third condition adapts the first preparatory condition of a request. It is not Searle’s but added here in order to bring out the request-like character of a question.



The syntax and semantics of discourse particles  

 27

3.2 Modifying FCs We assume that just like please in (16) above, discourse particles can modify specific components of the illocutionary force of their host utterance, viz., felicity conditions. This is in line with intuitions of previous research on discourse particles. For example, while discussing examples like (17), Zimmermann (2012: 2013) claims that “[a] difference in the choice of the particle [...] leads to a difference in felicity conditions, [...] such that each sentence will be appropriate in a different context”. (17) a. Max ist ja auf See. b. Max ist doch auf See. c. Max ist wohl auf See. ‘Max is PRT at sea.’ Karagjosova (2004) also analyses DiPs as modifiers of felicity conditions. She argues that they “contribute to the meaning of the utterance by providing additional preparatory conditions, that partly determine the speech act they occur in.” (Karagjosova 2004: 8). In the same vein, Waltereit (2001) analyses the French quand même as modifier of the preparatory conditions of the speech acts it occurs in. What all these approaches have in common is that they do not spell out in detail how discourse particle and felicity conditions interact. We propose that the felicity conditions can actually be seen as binary features which can be negated by discourse particles. For instance, the DiP wohl introduces uncertainty towards the propositional content of the utterance in assertions, due to factors like the lack of evidence. Following Tab. 2, this negates the first preparatory condition of assertions, which states that, to felicitously make an assertion, the speaker needs to have enough evidence for what he is asserting. (19) shows the formalization of this modification. (“Ͱ” represents the illocutionary type of assertions and “PCn”, the nth preparatory condition.) (18) Hein ist wohl auf See. Hein is PRT at sea ‘Hein is presumably at sea.’ (19) F(Φ)=[M,C,P={PC1,PC2},S,D,Ͱ](Φ ⋅ wohl) = [wohl][M,C,P={PC1,PC2},S,D,Ͱ](Φ) = [M,C,wohl ⋅ P={PC1,PC2},S,D,Ͱ](Φ) = [M,C,P={PC1,¬PC2},S,D,Ͱ](Φ)

28 

 Markus Egg and Johannes Mursell

The contribution of wohl in questions can be analysed in a very similar fashion. Above, it was argued that the third preparatory condition of questions is that the speaker actually believes that the addressee knows the answer to the question, so that he can fully commit to the answer. Thus, in asking a question like (20), the speaker does not presuppose that the addressee knows the answer and consequently, the addressee is permitted to give an assumption as answer. (20) A: Wen hat er gestern wohl getroffen? B: Klaus, whom has he yesterday PRT met Klaus glaube ich. believe I A: ‘Whom did he meet yesterday, what do you think?’ B: ‘Klaus, I believe.’ (21) F(Φ)=[M,C,P={PC1,PC2,PC3},S,D,?](Φ ⋅ wohl) = [wohl][M,C,P={PC1,PC2,PC3},S,D,?](Φ) = [M,C,[wohl] ⋅ P={PC1,PC2,PC3},S,D,?](Φ) = [M,C,P={PC1,PC2,¬PC3},S,D,?](Φ) The contribution of ja can be analysed analogously. It is restricted to assertions and marks the propositional content of the assertion as already part of the common ground (“already known”; Thurmair 1989), probably known to the addressee (Kratzer 1999), uncontroversial or obvious (Kaufmann 2010; Kaufmann and Kaufmann 2012), or as privileged information (Grosz 2010). It has been argued that a similar component is part of the meaning of doch. We propose for both ja and doch that these intuitions from the previous literature can be summarised in terms of negating the second preparatory condition of assertions, namely that it is not obvious to S and H that H knows p: (22) F(Φ)=[M,C,P={PC1,PC2},S,D,Ͱ](Φ ⋅ ja) = [ja][M,C,P={PC1,PC2},S,D,Ͱ](Φ) = [M,C,[ja] ⋅ P={PC1,PC2},S,D,Ͱ](Φ) = [M,C,P={PC1,¬PC2},S,D,Ͱ](Φ) Analysing the modification of illocutionary force by discourse particles as a negation of felicity conditions enables us to analyse the contribution of various discourse particles in the same way, without any further assumptions. As has been shown, this approach is also in line with many intuitions offered in the literature.

4 Syntactic background Describing the interaction of DiP and illocutionary force as a modification of felicity conditions raises the question of how the two elements are syntactically

The syntax and semantics of discourse particles  



 29

connected. After all, they are not in a local relation, since the DiP is an element of the TP, while the illocutionary force is hosted in the highest projection in the left periphery. Following the idea of replacing LF movement with agree (Chomsky 2000, 2001), connecting DiPs and left periphery based on agree raises the question of which features are involved in the agreement process. After introducing two additional assumptions that are necessary for a syntactic treatment of discourse particles, the split ForceP hypothesis in Section 4.1 and a slightly modified agreement system in 4.2, we present two recent syntactic approaches towards discourse particles based on agree and discuss their problems.

4.1 Split ForceP As pointed out above, we assume that discourse particles modify the illocutionary force of the clause. Zimmermann (2004a) assumed that the connection between the two elements is established via LF movement of the particle to the specifier of Rizzi’s ForceP, which makes possible modifying the illocutionary force of the clause in a specifier-head relation. However, Rizzi (1997) did not distinguish between clause type and illocutionary force proper. Thus, it is actually not illocutionary force hosted in the highest projection in the left periphery but clause type information, where specifications such as DECL for “declarative”, or Q for “interrogative” are located. To make illocutionary force information accessible for modification in the syntax, we follow researchers like Coniglio and Zegrean (2012), Heim et al. (2014), and Paul and Pan (2017) in assuming a split ForceP, split into a projection for illocutionary force, ILL, which dominates a projection for clause type, CT: (23)

ILL ILL0

CT CT0

...

Roussou (2000) and Endo (2012) cite two kinds of evidence for this hypothesis. First, discourse particles are restricted to certain clause types. In addition, they also seem to be independently dependent on the type of the illocutionary force. E.g., denn is incompatible with rhetorical questions like (24b), even though it is fine with literal questions like (24a): (24) a. Wer ist denn Weltmeister geworden? who is DENN world.champion became ‘Who made the world championship in the end?’

30 

 Markus Egg and Johannes Mursell

b. # Könntest du denn bitte das Fenster schließen? could you DENN please the window close ‘Could you close the window please?’ Furthermore, some languages overtly encode clause type and illocutionary force distinctly. E.g., the Japanese DiP yo is acceptable in imperatives expressing solicitation but not in imperatives which are intended as orders. (25) a. Are-o mi-ro-yo! that-acc look-IMP-YO ‘Look at that!’ b. *‘Hayaku tabe-ro-yo’ to meirei sareta. quickly eat-IMP-YO C order pass int.:‘I was ordered to eat quickly.’ (Endo 2012: 2)

4.2 Multiple agree We then modify one of the standard assumptions of current minimalist syntax in that we replace the agree system introduced in Chomsky (2000, 2001) with the one of Pesetsky and Torrego (2007). The former assumes (26) and (27), while the latter is more flexible when it comes to agree, feature valuation, and feature deletion. (26) agree (Assignment version) a. An unvalued feature F (a probe) on a head H scans its c-command domain for another instance of F (a goal) with which to agree. b. If the goal has a value, its value is assigned as the value of the probe. (27) Valuation/Interpretability Biconditional A feature F is uninterpretable iff F is unvalued. The definition in (26) states that agreement always takes place between a higher uninterpretable unvalued feature, the probe, c-commanding a lower interpretable valued feature, the goal. If agree takes place, the value of the lower feature is taken on by probe, but no other connection is established between the two features. Importantly, the uninterpretable feature is necessarily also unvalued, as stated in (27). Pesetsky and Torrego (2007) abandon the biconditional in (27), allowing interpretable features to be unvalued and uninterpretable features to carry a value



The syntax and semantics of discourse particles  

 31

when introduced into the syntactic computation. What they retain, however, is the restriction that only unvalued features can serve as probe for the agree mechanism. The new agree procedure is given in (28). (28) Agree (Feature sharing version) a. An unvalued feature F (a probe) on a head H at syntactic location α (Fα) scans its c-command domain for another instance of F (a goal) at ­location β (Fβ) with which to agree. b. Replace Fα with Fβ, so that the same feature is present in both locations. (28b) is an important deviation from the Chomskyan system in that it establishes a connection between the two instances of the feature after agreement has taken place. Consequently, nothing prevents agree between two uninterpretable features, as long as the c-commanding one is unvalued, creating what the authors call a feature chain, as exemplified in (29). In the first step (29a), agreement is established between two uninterpretable features, uFβ[] and uFϒ[]:val, so that the two features share the value of uFϒ. This is marked by assigning the same integer to both features, in this case [3]. In a second step, this shared feature can be targeted as a goal by a higher probe, iFα. After iFα agrees with the feature shared by uFβ and uFϒ, the same feature is present in three locations. Put differently, due to two agreement processes, a feature chain is created in which the same feature is present in three instances with one instance, iFα, being interpretable. (29) a. uFβ[] .. uFϒ[]:val → uFβ[3] ... uFϒ[3]:val b. iFα[] .. uFβ[3] .. uFϒ[3]:val → iFα[3] .. uFβ[3] .. uFϒ[3]:val In theory, nothing restricts the size of a feature chain. However, a derivation is only successful if at least one instance of the feature chain is interpretable at the end of the derivation. This is enforced by Brody’s thesis of radical interpretability: (30) Thesis of Radical Interpretability (Brody 1997: 143–144) Each feature must receive a semantic interpretation in some syntactic location.

4.3 Previous approaches to the syntax of DiPs This section briefly discusses two previous approaches that employ the modifications introduced above. We will show that these modifications cannot yet account for the way in which discourse particles interact with the illocutionary force of the clause. In particular, the problem for both theories relates to the features that

32 

 Markus Egg and Johannes Mursell

are taken to be responsible for the agreement process between illocutionary force and the discourse particle. Bayer and Obenauer (2011) investigate the discourse particle denn, which is restricted to questions. Briefly, they use the modified agree system just presented, and assume that the DiP is merged above the vP, carrying an unvalued uninterpretable feature [uQForce]. (31) [PrtP PrtuQForce[] [vP/VP ... ]] As the last step of the derivation, the Force/Fin head is merged, carrying its valued interpretable counterpart. This head probes and an agree relation is established between illocutionary force and discourse particle: (32) [FinP/ForceP Wh Force/FiniQForce:val[4] ... [PrtP PrtuQForce[4] [vP/VP ... ]]] Bayer and Obenauer’s analysis accounts for the particles’ optionality since the valuation of a feature in illocutionary force is not dependent on the DiP. In addition, it provides a very convenient way for modelling the occurrence of several particles inside the same clause. Pesetsky and Torrego (2007) explicitly allow for two uninterpretable features to enter into an agreement relation. Thus, when several discourse particles occur in the same clause, all the DiPs enter into an agreement relation based on their uQForce features, so that all the particles share the same unvalued feature (cf. the discussion of [29b]). This feature then is valued via agreement with the Force/Fin head on all particles at the same time because it is the same feature (Bayer and Obenauer 2011: ex. 32). (33) [FinP/ForceP Wh Force/FiniQForce:val[ ] ... [PrtP1 Prt1uQForce[ ] [PrtP2 Prt2uQForce[ ] [PrtP3 Prt3uQForce[ ] [vP/VP ... ]]]]] (34) [FinP/ForceP Wh Force/FiniQForce:val[12] ... [PrtP1 Prt1uQForce[12] [PrtP2 Prt2uQForce[12] [PrtP3 Prt3uQForce[12] [vP/VP ... ]]]]] Problematic for all agreement-based analyses of discourse particles is their ­ptionality. Since the particles are optional in almost all circumstances, the o ­illocutionary force of a clause cannot depend on the presence of a certain DiP. The consequence for any analysis based on agree is that a simple configuration like (35), in which the illocutionary force carries an unvalued feature, acts as probe, and agrees with the valued feature of the discourse particle will fail in the absence of the particle. If no element is present to provide a value for the feature of the illocutionary force, an unvalued feature will be sent to the interfaces, forcing the derivation to crash. (35) ILLiF[ ]:_ ... DiPuF[ ]: val → ILLiF[4]: val ... DiPuF[4]: val



The syntax and semantics of discourse particles  

 33

To handle this problem, Bayer and Obenauer (2011: 11) assume the following feature configuration. Talking about DiPs which are restricted to questions, they “… attribute to them the feature [QForce]. [QForce] is an unvalued uninterpretable feature which is valued by iQForce. iQForce is associated with the force/ fin-head”. Thus, the authors assume that the DiP carries an unvalued uninterpretable feature, while the Force/Fin head carries the valued interpretable counterpart. On the one hand, this makes the illocutionary force independent of the presence of a discourse particle, since the feature is already valued. On the other hand, it requires a further modification of the agreement system. As has been pointed out above, syntactic probes, be it in the system of Chomsky (2000, 2001) or the one of Pesetsky and Torrego (2007), are active as probes only due to an unvalued feature. Assuming that the DiP is the constituent that carries the unvalued feature and should therefore be active as a probe requires a different agree mechanism along the lines of so-called Upward Agree, as proposed by Zeijlstra (2012) and Bjorkman and Zeijlstra (2014) (but cf. Preminger 2013). Coniglio and Zegrean’s (2012) analysis of DiPs also employs the modified version of agree but differs from Bayer and Obenauer (2011) in two important respects. First, a more complex interaction of the DiP with the left periphery is assumed, which leads to different agreement processes, and second, as a consequence of the first point, the problem of optionality is solved in a different way: based on the modified agree process and the split ForceP hypothesis, Coniglio and Zegrean postulate two features involved in the agree process, one for clause type [Type], and one for illocutionary force [Intent]. The particle enters the derivation with uninterpretable yet valued features for illocutionary force and clause type, while the clause type head and the head for the illocutionary force carry unvalued yet interpretable features of their respective type. In addition, the illocutionary force also carries an unvalued uninterpretable clause type feature, which is intended to reflect that, usually, a specific illocutionary force is realized by a certain clause type.13 (36) a. ILL: [uType][]/[iIntent][] b. CT: [iType][] c. Prt: [uType:val[]]/[uIntent:val[]] In the course of the derivation, the particle is merged as an adjunct to vP.

13 Note that in Bayer and Obenauer (2011), the force/fin head also carries two independent features, one for clause type, iQ, and one for illocutionary force, iQForce.

34 

 Markus Egg and Johannes Mursell

(37) [vP PrtuType:val[]/uIntent:val[] [vP/VP ... ]] After several other elements have been introduced, the clause type head is merged. Due to its unvalued clause type feature, it is active as a probe and probes for the valued feature on the particle. (38) [CT CTiType[4] ... [vP PrtuType:val[4]/uIntent:val[][vP/VP ... ]]] In a last step, the head carrying illocutionary force is merged. Like the clause type head, it is active as a probe, due to its unvalued features. The unvalued feature for illocutionary force probes for the valued feature on the particle and the unvalued feature for clause type probes for its counterpart on the clause type head. (39) [ILL ILLiIntent[7]/uType[4] [CT CTiType[4] ... [vP PrtuType:val[4]/uIntent:val[7] [vP/VP ... ]]] Due to their unvalued features, CT and ILL can actually probe for the particle and it becomes possible to establish agree relations. Consequently, all features can be valued and the derivation can converge. However, clause type and illocutionary force depend on the presence of a particle to value the respective feature, so that a derivation without particle would crash. Thus, Coniglio and Zegrean have to assume covert or unpronounced DiPs even in derivations without overt discourse particle, to ensure that CT and ILL can value their features. It then becomes necessary to postulate several covert particles, one for every possible combination of illocutionary force and clause type, to ensure that in every case, the correct particle to value the features of ILL and CT can be merged. In sum, both recent approaches to the syntax of DiPs encounter problems when they attempt to model the connection between discourse particle and left periphery, which are mostly due to the optionality of the particle. In both approaches, particle and left periphery are connected by a feature that is restricted to these two elements. Thus, in cases where DiPs are absent, the valuation of this specific feature becomes problematic. The two approaches discussed solve this problem differently: Bayer and Obenauer (2011) assume that the illocutionary force carries a valued feature and consequently does not depend on the presence of the particle, which, however, makes the illocutionary force inactive as a probe (cf. [28a]). On the other hand, Coniglio and Zegrean (2012) claim that the illocutionary force head carries an unvalued feature and probes for the discourse particle, but since the valuation of the feature depends on the presence of a DiP, the authors have to assume phonologically null discourse particles to guarantee successful derivations without overt DiPs. What we propose in the following sections is that the feature connecting DiP and illocutionary force is not restricted to that relation. We argue that the DiP enters an agreement relation that is established between illocutionary force and focus and



The syntax and semantics of discourse particles  

 35

thus does not depend on the presence of the particle. In the next section, we first outline this relation between illocutionary force and focus based on the theory of obligatorily bound focus of Jacobs (1984) before we turn to our own proposal concerning the connection between illocutionary force and discourse particle.

5 DiPs and focus – the case of wohl In this section, we argue based on the example of the DiP wohl that discourse particles can enter a relation that is established between illocutionary force and focus, basically serving as an intermediate agreement step. On the one hand, this reflects the intuition that DiPs modify the relation between speaker and propositional content of the utterance, because they do indeed modify the relation between (elements of) the propositional content and the intentions of the speaker, and on the other hand, the relation is completely independent from the presence of the particle, accounting for its optionality. Before turning to our proposal, however, we will discuss the theory of obligatorily bound focus of Jacobs (1984).

5.1 Bound focus Discourse particles modify the relation between the content of an utterance and the speaker. Regardless of the presence or absence of discourse particles, a relation between content of the utterance and speaker, or between propositional content and illocutionary force is established in every utterance. Following Jacobs (1984), we assume that this relation is established via focus, more specifically in our proposal, via agreement based on focus features. Traditionally, two types of focus are distinguished, viz., free focus like in (40), which has no truth conditional impact, and bound focus as in (41), which does affect truth conditions (cf. Zimmermann and Onea 2011: 1665). (40) a. [PETER]F  gave Mary a kiss. b. Peter gave [MARY]F a kiss. c. Peter gave Mary [a KISS]F. true in all worlds in which Peter gave Mary a kiss (41) a. Marek only bought [PIZZA]F in Gdansk. true in all those worlds in which pizza is the only thing that Marek bought in Gdansk

36 

 Markus Egg and Johannes Mursell

b. Marek only bought pizza [in GDANSK]F . true in all those worlds in which Gdansk is the only place where Marek bought pizza In a structured meaning account (Krifka 1992) that divides semantic representations A into background α and focus β (i.e., A = 〈α, β〉), this can be represented as follows. Without a focus sensitive particle, different focus placements or differently sized focus domains in the same sentence do not affect truth ­conditions, e.g., in (42). (42) a. [PETER]F besucht Gerdas Schwester. Peter visits Gerda’s sister ‘Peter visits Gerda’s sister.’ b. 〈λx.x besucht Gerdas Schwester, Peter〉 In contrast, the focus sensitive particles’ meanings make the meaning of the overall clause dependent on the choice and size of focus. For instance, only introduces the claim that for none of β’s focus alternatives X that differs from β, α(X). Here “≈” represents the relation of being a focus alternative: (43) only(〈α,β〉) = α(β)∧∀X[X ≈ β∧α(X) → X = β] Thus, (44a) and (44b) differ in their truth conditions. As the meaning of (44a), we get (45a), the claim that the only person introduced to Bill by John is Sue. For (44b), the meaning is (45b), which states that the only thing John did was introduce Bill to Sue: (44) a. John only introduced Bill to [SUE]F only (〈λx.introduce(john,x,bill),sue〉) b. John only [introduced Bill to SUE]F only (〈λP.P(john),λx.introduce(x,sue,bill)〉) (45) a. introduce(john,sue,bill) ∧∀x[x ≈ sue ∧ introduce(john,x,bill) → x=sue] b. introduce(john,sue,bill)∧ ∀P[P≈λx.introduce(x,sue,bill) ∧ P(john) → P= λx.introduce(x,sue,bill)] Jacobs (1984) claims that there is no distinction between free and bound focus and that apparently free focus is actually bound by the illocutionary type operator, which we assume to be equivalent to illocutionary force. (46) a. [PETER]F besucht Gerdas Schwester. Peter visits Gerda’s sister ‘Peter visits Gerda’s sister.’ b. ASS(〈λx.x besucht Gerdas Schwester,Peter〉)



The syntax and semantics of discourse particles  

 37

In analogy to (45), (46b) claims that the partitioning of the proposition into focus and background parts influences meaning because the ASS operator takes scope over the background-focus structure. This influence can be based on the characterization of ASSERT following Krifka (1992: 20): (47) ASSERT (〈α, β〉) maps a common ground c to a common ground c’, where c’ is the intersection of c with the set of possible worlds for which α(β) is true, i.e. c′ = c∩α(β) Felicity conditions (among others): a. c′ ≠ c (asserting α(β) makes a difference in CG) b. c′ ≠ 0 (the truth of α(β) must not be already excluded by c) c. There are X, with X ≈ β and X ≠ β, such that α(X) could have been asserted with respect to c. That is, it would have changed c, c∩[α(X)] ≠ c, it would not be excluded by c, c∩[α(X)] ≠ 0, and would have yielded a different output context, c∩[α(X)] ≠ c∩[α(β)]. The most important part in (47) is (47c), which states that it is important which contextually salient alternative is expressed, since asserting a different alternative could lead to differences in meaning. While for Krifka (1992) this effect is purely pragmatic, an effect on the felicity conditions, Jacobs (1984) places it on the level of illocutionary semantics (Zaefferer 1979, 1984). However, following Coniglio and Zegrean (2012) in assuming that the illocutionary force is part of the syntax of the clause, and taking into account what we have argued concerning the internal structure of illocutionary force, this effect on the illocutionary force can be seen as part of syntax and thus also as part of semantics. Note that the effect of focus on the illocutionary force of the clause is very similar to the effect discourse particles have, since both modifications of ILL can be seen as modifications of the felicity conditions of the speech act operator. Since it has been argued for DiPs that this modification takes place on a semantic rather than a pragmatic level, the same seems to be feasible for an analysis of focus. Syntactically, we assume that this binding of the focus by the illocutionary force is an agree relation that is established between focus and force. We assume that the focussed element carries an F-feature (Selkirk 1995), which is able to project, most likely assigned in the numeration (Aboh 2010 contra Chomsky 1995), which is formally an uninterpretable valued feature. The illocutionary force carries the interpretable unvalued counterpart14, so that it is active as a probe and an agree relation can be established. 14 We assume that the interpretable instance is in Force, since this is the position where the focus has a semantic impact, whereas the valued instance is marked on the constituent in focus.

38 

 Markus Egg and Johannes Mursell

(48) (dass) Peter [den WEIN]F trinkt. ‘(that) Peter is drinking the wine.’ In (49) and (50), it is shown how the agreement takes place between the focus features, (51) gives a slightly less detailed tree representation. (49) ... [vP [DP Peter] [v’ [VP [dp den Wein]uF[]:val trinkt ] trinkt ]]] (50) ... [ILL ILLiF[3] [CT ... [TP [DP Peter] [T’ [vP [DP Peter] [v’ [VP [DP den Wein]uF[3]:val trinkt ] trinkt]] trinkt ]]]] ILL

(51) ILL0

CT CT0

... TP

DP

T’

Peter

vP

T

DP

v’

Peter

VP

trinkt v

DPF

V0

denF WeinF

trinkt

trinkt

Thus, combining the bound focus approach of Jacobs (1984) with the powerful agreement system of Pesetsky and Torrego (2007) provides a simple way of modelling the contribution of “free” focus. Including the illocutionary force as part of the syntax makes this contribution more concrete, by describing it as an interaction with the illocutionary force. This similarity to the contribution of



The syntax and semantics of discourse particles  

 39

­ iscourse particles will be one of the components of the syntactic analysis of DiPs d developed in the next section.

5.2 wohl and focus In the preceding section we pointed out that previous syntactic approaches assumed an agreement relation between discourse particle and illocutionary force based only on features restricted to those two elements. Since DiPs are in most cases optional, this enforced certain additional assumptions in order to ensure successful derivations for sentences without DiPs. In this section, we argue that if the particle is analysed as being part of an agreement relation between left periphery and something else in the clause, more specifically between illocutionary force and focus, no such additional assumptions are necessary. To become part of the agree relation based on focus features, the DiP has to carry an unvalued focus feature as well, enabling it to probe for the valued uninterpretable feature on the focussed element. Evidence for the connection between DiP and focussed element comes especially from the behaviour of wohl, whose contribution seems to be restricted to the element in focus. (52) a. (dass) [PETER]F wohl den Wein trinkt. that Peter wohl the wine drinks ‘(that) someone is drinking the wine, and I think it’s Peter.’ b. (dass) Peter wohl [den WEIN]F trinkt. ‘(that) Peter is drinking something, and I think it’s the wine.’ c. (dass) Peter wohl [den wEIN trinkt]F. ‘(that) Peter is doing something, and I think it’s drinking the wine.’ d. (dass) Peter wohl den Wein trinkt. ‘(that) something is happening, and I think it’s that Peter is drinking the wine.’ The effect of wohl that the speaker does not fully commit to the truth of his statement is restricted to the focus domain of the sentences, which is reflected in the English translations. Note that in cases like (52b) and (52c), identical placement of the focus accent can still lead to different focus domains and consequently also to different meaning contributions of the DiP, since they only take the highest element marked by an F-feature into account. In cases like (52d), we assume vP focus, i.e., the vP carrying a focus feature and the contribution of the DiP being restricted to the complete proposition. Interactions between DiP and focus are less easily observable for other particles (but cf. Grosz 2016 for doch). Nevertheless, certain pieces of data suggest

40 

 Markus Egg and Johannes Mursell

a close connection between DiPs and information structure. However, the exact nature of this connection requires further in-depth investigation and lies beyond the scope of this paper, so that we restrict ourselves to a mere presentation of the data. First, the presence of a focus sensitive particle in the left periphery seems to affect the distribution of DiPs, and such a particle licenses the occurrence of DiPs in the left periphery, a position from which discourse particles are usually banned. (53) a. Wohl nur der neue ICE schafft es in zwei Stunden WOHL only the new ICE manages it in two hours nach Hamburg. to Hamburg ‘Probably only the new ICE manages it to get to Hamburg in two hours.’ b. *Wohl der neue ICE schafft es in zwei Stunden nach  WOHL the new ICE manages it in two hours to Hamburg. Hamburg int.: ‘Probably the new ICE manages it to get to Hamburg in two hours.’ Second, DiPs appear to interact with so-called verum focus. Egg and Zimmermann (2012) argue that DOCH is an instance of doch in a verum focus context, which, under our assumptions, could be modelled as association with verum. The same has been argued for JA by Gutzmann (2010). (54) A: Peter ist nicht krank. B: (Peter ist) DOCH (krank)! Peter is not ill Peter is DOCH ill A: ‘Peter isn’t ill.’ B:‘Yes, he is!’ SCHON seems to show a comparable behaviour (cf. Féry 2010). (55) A: Ich dachte Maria ist Vegetarierin. B: Sie isst SCHON Fleisch. I thought Maria is vegetarian she eats SCHON meat A: ‘I thought Maria is a vegetarian.’ B:‘No, she does eat meat.’ Similar to the interaction of illocutionary force with focus, we model the relation between discourse particles and focus as an agree relation based on focus features, exemplified by (56), which contains wide vP focus. (56) (dass) Peter wohl [den WEIN trinkt]F. ‘(that) something is happening, and I think it’s that Peter is drinking the wine.’ (57) ... [PrtP wohluF[3] [vP [dp Peter] [v’ [VP [DP den Wein] trinkt ] trinkt ]]uF[3]:val ]

The syntax and semantics of discourse particles  



(58)

 41

PrtP Prt0 wohl

vPF DP

v’F

Peter

v0F

VPF DPF

V0F

denF WeinF

trinktF

trinktF

We assume that the unvalued uninterpretable focus feature of the DiP probes in its c-command domain and agrees with the uninterpretable but valued focus feature of the focussed element, in this case the vP. Employing the modified version of agree proposed by Pesetsky and Torrego (2007), the derivation at this point would not be able to converge, since Brody’s (1997) thesis of radical interpretability is not satisfied, because no instance of the feature chain is interpretable. However, a solution to this problem is readily available. As discussed above, we assume that in addition to the connection between DiP and focus, there is also a connection established between illocutionary force and focus. Carrying an unvalued but interpretable feature, the illocutionary force probes in its c-command domain and agrees with the highest available focus feature, in this case the valued uninterpretable feature of the DiP. Consequently, all focus features are valued and the chain also contains an interpretable instance so that the derivation converges, as shown in (59) and the corresponding tree in (61). (59) ... [ILL ILLiF[3] [CT ... [TP [DP Peter] [T’ [PrtP wohluF[3] [vP [DP Peter] [v’ [VP [DP den Wein] trinkt ] trinkt]]uF[3]:val trinkt ]]]]] Comparing this to a derivation without discourse particle like (51) makes clear that the particle is in no way necessary for a successful derivation. It serves merely as an intermediate step for the agree relation between illocutionary force and focus. In this way it is possible to establish a connection between particle and left periphery without making the derivation dependent on the particle, accounting for its optionality. At the same time, combining the bound focus approach of Jacobs (1984) and the idea of DiPs carrying a focus feature enables proper agree between all features involved, resulting in a successful derivation.

42 

 Markus Egg and Johannes Mursell

This type of analysis makes certain predictions concerning the behaviour of discourse particles. First, the relation between illocutionary force and focus is not clause bound, so nothing should force the relation between illocutionary force and DiP to be clause bound either. Interestingly, it is possible for discourse particles in embedded clauses to modify the illocutionary force of the matrix clause.15 (60) a. Franziska ist eingefallen, dass Klaus ja allergisch auf Franziska is remembered that Klaus JA allergic on Nüsse ist. nuts is ‘Franziska remembered that Klaus is allergic to nuts (and that’s old information).’ b. Wie denkst du, dass es denn weitergehen soll mit euch? how think you that it DENN go-on should with you ‘How do you think that the two of you should carry on? (I’m wondering)’ ILL

(61) ILL0

CT CT0

... TP

DP

T’

Peter

T0

PrtP Prt0 wohl

vPF

trinkt v’F

DP Peter

V0F

VPF DPF

V0F

denF WeinF

trinktF

15 Thanks to the reviewer who brought this issue to our attention.

trinktF



The syntax and semantics of discourse particles  

 43

In both examples in (60), the respective discourse particle is not interpreted inside the embedded clause, i.e., modifying the illocutionary force of the embedded clause, but as modifying the illocutionary force of the matrix clause.16 In (60a), (12b) from above, it is shown that it cannot be known to Franziska that Klaus is allergic to nuts and therefore the contribution of the DiP ja refers back to the speaker of the complete utterance. The example in (60b), from Bayer and Obenauer (2011), makes an even stronger point since the DiP denn, being restricted to questions, is not even licensed in the embedded clause. In addition, the parenthesis in the translation, which is supposed to reflect the contribution of the DiP, makes it clear that again it is interpreted with respect to the speaker of the whole utterance. It needs to be noted that following Haegeman (2006), the embedded clause in example (60a) does not contain an independent illocutionary force, since it is a factive clause. Thus, the discourse particle, carrying an uninterpretable feature, and therefore being in need of agreement with an interpretable counterpart, necessarily has to agree with the illocutionary force of the matrix clause to satisfy radical interpretability. In (60b), Bayer and Obenauer (2011) link the possibility of having denn in an embedded declarative clause to wh-movement out of that clause. The second observation that can be linked to a syntactic relation between discourse particles and focus concerns intervention effects. In general, intervention effects occur when quantificational or focusing elements intervene between a whphrase and its licensing complementizer, resulting in ungrammaticality (Beck 2006). This is a rather widespread phenomenon, most obvious in wh-in-situ languages. In German, intervention configurations require multiple wh-questions. (62) *[Qi [ ... [ intervener [... wh-phrasei ... ] ] ] ] (63) a. *Wem hat nur [PETER]F was gegeben?  whom has only Peter what given int: ‘Whom did Peter only give what?’ b. Wem hat was nur [PETER]F gegeben? whom has what only Peter given ‘Whom did Peter only give what?’ Beck’s account for intervention effects is purely semantic. It is the binding of the variable introduced by the wh-element by an operator which is not Q that causes ungrammaticality due to the semantics of the operator. Importantly, in addition

16 This is not a general option for discourse particles. Usually, as pointed out above, their ­occurrence is restricted to clauses that contain independent illocutionary force and they are interpreted in a clause bound fashion. However, we believe that the availability of examples like (59a)–(59b) shows that other conditions need to be responsible for this restriction.

44 

 Markus Egg and Johannes Mursell

to focus sensitive particles and certain quantificational elements, discourse particles show intervention effects as well (64). The observation that the sentence can be rescued by scrambling the wh-element across the intervener shows that the ungrammaticality is indeed due to an intervention effect (65). (64) Wann hat Peter was mitgebracht? a. *Wann hat nur [PETER]F was mitgebracht? b. *Wann hat wohl [PETER]F was mitgebracht? c. *Wann hat doch [PETER]F was mitgebracht? when has PRT Peter what brought int.: ‘When did Peter bring what?’ (65) Wann hat Peter was mitgebracht? a. Wann hat was nur [PETER]F mitgebracht? b. Wann hat was wohl [PETER]F mitgebracht? c. Wann hat was doch [PETER]F mitgebracht? when has what PRT Peter brought ‘When did Peter bring what?’ This analysis of intervention effects requires the intervener to have a truth conditional effect when applied to its complement. However, as mentioned above, discourse particles do not have a truth conditional effect at all but modify the illocutionary force of their host utterance. Consequently, a semantic analysis cannot account for intervention effects caused by discourse particles, whereas a syntactic analysis based on focus features can provide a solution to the problem (Li and Cheung 2015). Under the assumption that one part of the relation between the Q operator in the left periphery and the wh-element is an agreement relation based on focus features (Breul 2004; Haida 2007), the intervention effect can be analysed as a Relativized Minimality effect (Rizzi 1990; Starke 2001): the discourse particle provides an appropriate goal, a uF feature, which is closer to Q than the uF feature of the wh-element and by this means prevents the agreement relation between Q and the wh-element. Thus, our analysis correctly predicts the option of interpreting an embedded discourse particle as modifying the illocutionary force of the matrix clause and it can also account for Intervention Effects caused by DiPs.

6 Conclusion In this paper we have presented a new way of how discourse particles contribute to their host clause and how this contribution can by syntactically modelled. We argued that, following much previous work, the contribution of DiPs can indeed



The syntax and semantics of discourse particles  

 45

be analysed as a modification of illocutionary force. Spelling this idea out in more detail, we proposed that they actually modify only a certain component of force, viz., its felicity conditions. By assuming that they negate one of the several felicity conditions of illocutionary force, it becomes possible to analyse their contribution in a meaningful way. In the syntax, the connection between illocutionary force and discourse particle has been modelled via an agree relation based on focus features. This relation is primarily established between force and the highest F-feature in its c-command domain. If a discourse particle is present, it only serves as an intermediate step in this relation, which is in line with the general optionality of discourse particles. Modelling the relation between DiP and illocutionary force without using features unique to those two elements avoids the problems encountered by earlier syntactic approaches. Many open questions remain, most importantly regarding the relation between discourse particles and focus and how this relation can account for the phenomena discussed above. Nevertheless, further in-depth investigation of the information structural impact of discourse particles might provide important answers regarding the nature of DiPs and information structure in general.

References Aboh, Enoch O. 2010. Information structuring begins with the numeration. Iberia 2. 12–42. Abraham, Werner. 1991a. Discourse particles in German: How does their illocutive force come about? In Werner Abraham (ed.), Discourse particles, 203–252. Amsterdam: Benjamins. Abraham, Werner. 1991b. Introduction. In Werner Abraham (ed.), Discourse particles, 1–11. Amsterdam: Benjamins. Austin, John L. 1962. How to do things with words. Oxford: Clarendon. Bayer, Josef & Hans-Georg Obenauer. 2011. Discourse particles, clause structure, and question types. The Linguistic Review 28(4). 449–491. Beck, Sigrid. 2006. Intervention effects follow from focus interpretation. Natural Language Semantics 14(1). 1 –56. Bjorkman, Bronwyn & Hedde Zeijlstra. 2014. Upward Agree is superior. Ms., Götting / Toronto. Breul, Carsten. 2004. Focus structure in Generative Grammar. Amsterdam: Benjamins. Brody, Michael. 1997. Perfect chains. In Liliane Haegeman (ed.), Elements of grammar, 139–167. Dordrecht: Kluwer. Chomsky, Noam. 1995. The Minimalist Program. Cambridge, MA: MIT Press. Chomsky, Noam. 2000. Minimalist inquiries: The framework. In Roger Martin, David Michaels & Juan Uriagereka (eds.), Step by step, 89–156. Cambridge, MA: MIT Press. Chomsky, Noam. 2001. Derivations by phase. In Michael Kenstowicz (ed.), Ken Hale: A life in language, 1–52. Cambridge, MA: MIT Press. Coniglio, Marco. 2007. German modal particles in root and embedded clauses. University of Venice Working Papers in Linguistics 17. 109–141.

46 

 Markus Egg and Johannes Mursell

Coniglio, Marco. 2011. Die Syntax der Deutschen Modalpartikeln: Ihre Distribution und Lizensierung in Haupt- und Nebensätzen. Berlin: Akademie Verlag. Coniglio, Marco & Iulia Zegrean. 2012. Splitting up force: Evidence from discourse particles. In Lobke Aelbrecht, Liliane Haegeman & Rachel Nye (eds.), Main clause phenomena. New horizons, 229–255. Amsterdam: Benjamins. Diesing, Molly. 1992. Indefinites. Cambridge, MA: MIT Press. Döring, Sophia. 2013. Modal particles and context shift. In Daniel Gutzmann & Hans-Martin Gärtner (eds.), Beyond expressives. Explorations in use-conditional meaning, 95–123. Leiden: Brill. Egg, Markus. 2013. Discourse particles, common ground, and felicity conditions. In Daniel Gutzmann & Hans-Martin Gärtner (eds.), Beyond expressives. Explorations in use-conditional meaning, 125–149. Leiden: Brill. Egg, Markus & Malte Zimmermann. 2012. Stressed out! Accented discourse particles – the case of DOCH. In Ana Aguilar Guevara, Anna Chernilovskaya & Rick Nouwen (eds.), Proceedings of Sinn und Bedeutung 16, vol.1, 225–238. Cambridge, MA: MITWPL. Endo, Yoshio. 2012. Illocutionary force of non-root sentences. Paper presented at University of Venice, March 22nd. Erlewine, Michael Y. 2014. Movement out of focus. Cambridge, MA: MIT dissertation. Féry, Caroline. 2010. Information structure of schon. In Thomas Hanneforth & Gisbert Fanselow (eds.), Language and logos. A Festschrift for Peter Staudacher, 160–176. Berlin: Akademie Verlag. Grosz, Patrick. 2010. German doch: An element that triggers a contrast presupposition. Proceedings from the Annual Meeting of the Chicago Linguistic Society 46(1). 163–177. Grosz, Patrick. 2005. ‘dn’ in Viennese German. The syntax of a clitic version of the discourse particle ‘denn’. Vienna: University of Vienna MA thesis. Grosz, Patrick. 2016. Information structure and discourse particles. In Caroline Féry & Shin Ishihara (eds.), The Oxford handbook of information structure, Oxford: Oxford University Press. Gutzmann, D. 2010. Betonte Modalpartikeln und Verumfokus. In Theo Harden & Elke Hentschel (eds.), 40 Jahre Partikelforschung, 119–138. Tübingen: Stauffenburg. Gutzmann, Daniel. 2008. On the interaction between modal particles and sentence mood in German. Mainz: Johannes Gutenberg-Universität Mainz MA thesis. Gutzmann, Daniel. 2015. Use-conditional meaning. Studies in multidimensional semantics. Oxford: Oxford University Press. Gutzmann, Daniel. 2017. Modal particles ≠ modal particles (= modal particles). In Josef Bayer & Volker Struckmeier (eds.), Discourse particles: Formal approaches to their syntax and semantics, 144–172. Berlin: de Gruyter. Haegeman, Liliane. 2002. Anchoring to speaker, adverbial clauses and the structure of CP. Georgetown University Working Papers in Theoretical Linguistics 2. 117–180. Haegeman, Liliane. 2004. Topicalization, CLLD and the left periphery. In Benjamin Shaer, Werner Frey & Claudia Maienborn (eds.), Proceedings of the Dislocated Elements Workshop, ZAS Berlin, November 2003 (ZAS Papers in Linguistics 35), 157–192. Berlin: ZAS. Haegeman, Liliane. 2006. Conditionals, factives and the left periphery. Lingua 116(10). 1651–1669. Haida, Andreas. 2007. The indefiniteness and focusing of wh-words. Berlin: Humboldt University Berlin dissertation.



The syntax and semantics of discourse particles  

 47

Heim, Johannes, Hermann Keupdjio, Zoe Wai-Man Lam, Adriana Osa-Gómez & Martina Wiltschko. 2014. How to do things with particles. In Proceedings of the 2014 Annual Conference of the Canadian Linguistic Association, 1–15. http://cla-acl.ca/wp-content/ uploads/Heim_Keupdjio_Lam_Osa-Gomez_Wiltschko-2014.pdf (accessed 20 May 2016) . Helbig, Gerhard. 1988. Lexikon deutscher Partikeln. Leipzig: Verlag Enzyklopädie. Hentschel, Elke. 1986. Funktion und Geschichte deutscher Partikeln: ‘ja’, ‘doch’, ‘halt’, und ‘eben’. Tübingen: Niemeyer. Jacobs, Joachim. 1984. Funktionale Satzperspektive und Illokutionssemantik. Linguistische Berichte 91. 25–58. Jacobs, Joachim. 1991. On the semantics of modal particles. In Werner Abraham (ed.), Discourse particles, 141–162. Amsterdam: Benjamins. Karagjosova, Elena. 2004. The meaning and function of German modal particles. Saarbrücken: Saarland University dissertation. Kaufmann, Magdalena & Stefan Kaufmann. 2012. Epistemic particles and performativity. Semantics and Linguistic Theory 22. 208–225. Kaufmann, Stefan. 2010. Strong and weak presupposition: German ja under quantifiers. Ms., Northwestern University. http://homepages.uconn.edu/ stk12004/Papers/ja.pdf (accessed 20 May 2016). Kratzer, Angelika. 1999. Beyond ‘Oops’ and ‘Ouch’. How descriptive and expressive meaning interact. Paper presented at the Cornell Conference on Theories of Context Dependency, Ithaca, NY, March 26, 1999. http://semanticsarchive.net/Archive/WEwNGUyO/ (accessed 20 May 2016). Kratzer, Angelika & Lisa Matthewson. 2009. Anatomy of two discourse particles. Paper presented at SULA5 – Semantic of under-represented Languages in the Americas V, MIT / Harvard University, May 15–17, 2009. Krifka, Manfred. 1992. A compositional semantics for multiple focus constructions. In Joachim Jacobs (ed.), Informationsstruktur und Grammatik, 17–53. Opladen: Westdeutscher Verlag. Li, Haoze & Candice Chi-Hang Cheung. 2015. Focus intervention effects in Mandarin multiple wh-questions. Journal of East Asian Linguistics 24(4). 361–382. Ormelius-Sandblom, Elisabet. 1997. Die Modalpartikeln ‘ja’, ‘doch’, ‘schon’. Zu ihrer Syntax, Semantik und Pragmatik. Stockholm: Almqvist & Wiksell. Paul, Waltraud & Victor Junnan Pan. 2017. What you see is what you get: Chinese sentencefinal particles as head-final complementisers. In Josef Bayer & Volker Struckmeier (eds.), Discourse particles: Formal approaches to their syntax and semantics, 49–77. Berlin: de Gruyter. Pesetsky, David & Esther Torrego. 2007. The syntax of valuation and the interpretability of features. In Simin Karimi, Vida Samiian & Wendy Wilkins (eds.), Phrasal and clausal architecture: Syntactic derivation and interpretation, 262–294. Amsterdam: Benjamins. Petrova, Svetlana. 2017. On the status and the interpretation of the left-peripheral sentence particles inu and ia in Old High German. In Josef Bayer & Volker Struckmeier (eds.), Discourse particles: Formal approaches to their syntax and semantics, 304–331. Berlin: de Gruyter. Potts, Christopher. 2007. The expressive dimension. Theoretical Linguistics 33(2). 165–197. Preminger, Omer. 2013. That’s not how you agree: A reply to Zeijlstra. The Linguistic Review 30(3). 491–500. Rizzi, Luigi. 1990. Relativized minimality. Cambridge, MA: MIT Press.

48 

 Markus Egg and Johannes Mursell

Rizzi, Luigi. 1997. The fine structure of the left periphery. In Liliane Haegeman (ed.), Elements of grammar, 281–337. Dordrecht: Kluwer. Ross, John R. 1970. On declarative sentences. In Roderick A. Jacobs & Peter S. Rosenbaum (eds.), Readings in English Transformational Grammar, 222–272. Waltham, MA: Ginn & Co. Roussou, Anna. 2000. On the left periphery: Modal particles and complementisers. Journal of Greek Linguistics 1. 65–94. Searle, John R. 1969. Speech acts. An essay in the philosophy of language. Cambridge: Cambridge University Press. Searle, John R. & Daniel Vanderveken. 1985. Foundations of illocutionary logic. Cambridge: Cambridge University Press. Selkirk, Elisabeth. 1995. Sentence prosody: Intonation, stress, and phrasing. In John A. Goldsmith (ed.), The handbook of phonological theory, 550–569. Oxford: Blackwell. Speas, Peggy. & Carol L. Tenny. 2003. Configurational properties of point of view roles. In Maria Di Sciullo (ed.), Asymmetry in grammar, 315–344. Amsterdam: Benjamins. Starke, Michael. 2001. Move dissolves into merge: A theory of locality. Geneva: University of Geneva dissertation. Thurmair, Maria. 1989. Modalpartikeln und ihre Kombinationen. Tübingen: Niemeyer. Vanderveken, Daniel. 1990. Meaning and speech acts. Cambridge: Cambridge University Press. Waltereit, Richard. 2001. Modal particles and their functional equivalents: A speech-acttheoretic approach. Journal of Pragmatics 33(9). 1391–1417. Zaefferer, Dietmar. 1979. Sprechakttypen in einer Montague-Grammatik. In Günther Grewendorf (ed.), Sprechakttheorie und Semantik, 386–417. Frankfurt a. M.: Suhrkamp. Zaefferer, Dietmar. 1984. Frageausdrücke und Fragen im Deutschen. München: Fink. Zeijlstra, Hedde. 2012. There is only one way to agree. The Linguistic Review 29(3). 491–539. Zimmermann, Malte. 2004a. Discourse particles in the left periphery. In Benjamin Shaer, Werner Frey & Claudia Maienborn (eds.), Proceedings of the Dislocated Elements Workshop, ZAS Berlin, November 2003 (ZAS Papers in Linguistics 35), 543–566. Berlin: ZAS. Zimmermann, Malte. 2004b. Zum Wohl: Diskurspartikeln als Satztypmodifikatoren. Linguistische Berichte 199. 253–286. Zimmermann, Malte. 2012. Discourse particles. In Klaus von Heusinger, Claudia Maienborn & Paul Portner (eds.), Semantics. An international handbook of natural language meaning, vol. 2, 2011–2038. Berlin: de Gruyter Mouton. Zimmermann, Malte & Edgar Onea. 2011. Focus marking and focus interpretation. Lingua 121(11). 1651–1670.

Waltraud Paul and Victor Junnan Pan

What you see is what you get: Chinese sentence-final particles as head-final complementizers The present article presents an in-depth analysis of the head-final three-layered split CP realized by sentence-final particles (SFPs) in the SVO language Mandarin Chinese. These SFPs are shown to be fully-fledged functional heads with a complex feature make-up, on a par with C elements in e.g. Indo-European languages. Chinese SFPs select and project, as evidenced by the strict hierarchy for ­co-occurring SFPs in the split CP. This structure must be merged as such and cannot be derived by postulating movement from a head-initial CP. It straightforwardly invalidates empirically superficial statements that attempt to turn Chinese SFPs into a grammatical quantité négligeable in order to uphold problematic word order generalizations such as the Final-over-Final Constraint.

1 Introduction In Chinese, zhùcí ‘particle’ has been used as a cover term for (mostly, but not exclusively monosyllabic) items whose categorial status is unclear. However, already in the 1980s, Zhu Dexi (1982) identified sentence-final particles (SFP) (yǔqì zhùcí ‘mood particles’) as a closed set of items, which he characterized as combining with the sentence as a whole and constituting three distributional classes. He obtained these classes by determining the paradigmatic and syntagmatic relations among SFPs: mutually exclusive SFPs were assigned to the same class, while co-occurring SFPs were assigned to different classes in terms of their strict ordering restrictions. Zhu Dexi’s (1982) three-partite division can

Acknowledgements: This article has greatly benefited from the comments by anonymous ­reviewers and by the ­editors, Josef Bayer and Volker Struckmeier. Waltraud Paul: Centre de recherches linguistiques sur l’Asie orientale CNRS-EHESS-INALCO, Paris, e-mail: [email protected] Victor Junnan Pan: Laboratoire de Linguistique Formelle, CNRS & Université Paris Diderot, Paris, e-mail: [email protected] DOI 10.1515/9783110497151-003

50 

 Waltraud Paul and Victor Junnan Pan

be easily recast into a split CP à la Rizzi (1997, 2004), modulo the addition of an additional speaker/hearer-related projection (Attitude Phrase) above Rizzi’s ForceP. As in the languages examined by Rizzi (1997, 2004), the strict relative order observed in the Chinese split CP results from the fixed hierarchy among its subprojections. The present article presents an in-depth analysis of the three-layered split CP realized by SFPs in Mandarin Chinese. It provides extensive evidence for their status as projecting and selecting C-heads, on a par with the C-heads in e.g. Romance and Germanic languages. This straightforwardly invalidates the various recurring statements by Biberauer, Holmberg, and Roberts (2007, 2008, 2010, 2014) – based on a superficial discussion of the two Chinese SFPs ma and ne – that SFPs are basically “acategorial” and therefore a quantité négligeable that does not “count” for grammar. The organization of the article is as follows. Section 2 introduces the basic structure of the split CP in Chinese. Section 3 corrects some of the major misconceptions commonly encountered in the literature, among them the alleged optionality of SFPs. Section 4 begins with an examination of the “innermost” SFPs nearest to TP, labelled Low C here. Even though the SFPs realizing Low C are often described as tense/aspect-related, they are clearly located above TP. Section 5 turns to the second-highest level, i.e. ForceP. Section 6 discusses the topmost level, viz. the speaker/hearer-related AttitudeP. As in other languages, the exact semantic contribution of these SFPs is the most difficult to describe. Section 7 illustrates the strict order for co-occurring SFPs and shows it to reflect the hierarchy of the respective layers. Importantly, the resulting split CP is observed in root-contexts only. Section 8 examines the issue of SFPs from a typological perspective and invalidates current proposals claiming the non-existence of head-final CPs in VO languages. Section 9 ­concludes the article.

2 The three-layered split CP in Mandarin Chinese Extending Lee Hun-tak’s (1986) analysis of the yes/no question SFP ma as C to all SFPs, Paul (2008, 2014, 2015) established a three-layered CP for Chinese: Low C < Force < Attitude. This split CP replicates Zhu Dexi’s (1982: 207–213) division of the SFPs into three distributional classes, based on their rigid relative ordering. As to be argued for in this article, this split CP needs to be refined insofar as Low CP and AttitudeP can be further divided into two subprojections.

Chinese SFPs as head-final complementizers 



 51

(1) The three layers in the split root CP C1 (Low C) le currently relevant state láizhe recent past

C2 (Force) éryǐ ‘only’

C3 (Attitude)

baImp (advisative ba)

a softening

baQconfirmation

ei gentle reminder

ma yes/no question ………

ou impatience, surprise ma dogmatic assertion zhene intensifier ……… ne3 exaggeration

ba probability

(N.B. The semantic values indicated for each SFP can give a rough approximation only.)

The first class of SFPs (corresponding to our Low C) occurs nearest to the sentence (TP) and is claimed to express “tense” by Zhu Dexi (1982: 9); it comprises SFPs such as le and láizhe, cf. (2)–(3) below. The SFP éryǐ (not discussed by Zhu Dexi [1982]) also belongs to Low CP, but in a subprojection higher than the one hosting láizhe and le (cf. Section 3 below). The SFPs of the second class (Force) convey notions such as yes/no question (ma), confirmation-seeking question (baQconf) and imperative (baImp), cf. (4) below. The third, “outermost” class of SFPs (Attitude), finally, is explicitly stated to be different from the two other classes, because it involves the speaker’s attitude or feelings; SFPs belonging to this class are e.g. a, ei etc., cf. (5) below. Zhu Dexi (1982: 208) emphasizes that co-occurring SFPs belong to hierarchically different levels. SFPs of the same class are mutually exclusive, such as e.g. le and láizhe, which both belong to the innermost class, cf. (3) below. (2) Tā gāngcái hái zài bàngōngshì láizhe.1 3sg just.now still at office LowC ‘He was in his office just now.’ (3)  Zuótiān xià yǔ le / láizhe yesterday fall rain LowC / LowC / {* le láizhe /*láizhe le }. LowC LowC/ LowC LowC ‘It rained yesterday.’

1 The following abbreviations are used in glossing examples: CLF = classifier; EXP = experiential aspect; NEG = negation; PL = plural; PRF = perfective aspect; PROG = progressive aspect; SG = singular; SUB = subordinator; 1/2/3 = 1st/2nd/3rd person.

52 

 Waltraud Paul and Victor Junnan Pan

(4) a. Nǐ míngnián qù Běijīng ma? 2sg next.year go Beijing force ‘Will you go to Beijing next year?’ b. Nǐ jīntiān xiàwǔ zài lái ba. 2sg today afternoon again come force ‘Please come again this afternoon.’ (5) [CP [TP Jīntiān xīngqīsān ei! Nǐ bié wàngle today Wednesday att 2sg neg forget xiàwǔ děi shàng kè] ei!] afternoon must attend class att ‘Today is Wednesday (mind you)! Don’t forget you have classes in the ­afternoon!’ (slightly changed example from Zhu Dexi 1982: 213) The highest layer established for Mandarin Chinese, AttitudeP, is absent from Rizzi’s (1997, 2004) original hierarchy, but is attested in other languages as well (cf. among others Munaro and Poletto [2006] for discourse-related SFPs in the Italian dialects Pagotto and Veneto, and Haegeman’s [2014] DiscourseP postulated for West-Flemish). In Mandarin Chinese, the Attitude head ei e.g. indicates that the speaker assumes the co-speaker to be up to date concerning the matter at hand, but nevertheless issues a reminder. This is reminiscent of German ja and doch as well as the particle to in Bangla and Hindi.2 As already observed by Zhu Dexi (1982), the SFPs realizing Low C (as “innermost” SFPs) are sensitive to the properties of the TP-internal predicate (cf. Section 3 below for further discussion) and in that respect are comparable to Rizzi’s FiniteP, which entertains a close relationship with the [± finite] nature of the extended verbal projection within TP. Given the controversial nature of the [± finite] distinction in Chinese, the more neutral label “Low C” has been chosen for this layer.

3 Against the so-called optionality of SFPs As will become clear in the remainder of this article, whenever one intends to express the meaning encoded by a given SFP, this SFP is obligatory. For example, the “optionality” of the yes/no question Force head ma only exists insofar as a sentence remains acceptable without it, modulo the associated interpretational difference between a declarative and an interrogative sentence. The same observation

2 Sentence (5) can be roughly translated into German as follows: Heute ist doch Mittwoch! Vergiss ja nicht, dass du heute nachmittag zum Unterricht musst!



Chinese SFPs as head-final complementizers 

 53

holds for all SFPs in any given language, including those realizing AttitudeP. As pointed out by Biberauer, Haegeman, and van Kemenade (2014: 9) it is misleading to characterize discourse particles as “optional”, given that the absence of such a particle inevitably leads to a different interpretation. Accordingly, it does not make much sense to talk about “optionality” as a general feature of SFPs, as Biberauer, Holmberg, and Roberts (2014: 200) do: “In a survey of about 80 VO languages with final question particles, Bailey (2010, 2012) observed that these particles are very often optional (this is true of Mandarin ne and ma, for example).” [emphasis ours, W.P., V.J.P.].3 Importantly, the alleged optionality of ne and ma is precisely not true, as well-known by everybody working on Chinese. This is not meant to gloss over yes/no questions formed by a rising intonation alone (6c) (cf. Lu Jianming 1985: 236; Pan Victor Junnan 2011: 67), which do exist in Chinese, in addition to yes/no questions formed by adding the yes/no question SFP ma to a declarative sentence, cf. (6a)–(6b). (6) a. Tā huì shuō bāfálìyàyǔ. 3sg can speak Bavarian ‘He can speak Bavarian.’ b. Tā huì shuō bāfálìyàyǔ ma? 3sg can speak Bavarian force ‘Can he can speak Bavarian?’ c. Tā huì shuō bāfálìyàyǔ ↑? 3sg can speak Bavarian ‘Can he speak Bavarian?’ However, in many syntactic contexts the option of using intonation to encode a question is excluded. In tag questions with bù shì ma ‘isn’t it (so)?’, the SFP ma is obligatory and cannot be “replaced” by a rising intonation. (7) Nǐ zài Sītújiātè jiāo shū, bù shì *(ma)? 2sg at Stuttgart teach book not be force ‘You teach in Stuttgart, don’t you?’

3 While Biberauer, Holmberg, and Roberts (2014) refer to Bailey’s doctoral dissertation as Bailey (2012), elsewhere it is cited as Bailey (2013) (2013 being the year of submission). In the following, we settle for Bailey (2012/2013) in order to indicate that we refer to the same work as Biberauer, Holmberg, and Roberts (2014). Note that Bailey’s (2012/2013) starting point is that SFPs in VO languages such as Chinese only superficially violate the Final-over-Final Constraint postulated in Biberauer, Holmberg, and Roberts (2010, 2014). Cf. Section 5 below for further discussion.

54 

 Waltraud Paul and Victor Junnan Pan

Similarly, in the presence of wh-indefinite construals ‘something, someone’, a yes/no question requires the presence of ma, because otherwise the sentence – due to the rising intonation – is analysed as a wh question (cf. Victor Junnan Pan 2011: Ch. 5): (8) a. Nǐ xiǎng chī diǎn shénme ↑? 2sg want eat a.bit what ‘What do you want to eat?’ b. Nǐ xiǎng chī diǎn shénme ma? 2sg want eat a.bit what force ‘Do you want to eat a little something?’ Tā pà shéi huì dǎ tā ↑? (9) a. 3sg fear who will beat 3sg ‘Who does he fear will beat him?’ b. Tā pà shéi huì dǎ tā ma? 3sg fear who will beat 3sg force ‘Does he fear that someone will beat him?’ In this respect, Chinese is on a par with English and many other languages, where a yes/no question can be either formed by subject-auxiliary inversion (SAI) or by a rising intonation. Evidently, this does not imply that they are equivalent, or that the existence of rising intonation renders SAI “optional”. Negative Polarity Items, for example, are licensed in SAI only, not in yes/no questions formed by rising intonation: (10) a. *You saw anyone ↑? b. Did you see anyone? Furthermore, as in Chinese, in English as well tag questions cannot be formed by a rising intonation, but require SAI instead: (11) You teach in Stuttgart, don’t you / *you don’t ↑? Concerning the second allegedly optional SFP mentioned by Biberauer, Holmberg, and Roberts (2014: 200), i.e. ne, note first of all that it is not a question particle on a par with ma (pace Lisa Lai-Shen Cheng 1991), a fact again well-documented in the literature (cf. Hu Mingyang 1981: 418; Paris 1981: 389; Li and Thompson 1981: 305; Lin William C. 1984: 220, among others).4 In other words, in a wh question,

4 Given that ne is not a wh-question typing particle à la Lisa Lai-Shen Cheng (1991), it does not qualify as an interrogative Force head and can therefore not be analysed as the overt realization of the null operator present in wh questions, either (contra Aoun and Li [1993]). For a critical appraisal of Lisa Lai-Shen Cheng’s (1991) Clause typing hypothesis in general, cf. Bruening (2007).



Chinese SFPs as head-final complementizers 

 55

cf.  (12), or in an A-not-A polar question (formed by the juxtaposition of the predicate in its positive and negative form; cf. [13]), the Attitude head ne is not obligatory, for the simple reason that ne does not encode the interrogative force. However, if one wants to signal the discourse function associated with ne, which inter alia is to solicit the co-speaker’s attention, rendered here by “listen, and you…”, ne is evidently obligatory (cf. Wu Guo 2005; Li Boya 2006; Victor Junnan Pan 2007, 2011, among others): (12) a. Nǐ zuì xǐhuān hē nǎ 2sg most like drink which ge páizi de déguó píjiǔ? clf brand sub German beer ‘Which brand of German beer do you like most?’ b. Nǐ zuì xǐhuān hē nǎ 2sg most like drink which ge páizi de déguó píjiǔ ne? clf brand sub German beer att ‘Listen, and you, which brand of German beer do you like most?’ (13) a. Tā huì bù huì shuō bāfálìyàyǔ? 3sg can neg can speak Bavarian ‘Can he speak Bavarian?’ Tā huì bù huì shuō bāfálìyàyǔ ne? 3sg can neg can speak Bavarian att ‘And he, can he speak Bavarian?’ Being a head realizing AttittudeP, ne can also select a non-interrogative complement (cf. Section 6 below for further discussion): (14) Déguó yǔyánxuéjiā kě duō *(ne)! German linguist really many att ‘There really are a lot of German linguists!’ As indicated, ne is obligatory in the presence of the speaker-oriented emphatic adverb kě ‘really’. Finally, to round off this discussion of the alleged optionality of SFPs, the Low C le is often required in order to syntactically “close off” a sentence (in the absence of any clearly definable meaning associated with it), the sentence in question simply being unacceptable without it. This is another well-known and broadly documented fact, as evidenced by e.g. the sixty pages in Li and Thompson (1981) devoted to the SFP le alone (also cf. Section 4 below).

56 

 Waltraud Paul and Victor Junnan Pan

(15) Wǒ tài kāixīn *(le)!5 1sg too happy LowC ‘I’m too happy!’ To sum up, the alleged “optionality” of SFPs invoked by Biberauer, Holmberg, and Roberts (2014) is not only incorrect for ne and ma, but for SFPs in Chinese in general, as to be amply documented in the remainder of this article. In order for the associated semantics to be encoded, the SFP must evidently be present.

4 Low CP The SFPs realizing Low C can be further divided into two classes, viz. láizhe and le, on the one hand, and éryǐ ‘only’, on the other. As already stated above, le and láizhe as the “innermost” SFPs are sensitive to the properties of the sentenceinternal extended verbal projection (e.g. aktionsart of the verb, type of negation etc.). Zhu Dexi (1982: 208) grasped this dependence of Low C on TP-internal material by characterizing them as “tense-related” and illustrated this in the minimal pair below: (16) [LowCP [TP Xià yǔ] le ].6 fall rain LowC ‘(Look), it’s raining.’ (Zhu Dexi’s comment: It didn’t rain before.) (Zhu Dexi 1982: 209) (17) [LowCP [TP Xià yǔ] láizhe ]. fall rain LowC ‘It was raining just now.’ (Zhu Dexi’s comment: It just rained.) (Zhu Dexi 1982: 209)

5 The Low C le in sentences such as (15) is obligatory, whereas it is unacceptable in (i). This pair with nearly identical lexical material provides additional evidence to show that the absence/ presence of le is constrained by syntax, not by prosody. Thanks to a reviewer for asking us to be more precise on this point. (i) Wǒ hěn kāixīn (*le)! 1sg very happy LowC ‘I am happy!’ 6 Note that there is the homophonous perfective aspect verb suffix -le to be distinguished from the Low C le. (i) [LowCP [TP Jīntiān xià-le dà yǔ] le]. today fall-prf big rain LowC ‘(Look), it has heavily rained today!’



Chinese SFPs as head-final complementizers 

 57

On the basis of these examples, Zhu Dexi (1982: 209) proposed the following interpretative values: láizhe indicates that the event has occurred in the recent past, le signals that the situation at hand is (conceived of as) new. Naturally, this characterization is not meant to postulate tense as a verbal category for Chinese. It rather attempts to capture the semantic import of the SFP, which is also reflected in the constraints imposed on the complement type (TP or CP-subprojection) each SFP can select, to be examined in detail in the following sections. Since the description of le as signaling a new situation is not appropriate to cover all cases, we adopt Li and Thompson’s (1981: 240) more general characterization of the Low C le as indicating “currently relevant state”. Though “tense-related”, Low C are clearly in the left periphery above TP, not at the vP edge within TP (contra Tang Sze-Wing [1998: 42, 51], among others, who locates Low Cs in the Tense head, with subsequent movement of T0 to C and of the TP-remnant to Spec, Low CP). As evidenced by the interpretation of sentence (18), the negation mei ‘have not’ inside TP only scopes over the vP, whence the indefinite reading for shénme ‘what’. It does not scope over láizhe; instead, it is láizhe that takes wide scope over the entire TP-complement. The same holds for le in (19), which relates the proposition (‘not going to Paris’) to the speech moment and signals that it does no longer hold. (18) [LowCP [TP Wǒ gānggāng méi zuò shénme] [LowC láizhe ]]. 1sg just.now neg do what LowC ‘Just now, I didn’t do anything.’ (recent.past>¬) NOT: ‘It is not the case that [I did anything just now].’ (# ¬>recent.past ) (19) [LowCP [TP Tā [méi [qù Bālí]] [LowC le ]] 3sg neg go Paris LowC ‘He doesn’t go to Paris anymore.’ NOT: ‘He hasn’t been to Paris .’

(le > ¬) (# ¬> le )

The position of le above TP is also confirmed by (20b). Here le signals that previously, unlike the situation this year, he worked during minor holidays only and did take a few days off for Christmas. This is made explicit by the acceptability of kāishǐ ‘start’ in (20b), and its unacceptability in (20a): (20) a. Tā jīnnián lián shèngdànjiē 3sg this.year even Christmas dōu (*kāishǐ) bù fàng jià. all start neg take holiday ‘He doesn’t even take holidays on Christmas this year.’

58 

 Waltraud Paul and Victor Junnan Pan

b. Tā jīnnián lián shèngdànjiē dōu 3sg this.year even Christmas all (kāishǐ) bù fàng jià le. start neg take holiday LowC ‘He doesn’t even take holidays on Christmas (starting from) this year.’ Besides these data clearly showing the position of the Low C above TP, there are also principled objections against locating Low C in T0 (and stipulating obligatory T-to-C movement). Given that Chinese lacks v-to-T movement (Huang C.-T. James 1994), it is rather ad hoc to postulate obligatory T-to-C movement; moreover, it must be excluded that after the verb has raised to Asp0, cf. (21a), it further moves on to T0, picks up the SFP and raises as a complex head to C itself, cf. (21b): (21) a. [TP Tā yǐjīng [AspP [Asp° qù-guò] [vP tqu déguó]]] le. 3sg already go-exp Germany LowC ‘He has been to Germany before.’ b. *[LowCP [TP tā yǐjīng [AspP [Asp° qù-guò] [vP tqu déguó ]]] [C° qù-guò-le]]. go-exp Germany go-exp-LowC 3sg already In any case, as soon as one takes into account the Low C éryǐ, which can roughly be translated as ‘only; this is all I have to contribute’ and which is clearly not “tense-related”, the location of Low C in Tense is completely implausible. (22) [LowCP [TP Wǒ zhǐbúguò shuō shuo] éryǐ ]. 1sg merely say say LowC ‘I’m just talking. (Don’t take me serious.)’ More precisely, as illustrated in the table in (1) above, éryǐ is located in the subprojection of Low CP that is higher than the one hosting le and láizhe, as evidenced by their co-occurrence in the order ‘{le/láizhe}+ éryǐ’(the opposite order being excluded). (23) [LowC2P [LowC1P [TP Tāmen gāngcái zhǐbúguò chǎo jià] [LowC1 láizhe]] [LowC2 éryǐ ]].             3pl just.now   merely  quarrel fight LowC1  LowC2 ‘They were only quarrelling right now (not fighting.), that’s all.’ While Low Cs clearly occupy a TP-external position in the left periphery, there is nevertheless an interaction with TP-internal material including temporal adverbs such as gāngcái ‘just now’: (24) Tā gāngcái hái zài bàngōngshì láizhe / *le. 3sg just.now still at office LowC / LowC ‘He was in his office just now.’



Chinese SFPs as head-final complementizers 

 59

There is a conflict between gāngcái ‘just now, a moment ago’ and le. Gāngcái explicitly locates the event in the past, whereas le relates the very same event to the speech time. However, this is not the case for láizhe, which does not establish such a relation.

5 ForceP The particles typing the clause belong to ForceP: yes/no question ma, confirmation seeking question baQconf and imperative baimp. Note that a Force head not only can, but must determine the nature of the resulting sentence in terms of its respective clause typing features. The presence of ma is obligatory for question formation, on a par with SAI in English, modulo the constrained possibility of forming questions by rising intonation discussed in Section 3 above.7 (25) a. [TP Nǐ shì déguórén]. 2sg be German ‘You are German.’ b. [ForceP [TP Nǐ shì déguórén] [Force° 2sg be German ‘Are you German?’

*(ma)]]? force

As illustrated in (26), the confirmation seeking Force head ba neatly contrasts with the yes/no Force head ma, as evidenced by the different answering possibilities. (26) a. [ForceP [TP Nǐ shì déguórén] [Force° 2sg be German ‘You are German, aren’t you?’ b. Nǐ shì zěnme zhīdào de? 2sg be how know DE ‘How come you know that?’

baQconf]]? force

Importantly, (26b) would be completely infelicitous as answer for the yes/no question with ma in (25b). Contrary to the information seeking question Force head ma, baQconf is compatible with adverbs of the type dàgài ‘probably’. Note that without ba, (27) would be a declarative sentence.

7 As illustrated in (13) above, polar questions can also be in the form of A-not-A questions. For the numerous differences between the latter and the yes/no question with ma, cf. Hagstrom (2005).

60 

 Waltraud Paul and Victor Junnan Pan

(27) [ForceP [TP Nǐ dàgài shì déguórén] [Force° 2sg probably be German ‘You probably are German, aren’t you?’

baQconf /*ma]]? force /force

The data above nicely confirm the contrast between these two types of question SFPs and highlight the different constraints at work for each SFP. The SFP baIMP, homophonous with the confirmation seeking question SFP baQconf, is called “advisative” by Chao Yuen Ren (1968: 807) because of its “softening” effect. Accordingly, an imperative containing baIMP is understood as less harsh an order than the corresponding imperative sentence without baIMP (also cf. Hu Mingyang 1981: 416): (28) [ForceP [TP Qù Kāngsītǎncí niàn shū] [Force° go Konstanz study book ‘Go study in Konstanz!’

ba]]! force

(29) Kuài diǎnr zǒu ba! quick a.bit go force ‘Better hurry up and go!’ (Chao Yuen Ren 1968: 807) (30) Bié chàng {le ba } / {*ba le}! neg sing LowC force / force LowC ‘Better stop singing.’ (Hu Mingyang 1981: 416) Note the rigid ordering between the Low C le and the Force head baIMP, illustrating the hierarchy ‘LowCP < ForceP’. Let us return now to the yes/no question Force head ma. Its analysis as Force head in a head-final CP dominating a head-initial TP and a likewise head-­ initial extended verbal projection challenges Biberauer, Holmberg, and Roberts’ (2014, and earlier versions) Final-over-Final Constraint (FOFC), which excludes the structure where a head-final XP immediately dominates a head-initial YP. (This echoes Dryer’s [1992, 2009] “near-absolute” universal that SFPs are excluded from VO languages such as Chinese.) In their attempt to maintain the FOFC notwithstanding languages such as Chinese, Biberauer, Holmberg, and Roberts (2014: 200–201) implement Bailey’s (2012/2013) analysis which they summarize as follows:

[…] at least some of the apparently FOFC-violating final question particles may actually be initial negative disjunctions of an elided disjunct clause. The structure of these yes/ no questions would be [Q [TP [OR-NOT TP]]], where ellipsis of the second TP, identical



Chinese SFPs as head-final complementizers 

 61

with the first TP, leaves the negative disjunction as an apparently clause-final particle. (Biberauer, Holmberg, and Roberts 2014: 200–201).8

However, putting aside the initial motivation for Bailey’s analysis, viz. to save the FOFC, several problems arise immediately. First, in her attempt to defend the conjunction scenario for ma, Bailey (2012/2013) glosses over the existence of true disjunctive questions with háishi ‘or’ in Chinese, where the second TP can never be elided (whether it is identical with the first TP or not), “stranding” háishì with or without bù: (31) a. Míngtiān tā lái wǒ jiā háishì (tā) bù lái tomorrow 3sg come my home or 3sg neg come wǒ jiā / *háishì (bù)? my home / or neg ‘Will he come to my place or will he not come to my place tomorrow?’ b. Míngtiān nǐ lái wǒ jiā háishì wǒ qù nǐ jiā? tomorrow 2sg come my home or 1sg go your home ‘Will you come to my place or shall I go to your place tomorrow?’ Importantly, ma in such a disjunctive question is completely ungrammatical. (32) a. *Nǐ lái wǒ jiā háishì wǒ qù nǐ jiā ma 2sg come my home or 1sg go your home force b. *Nǐ lái wǒ jiā ma háishì wǒ qù nǐ jiā ma? 2sg come my home force or 1sg go your home force This ungrammaticality holds irrespective of whether there is one ma per clause or one ma for the entire disjunctive structure. Both (32a) and (32b) are excluded because the yes/no question force is in conflict with the disjunctive question force inherent in háishì ‘or’ (cf. the discussion immediately below). In turn, this shows that yes/no questions with ma and disjunctive questions must be distinguished and cannot be analysed uniformly (cf. C.T. James Huang 1982; Huang, Li, and Li 2009). As a result, the yes/no question with ma cannot be derived from a disjunctive structure as postulated by Bailey (2012/2013).

8 As indicated by the position of the Q-operator in the structure [Q [TP [OR-NOT TP]]], the Force head is assumed to take a disjunction of two TPs as complement. As we will show below, cf. (34), this is not the case in Chinese. By contrast, it is the disjunctive operator háishi ‘or’ itself that scopes over the respective propositions.

62 

 Waltraud Paul and Victor Junnan Pan

Furthermore, in addition to the interrogative disjunctor haishi ‘or’, Chinese also has the declarative disjunctor huòzhě ‘or’. (33) a. Nǐ lái wǒ jiā huòzhě wǒ qù nǐ jiā. 2sg come my home or 1sg go your home ‘Either you come to my place or I go to your place.’ b. Nǐ lái wǒ jiā háishì wǒ qù nǐ jiā? 2sg come my home or 1sg go your home ‘Will you come to my place or shall I go to your place?’ As illustrated in (33b), with haishì ‘or’ instead of huòzhě ‘or’ we automatically obtain a disjunctive question. This is different from English and German where the formation of a disjunctive question not only requires ‘or’, but in addition requires SAI. In other words, háishì in Chinese involves both a disjunction and  an interrogative operator. This is the reason for the incompatibility between a disjunctive question and the yes/no question SFP ma observed in (32) above. Finally, Bailey’s (2012/2013) conjunction scenario fails completely in the case of disjunctions where each conjunct bears a sentence final particle, such as the Attitude head ne (discussed in detail in Section 6 below). Also note that the very presence of háishì ‘or’ itself is surprising in her account, given that the SFPs themselves are considered to be disjunctors. (34) Nǐ qù Bólín ne háishì bù qù Bólín ne? 2sg go Berlin att or neg go Berlin att ‘Listen, will you go to Berlin or not?’ Bailey (2012/2013) would have to postulate an underlying disjunction per SFP ne in order to account for its presence on each clause in a disjunction. She also wrongly predicts the acceptability of (35) where háishì ‘or’ – with or without bù ‘not’ – is “stranded” after deletion of the second TP conjunct: (35) *Nǐ qù Bólín ne háishì (bù)? 2sg go Berlin att or neg In addition, the status of the “negative disjunction” OR-NOT in the structure [Q [TP [OR-NOT TP]]] is not clear. If it stands for a conjunction followed by negation after the deletion of the TP, háishì ‘or’ and bù ‘not’ will be stranded, which leads to the ungrammaticality of the sentence, cf. (35); if OR-NOT stands for a conjunction with negation incorporated, which seems to be taken to correspond to the semantics of the yes/no question Force head ma, the sentence is ungrammatical as well, cf. (36).



Chinese SFPs as head-final complementizers 

 63

(36) *Nǐ qù Bólín ne háishi / ma? 2sg go Berlin att or / force As already mentioned in the discussion of examples (12)–(14) above, the SFP ne is not a wh-question “typing particle” à la Lisa Lai-Shen Cheng (1991). Instead, ne is an Attitude head; it can select different types of questions, such as disjunctive questions, cf. (34) above, polar A-bù-A questions, cf. (37), rhetorical questions, cf. (39), as well as declaratives (cf. Section 6 immediately below).9 (37) a. [ForceP [TP Nǐ qù bú qù Fǎlánkèfú]]? 2sg go neg go Frankfurt ‘Do you go to Frankfurt or not?’ b. [AttP [TP Nǐ qù bú qù Fǎlánkèfú] [Att° 2sg go neg go Frankfurt ‘Listen, do you go to Frankfurt or not?’

ne ]]? att

If ne were a wh-question typing particle in the sense of Cheng Lisa Lai-Shen (1991), i.e. obligatory for wh-in-situ languages, its presence in polar A-bù-A questions would force us to treat the latter as a type of wh questions as well, clearly an undesired result. In addition, as is well known, the question interpretation obtains in the absence of ne, both in polar A-bù-A questions, cf. (37a), and wh questions, cf. (38a): (38) a. [ForceP [Force° Op] [TP Nǐ xǐhuān nǎ zuò chéngshì]]? 2sg like which clf city ‘Which city do you like?’ b. [AttP [ForceP [Force° Op] [TP   Nǐ   xǐhuān  nǎ   zuò  chéngshì]] [Att° ne ]]?                 2sg like    which  clf   city att ‘Listen, which city do you like?’ The discourse-related semantics associated with ne can approximately be rendered by ‘listen, look’. Its alleged “clause typing” function is also invalidated by its compatibility with rhetorical questions, which are standardly

9 Interestingly enough, Petrova (2017) observes and corrects a similar misanalysis for the Old High German particles inu and ia. These are in general considered to encode interrogative force, notwithstanding their occurrence in declarative sentences and their optionality in yes/no questions and wh questions. In fact, already in Old High German question formation involved verb fronting and wh fronting.

64 

 Waltraud Paul and Victor Junnan Pan

assumed to have a null negative operator as the head of ForceP (cf. Han and Siegel 1996). (39) a. [ForceP [Force° ¬ ] [TP Tā nǎr huì shuō déyǔ]]?! 3sg where can speak German ‘In which world can he speak German?!’ = ‘He cannot speak German at all!’ b. [AttP [ForceP [Force° ¬] [TP Tā nǎr huì shuō déyǔ]] [Att°  ne ]]?! 3sg where can speak German    att ‘Oh, come on, he cannot speak German at all!’ Even though the exact semantic contribution of ne is difficult to capture, a problem typical of Attitude heads (cf. Section 6 immediately below), it is evident that ne is obligatory if the associated meaning is to be expressed.

6 AttitudeP The SFPs realizing AttitudeP involve the speaker/hearer’s point of view and subjective judgements; this type of SFP is very widespread across language families. As illustrated in the examples below, the exact meaning of the SFPs in AttitudeP is difficult to pin down and strongly depends on the intonation and the context. This is typical of particles relating to the discourse, as inter alia observed for the SFPs in the dialects Pagotto and Veneto from the NorthEastern area of Italy (cf. Munaro and Poletto 2006) and in West-Flemish (cf. Haegeman 2014). In this respect, Chinese is not “exotic” at all (pace Biberauer and Sheehan [2011: 391]), but clearly patterns with other well-studied languages.10 Accordingly, the characteristics of SFPs realizing DiscourseP (the equivalent of AttitudeP) established by Haegeman and Hill (2013) also hold for Attitude SFPs in Chinese. First, AttitudeP does not concern nor affect the truth value of the proposition at hand. This contrasts with the SFPs instantiating ForceP, where as we have seen, baQconfirmation conveys the speaker’s belief that the proposition is true, and ma is a request as to the truth value (yes/no) of the proposition. Attitude SFPs are thus fundamentally distinct from both

10 “[…] Cantonese and Mandarin have an exotic range of sentence-final discourse-particles (SFP) that can be combined to express subtly nuanced (and notoriously difficult to translate) meanings [...]” (Biberauer and Sheehan 2011: 391).



Chinese SFPs as head-final complementizers 

 65

Low C and Force heads, an observation already made by Zhu Dexi (1982: 208), although not elaborated upon. Second, Attitude SFPs indicate the speaker’s commitment to the sentence content; they are interactional and imply the obligatory presence of a hearer. Third, Attitude SFPs are deictic, i.e. they are directly correlated with the speech act, but do not require a preceding utterance as “trigger”. Finally, Haegeman and Hill (2013) concede that it is difficult to determine the precise interpretive properties of Attitude SFPs, even though their semantic import is clearly discernible when comparing sentences with and without them. Starting with the Attitude head ne discussed in the preceding section, its compatibility with a declarative clause is an additional argument against its alleged role as a wh-question clause typer: (40) [AttP [TP Mùníhēi jīnnián dōngtiān méi xià xuě] ne]! Munich this.year winter neg fall snow att ‘Surprisingly, it didn’t snow in Munich this winter!’ The Attitude head ma (henceforth maAtt) implies that the speaker presupposes the hearer not to be up to date and provides a correction of the hearer’s belief, conveying something like ‘this is self-evident’, ‘you should know’, ‘don’t you see?’ (cf. Chao Yuen Ren’s [1968: 801] term “dogmatic assertion”): (41) Tā bù shì Lǎolǐ ma? Ràng tā jìnlái maAtt! 3sg neg be Laoli force let 3sg come.in att ‘Isn’t that Laoli? Let him come in. (Why do I have to tell you ?)’ (Lü Shuxiang 2000: 375) (42) Wǒ shuō jīntiān shì xīngqīsān maAtt! Nǐ shuō bù  shì! 1sg say today be Wednesday att 2sg say neg  be ‘I say it’s Wedndesday today! You say it isn’t!’ (Zhu Dexi 1982: 213) The Attitude head maAtt is clearly distinct from the Force head ma encoding yes/no questions, as generally acknowledged in the literature (cf. Chao Yuen Ren 1968: 800–801; Zhu Dexi 1982: 211–213; Lü Shuxiang 2000: 375– 376, among others) and neatly illustrated by (41), where both SFPs occur in successive sentences. Whereas in the yes/no question, the intonation rises at the end of the sentence and ma cannot be stressed, the second sentence is pronounced with a falling intonation towards the end and maATT can, but  need not be stressed. (Contra Li Boya [2006: 64–65] who postulates a

66 

 Waltraud Paul and Victor Junnan Pan

single ma “mark[ing] a high degree of the strength of the assertive or directive force”.)11 The Attitude head a has a rather complicated morphophonemics depending on the preceding word, which is often reflected in different transliterations: ia, (u)a, (n)a, (ng)a etc. (cf. Chao Yuen Ren [1968: 803], Zhu Dexi [1982: 212], Yang-Drocourt [2007: 192–195] for detailed discussion). For ease of exposition, we gloss over these phonological alternations and use the transliteration a throughout. The Attitude head a is rather ubiquitous and occurs with all kinds of sentence types (declaratives, questions, imperatives, exclamatives), which makes its semantic characterization very difficult. Scholars agree that a conveys the personal implication of the speaker and has a general softening effect; the different interpretations observed for a are then due to the different sentence types it combines with (cf. Chao Yuen Ren 1968: 803–806; Zhu Dexi 1982: 212; Li and Thompson 1981: 313–317; Beutel 1988, among others). For example, Chao Yuen Ren (1968: 804) observes that a question with the SFP a is less blunt than one without it, an effect which can be paraphrased as ‘by the way’ or ‘excuse me’ etc. (43) Nǐ míngtiān chūqù bù chūqù a? 2sg tomorrow go.out neg go.out att ‘(By the way) are you going out tomorrow?’ Likewise, an imperative with the Attitude head a has less the flavour of a command than an imperative without it: (44) Shuō a, bié hàipà a! say att neg be.afraid att ‘Come on, say it, don’t be afraid! In an exclamative, a expresses the emotion of the speaker, which, depending on the sentence meaning, can be anger, astonishment, enthusiasm etc.:

11 Chao Yuen Ren (1968: 801) explicitly addresses the distinction Force head ma vs Attitude head ma and notes the latter as me: “Because particles are in the neutral tone and unstressed, the low vowel a and the midvowel e are indistinguishable. However, in questions ending in ma [i.e. the Force head; W.P., V.J.P.], the sentence intonation is usually fairly high and ends in a slight drawl. It is therefore distinguishable from P5 me [i.e. the Attitude head; W.P., V.J.P.] below, which is always short.” Since nowadays the Attitude head is pronounced ma, we do not follow Chao Yuen Ren (1968), but note it as maAtt. Also note that for reasons poorly understood, the yes/no question force head ma, unlike the imperative force head ba, cf. (51) below, is incompatible with any Attitude head.



Chinese SFPs as head-final complementizers 

 67

(45) Nǐ kàn a, biànhuà duō dà a! 2sg see att change much big att ‘Look, how much everything has changed!’ (Yang-Drocourt 2007: 311) (46) Nǐ yě yào qù a?! 2sg also want go att ‘Oh, you are going, too?!’ Zhene is another Attitude head. It does not only convey an exaggeration by the speaker, but also corrects the presupposition of the co-speaker who underestimates the degree of the property in question. (47) Tā (*fēicháng) piāoliang zhene! 3sg extremely pretty att ‘She is really pretty!’ (48) Tā kě gāo zhene! 2sg unmistakably tall att ‘He is tall indeed!’ Given that zhene already indicates a maximal degree, the presence of a degree adverb such as fēicháng ‘extremely’ is excluded; by contrast, speaker-oriented adverbs such as kě ‘unmistakably’ are acceptable. This sensitivity of the Attitude head zhene to TP-internal material such as adverbs indicates that in the absence of intervening projections, the TP is accessible to a high C-head such as AttitudeP. In addition to the two Force heads baQconf and baImp, there is a third ba realizing AttitudeP: (49) Wàimiàn zài xià yǔ ba. outside prog fall rain att ‘Probably, it is raining outside.’ BaAtt indicates probability; thus (49) would be felicitous when uttered in a room without any windows where the speaker makes a guess based on the noise of the falling rain. Finally, like LowCP, AttP must be further divided into two subprojections. (50) [Att2P [Att1P [TP Sānshí nián qián hái méi yǒu shǔbiāo] ne] ba]. thirty year before still neg have mouse att1 att2 ‘Thirty years ago, very probably there didn’t even exist anything like a ­computer mouse.’

68 

 Waltraud Paul and Victor Junnan Pan

This sentence contains two Attitude heads, ne and ba; it is part of a conversation about video games found on the web (and double-checked with native speakers for its acceptability). The order of ne and ba is fixed, the sequence *ba ne being ungrammatical. Their exact semantic contribution is difficult to tease apart here; however, the import of ba can be rendered by “very probably”.

7 The hierarchy of co-occurring SFPs and the root vs non-root asymmetry in the Chinese CP Having described the three layers, LowCP, ForceP and AttP, we can now proceed to examples where several SFPs co-occur in the same sentence and obey the strict order corresponding to the hierarchy of their respective subprojections. Note that for semantic reasons, it is quasi-impossible to find examples where each of the three layers (LowCP, ForceP and AttP) is realized. In fact, Zhu Dexi (1982: 208) was well aware of this problem. Accordingly, when establishing the relative order between several SFPs, he applied the notion of transitivity: if a given SFP A is shown to precede the SFP B and SFP B precedes the SFP C, then necessarily SFP A likewise must precede C. This same notion of transitivity also underlies Zhu Dexi’s (1982: 208) statement that the relative order always holds, i.e. also when a given SFP position remains empty, as in the combination of the Low C le with the Attitude head ou: (51) Bù zǎo l’ou [= le + ou], kuai zou b’ou [= ba + ou] neg early SFP(fusion) fast go SFP(fusion) ‘It’s getting late! Hurry up and go!’ (Chao Yuen Ren 1968: 808) Given that the Attitude head ou (expressing the speaker’s impatience) consists of a single vowel, it fuses phonetically with the preceding SFP (le and ba) into a single syllable, resulting in l’ou and b’ou. The examples below illustrate the split LowCP followed by either an Attitude or a Force head: (52) Wǒ zhǐbúguò chū chāi le éryǐ a! 1sg merely go.out business.trip LowC1 LowC2 att ‘I only went on a business trip (i.e. it is not that I wouldn’t come back)!’ zì le éryǐ ma? (53) Tā jǐnjǐn dǎ-cuò 3sg only type-wrong character LowC1 LowC2 force ‘Did he only make spelling mistakes?’



Chinese SFPs as head-final complementizers 

 69

Note that the combinations *éryǐ le ma, *le ma éryǐ, *ma le éryǐ, and *éryǐ ma le are all excluded and thus confirm the requirement of a rigid ordering. As illustrated in example (48) above with only one SFP realizing Attitude, in the absence of an intervening projection, the highest C can select the TP directly as its complement. On the other hand, when projections are spelt out, it is always the leftmost SFP which s-selects the complement (TP or CP) to its left, not the next highest SFP. (54) a. Nǐ míngtiān chī xīcān ma? 2sg tomorrow eat western.food force ‘Will you eat western food tomorrow?’ b. *Nǐ míngtiān chī xīcān láizhe 2sg tomorrow eat western.food LowC c. *Nǐ míngtiān chī xīcān láizhe ma? 2sg tomorrow eat western.food LowC force As indicated in (54b), láizhe is incompatible with a TP containing the adverb míngtiān ‘tomorrow’, and the same incompatibility is observed in (54c); accordingly, this unacceptable LowCP cannot serve as the complement for ma. The fact that ma itself allows for a TP complement with míngtiān (54a) cannot save (54c). So far we have limited our discussion to SFPs occurring in matrix sentences, i.e. root contexts. This is important because most C-elements in Chinese are ­prohibited in embedded, non-root contexts.12 More precisely, only Low Cs are acceptable in embedded contexts, cf. (55)–(57), whereas Force and Attitude heads, cf. (58)–(60), are completely excluded here and only acceptable in root contexts (cf. Paul [2014, 2015: Ch. 7] for further discussion). Accordingly, the three-layered split CP ‘LowCP < ForceP < AttitudeP’ exclusively holds for root contexts.

12 The literature on the Chinese C-system, from Lisa Lai-Shen Cheng (1991) up to the more recent studies by Li Boya (2006), Xiong Zhongru (2007), Hsieh and Sybesma (2008), Huang, Li, and Li (2009: 34–35), among others, has so far not acknowledged the systematic character of the root/non-root asymmetry and has at best stated the root-only distribution as the idiosyncrasy of individual SFPs (cf. Li and Thompson 1981: 557; Tang Ting-chi 1988: 363–365, for ma; and Lisa Lai-Shen Cheng 1991; Li Yen-hui 1992: 153, for ne). Whether Chinese has embedded root phenomena is still an open question. For first attempts to address this issue, cf. Lu Peng (2008), Pan Victor Junnan (2012), Paul (2011).

70 

 Waltraud Paul and Victor Junnan Pan

(55) Tā méi gàosù wǒ [ClowP [TP Lǐsì bù qù Bólín] le ] 3sg neg tell 1sg Lisi neg go Berlin LowC ‘He didn’t tell me that Lisi no longer wants to go to Berlin.’ (56) [DP [ClowP [TP Gāngcái dǎ diànhuà] láizhe] de rén] just strike phone LowC sub person dàodǐ shì shéi? in.fact be who ‘Who on earth was the person that called just now?’ (Pan Victor Junnan 2012: Ex. [41]) (57) [TPWǒ shì [CP [TP[-fin] two cónglái bù chōu yān] de ]]. 1sg be ever neg smoke cigarette [-root]C ‘(It is the case that) I have never smoked.’ While the complement of gàosù ‘tell’ in (55) and the relative clause in (56) can either be a CP or a TP, in the propositional assertion construction (57) the copula shi ‘be’ requires a projection headed by the non-root C de as complement (cf. Paul and Whitman 2008). Accordingly, de, which in turn selects a non-finite TP (hence the obligatory subject raising to the matrix TP), can be considered a subordinating C on a par with e.g. that. (58) a. Wǒmen yīqǐ qù baIMP! 1pl together go force ‘Let’s go there together.’ b. Tā yào wǒmen [ pro yīqǐ qù  (* baIMP)]. 3sg ask 1pl together go force ‘He asked us to go there together.’ (59) {*[ForceP[TP Ākiū lái] ma] / [TP Ākiū lái bù lái]} Akiu come force Akiu come neg come méi yǒu guānxi]. neg have relation ‘Whether or not Akiu comes doesn’t matter.’ (60) Tā bù zhīdao [ Ākiū lái bù lái] (*ne)]. 3sg neg know Akiu come neg come att ‘She doesn’t know whether Akiu will come or not.’ (59) illustrates the impossibility for a CP headed by the yes/no question ma to occur in an embedded context; by contrast, the polar A-not-A question is acceptable here.



Chinese SFPs as head-final complementizers 

 71

Note finally that Force heads were not always banned from non-root contexts in the history of Chinese. The yes/no question Force head hū was acceptable in matrix and embedded questions:13 (61) [ForceP [TP Lǔ kě qǔ] hū]? Duì-yuè [pro bù kě]. Lu can take force answer neg can ‘Can the state Lu be annexed? He answered: No, it cannot.’ (Zuozhuan, Min 1; 4th c. BC) (62) Bù zhī [ForceP [TP tiān qì Lǔ] hū]. neg know heaven abandon Lu force ‘I do not know whether Heaven has abandoned the state of Lu.’ (Shǐjì 33; 1542; 2nd c. BC) Apparently, the ban on Force and Attitude heads in non-root contexts is a rather recent development in the history of Chinese. This section has illustrated the well-known rigid ordering among SFPs in the three layers as well as within the subprojections of LowCP and ­AttitudeP.

8 The head-final CP in VO languages from a typological perspective The root vs non-root asymmetry holding for the Chinese C system just discussed is important in two respects. First, it demonstrates that in addition to their s- and c-selectional features, SFPs qua C-heads have to be specified for the feature [±root] as well. The SFP ma e.g. at least has the features [polar question force] and [+root] and c-selects a declarative TP only. The comparison with the SFP baQconfirmation, cf. (27) above, shows that the feature [polar question] needs to be further refined in order to distinguish the true information seeking question encoded by ma from

13 Biberauer, Holmberg, and Roberts (2014: 192) deplore “the paucity of long-term attestation of most of the world’s languages”. While this observation is correct, this is no reason to neglect Chinese with its more than 3000 years of documented history, which shows that the threelayered head-final CP has been attested since the 6th c. BC, against the backdrop of constant VO order (cf. Djamouri, Meisterernst, and Paul [2009], Paul [2008, 2014] and references therein).

72 

 Waltraud Paul and Victor Junnan Pan

the confirmation seeking question with ba, the other features being shared by both SFPs. With respect to their featural make-up, Chinese SFPs are therefore on a par with complementizers such as English that and if, which besides the features for Force (declarative or interrogative, respectively) and [+finite] also encode [-root]. This challenges Huang, Li, and Li’s (2009: 35) view that such complex feature bundles are a characteristic of functional categories in IndoEuropean languages, but not in Chinese. Second, this “syncretic” character makes it impossible to dismiss Chinese SFPs as “categorially deficient”, “syncategorematic” or “acategorial” (cf. Biberauer, Holmberg, and Roberts 2007, 2008, 2010, 2014; Biberauer, Newton, and Sheehan 2009).14 The dismissal is motivated by the intention to maintain the cross-categorial generalization associating the sentence-final position of C with OV languages only (cf. Dryer 1992: 102, 2009).15 This generalization is important for Biberauer, Holmberg, and Roberts because it states the non-existence of structures violating the Final-over-Final Constraint, which precisely excludes the configuration where a head-final phrase dominates a head-initial one. The different attempts by Biberauer, Holmberg and Roberts over the past to come to terms with the head-final CP in the VO-language Chinese, illustrated by their successive reformulations of the FOFC, all have in common that SFPs are likened to “extra-metrical” elements in phonology, i.e. elements not counting for rules, in this case the FOFC. Whether this type of “extra-metrical” element can indeed exist in syntax is not discussed, notwithstanding the far-reaching nature of this claim. Last but not least, the issue of how to account then for the rigid ordering among the different layers within the split CP, the intricate semantics of SFPs and their s- and c-selectional features is never addressed. Nor do Biberauer, Holmberg, and Roberts explain how overt elements that are invisible for constraints of UG, among which the FOFC, can be correctly acquired by the child. In any case, the at first sight solid-looking empirical basis for Dryer’s (1992, 2009) claim shrinks considerably under a more careful scrutiny. When correlating the feature 92a “polar question particle” in the World Atlas of Languages (cf. Dryer 2013a) (the category C not being searchable) with word order, OV and

14 In fact, C heads in Mandarin Chinese turn out to be more “syncretic” than Cs in e.g. German, where according to Struckmeier (2014), TP-internal modal particles spell out “surplus” features of C. In Chinese, by contrast, all these features are located on C alone. 15 Dryer (1992: 102) states that “[…] in fact it may be an exceptionless universal that final complementizers are found only in OV languages. […] complementizers are therefore verb patterners, while the Ss they combine with are object patterners.” In Dryer (2009), the 140 VO languages examined are all said to have a sentence-initial C.



Chinese SFPs as head-final complementizers 

 73

VO languages behave in fact more or less alike, insofar as for both word orders the sentence-initial position (observed for 37 OV and 82 VO languages, respectively) is much rarer than the sentence-final position (observed for 140 OV and 154 VO languages, respectively). Since against the backdrop of Rizzi’s (1997) split CP approach it is likely that many of the question particles can be analysed as complementizers, this considerably weakens Dryer’s (1992: 102, 2009: table [24]) claim that complementizers are verb patterners and that accordingly final complementizers are found only in OV languages. As a result, Chinese is very probably just one example among many where a VO language has a head-final (interrogative) CP. Concerning the position of SFPs in general, going beyond those encoding polar question Force, one has to fall back on the distribution given for adverbial subordinator in WALS (cf. feature 94; Dryer 2013b): 279 VO-languages with a sentence-initial adverbial subordinator vs 2 with a sentence-final subordinator. However, as discussed by Dryer (2013b) himself, the label adverbial subordinator is a cover term for different categories, among them adpositions and case suffixes. Accordingly, the results in WALS cannot be uncritically used as testing ground for the predictions made by the FOFC for the functional category C in embedded contexts (pace Biberauer, Holmberg, and Roberts [2014] and preceding publications). Also note that there are 30 VO languages with mixed order, among them ­Cantonese (a fact not mentioned in Biberauer, Holmberg, and Roberts’ [2014: 183] discussion of the figures in WALS). This is noteworthy insofar as the potential candidates in Mandarin Chinese falling under the label adverbial subordinator can be shown to include sentence-level adverbs and clause-selecting prepositions (cf. Paul 2014, 2015: Ch. 8), i.e. elements preceding the following clause, in addition to the sentence-final non-root C de in the propositional assertion construction (cf. Section 7 above). It is therefore not excluded that these “mixed” order languages precisely illustrate the Chinese case. In the face of the well-established existence of the FOFC violating configuration [[V O] C], the only way out to reconcile SFPs qua Cs in Chinese with the FOFC is an analysis à la Kayne (1994) where a head-final CP is derived from a head-intial CP by movement of the complement TP into the specifier position. However, this account does not work, either, because it shows a number of serious shortcomings, discussed in detail by Bayer (1999: §3) (also cf. Abels and Neeleman 2012). The most obvious problem for a Kayne-style analysis is the impossibility of the SFP to c-command its raised complement. This is clearly an undesirable result, because as demonstrated above, the construal of wh-indefinites ‘something, someone’, crucially depends on the c-command of the TP by the yes/no question Force head ma:

74 

 Waltraud Paul and Victor Junnan Pan

(63) a. Nǐ xiǎng chī diǎn shénme ↑? (= [8a]–[8b] above) 2sg want eat a.bit what ‘What do you want to eat?’ b. Nǐ xiǎng chī diǎn shénme ma? 2sg want eat a.bit what force ‘Do you want to eat a little something?’ Furthermore, as pointed out by Bayer (1999) it remains entirely stipulative that it is the entire TP that must move in order to check the movement triggering feature of C, for such a feature could very well be checked by moving a subconstituent of TP, e.g. the object or the subject. This requirement also runs counter the generally observed non-movability of TP to the left (including local movement). Bayer (1999) therefore concludes that head-final Cs should not be analysed as attractors of TP and that head-final CPs are indeed merged as such.

9 Conclusion The present article has provided extensive evidence to show that SFPs in Chinese are fully-fledged functional heads with a complex feature make-up, on a par with C elements in e.g. Indo-European languages. Chinese SFPs select and project, as evidenced by the strict hierarchy for co-occurring SFPs in the split CP. Structures with a head-final CP and VO order are not only attested in Chinese, but also in other languages (Vietnamese, Niger-Congo languages, etc.). This structure must be merged as such and cannot be derived by postulating movement from a headinitial CP. As a result, the FOFC, which precisely excludes the configuration [[V  O]  C], cannot be a principle of UG, but instead illustrates a statistical observation (cf. Whitman [2008], Paul [2015: Ch. 8], for further discussion).

References Abels, Klaus & Ad Neeleman. 2012. Linear asymmetries and the LCA. Syntax 15(1). 25–74. Aoun, Joseph & Y.-H. Audrey Li. 1993. Essays on the representational and derivational nature of grammar. The diversity of wh-constructions. Cambridge, MA: MIT Press. Bailey, Laura. 2012/2013. The syntax of question particles. Newcastle: Newcastle University dissertation. Bayer, Josef. 1999. Final complementizers in hybrid languages. Journal of Linguistics 35(2). 233–271. Beutel, Helga. 1988. Zur Spezifikation der Satzmodi durch Satzendpartikeln im modernen Chinesisch. In Ewald Lang (ed.), Studien zum Satzmodus, 134–175. Berlin: Akademie der Wissenschaften der DDR, Zentralinstitut für Sprachwissenschaft.



Chinese SFPs as head-final complementizers 

 75

Biberauer, Theresa & Michelle Sheehan. 2011. Introduction: Particles through a modern syntactic lens. The Linguistic Review 28(4). 387–410. Biberauer, Theresa, Liliane Haegeman & Ans van Kemenade. 2014. Putting our heads together: Towards a syntax of particles. Studia Linguistica 68(1). 1–15. Biberauer, Theresa, Anders Holmberg & Ian Roberts. 2007. Disharmonic word order systems and the Final-over-Final Constraint (FOFC). In Antonietta Bisetto & Francisco Barbieri (eds.), Proceedings of XXXIII Incontro di Grammatica Generativa, 86–105. Bologna: Dipartimento di Lingue e Letterature Straniere Moderne. Biberauer, Theresa, Anders Holmberg & Ian Roberts. 2008. Structure and linearization in disharmonic word orders. In Charles B. Chang & Hannah J. Haynie (eds.), Proceedings of the 26th West Coast Conference on Formal Linguistics, 96–104. Somerville, MA: Cascadilla Proceedings Project. Biberauer, Theresa, Anders Holmberg & Ian Roberts. 2010. A syntactic universal and its consequences. Ms., University of Cambridge/Stellenbosch University; Newcastle University; University of Cambridge. Biberauer, Theresa, Anders Holmberg & Ian Roberts. 2014. A syntactic universal and its consequences. Linguistic Inquiry 45(2). 169–225. Biberauer, Theresa, Glenda Newton & Michelle Sheehan. 2009. Limiting synchronic and diachronic variation and change: The Final-over-Final Constraint. Language and Linguistics 10(4). 701–743. Bruening, Benjamin. 2007. Wh-in-situ does not correlate with wh-indefinites or question particles. Linguistic Inquiry 38(1). 139–166. Chao, Yuen Ren. 1968. A grammar of spoken Chinese. Los Angeles, CA: California University Press. Cheng, Lisa Lai-Shen. 1991. On the typology of wh-questions. Cambridge, MA: MIT dissertation. Djamouri, Redouane, Barbara Meisterernst & Waltraud Paul. 2009. Particles in Classical Chinese: Complementizers and topic markers. Paper presented at the 29th International Conference on Historical Linguistics, Radboud University, Nijmegen, August 10–14, 2009. Dryer, Matthew S. 1992. The Greenbergian word order universals. Language 68(1). 81–138. Dryer, Matthew S. 2009. The branching direction theory revisited. In Sergio Scalise, Elisabetta Magni & Antonietta Bisetto (eds.), Universals of language today, 185–207. Berlin: Springer. Dryer, Matthew S. 2013a. Position of polar question particles. In Matthew S. Dryer & Martin Haspelmath (eds.), The World Atlas of Language Structures Online. Leipzig: Max Planck Institute for Evolutionary Anthropology. http://wals.info/chapter/94 (accessed 6 April 2016). Dryer, Matthew S. 2013b. Order of adverbial subordinator and clause. In Matthew S. Dryer & Martin Haspelmath (eds.), The World Atlas of Language Structures Online. Leipzig: Max Planck Institute for Evolutionary Anthropology. http://wals.info/chapter/94 (accessed 6 April 2016). Haegeman, Liliane. 2014. West Flemish verb-based discourse markers and the articulation of the speech act layer. Studia Linguistica 68(1). 116–139. Haegeman, Liliane & Virginia Hill. 2013. The syntacticization of discourse. In Raffaella Folli, Christina Sevdali & Robert Truswell (eds.), Syntax and its limits, 370–390. Oxford: Oxford University Press. Hagstrom, Paul. 2005. A-not-A questions. In Martin Everaert & Henk van Riemsdijk (eds.), The Blackwell companion to syntax, Vol. 1, 173–213. Oxford: Blackwell.

76 

 Waltraud Paul and Victor Junnan Pan

Han, Chung-hye & Laura Siegel. 1996. NPI licensing in adjunct wh-questions In Jan Johnson, Matthew L. Juge & Jeri L. Moxley (eds.), Proceedings of the 22nd Annual Meeting of the Berkeley Linguistics Society, 147–158. Berkeley, CA: Berkeley Linguistics Society. Hsieh, Feng-fan & Sybesma Rint. 2008. Shēngchéng yǔfa lǐlùn hé Hànyǔ yǔqì yánjiū [Generative grammar and the study of sentence final particles in Chinese]. In Yang Shen & Shengli Feng (eds.), Dangdai yuyanxue lilun he hanyu yanjiu [Contemporary linguistic theories and related studies on Chinese], 364–374. Beijing: Shangwu. Hu, Mingyang. 1981. Beijinghua de yuqi zhuci he tanci [Mood particles and interjections in the Beijing dialect]. Zhongguo Yuwen nr. 5, 347–350; nr. 6, 416–423. Huang, C.-T. James. 1982. Logical relations in Chinese and the theory of grammar. Cambridge, MA: MIT dissertation. [Published 1998 by Garland, New York]. Huang, C.-T. James. 1994. More on parametric word order and parametric theory. In Barbara Lust, Margarita Suñer & John Whitman (eds.), Syntactic theory and first language acquisition: Cross-linguistic perspectives. Vol. 1: Heads, projections and learnability, 15–35. Hillsdale, NJ: Erlbaum. Huang, C.-T. James, Y.-H. Audrey Li & Yafei Li. 2009. The syntax of Chinese. Cambridge: Cambridge Unversity Press. Kayne, Richard S. 1994. The antisymmetry of syntax. Cambridge, MA: MIT Press. Lee, Hun-tak Thomas. 1986. Studies on quantification in Chinese. Los Angeles, CA: University of California dissertation. Li, Boya. 2006. Chinese final particles and the syntax of the periphery. Leiden: University of Leiden dissertation. Li, Charles N. & Sandra A. Thompson. 1981. Mandarin Chinese. A functional reference grammar. Los Angeles, CA: California University Press. Li, Yen-Hui Audrey. 1992. Indefinite wh in Mandarin Chinese. Journal of East Asian Linguistics 1(2). 125–155. Lin, William C. 1984. What does the Mandarin particle ne communicate? Cahiers de Linguistique – Asie orientale 13(2). 217–240. Lu, Jianming. 1985. Guānyú xiàndài hànyǔ lǐ de yíwèn yǔqìcí [On question particles in contemporary Chinese]. Yufa yanjiu he tansuo 3. 233–246. Lu, Peng. 2008. Les phrases complexes en chinois. Beijing: Waiwen chubanshe (www.flp. com.cn). [Revised version of Lu, Peng. 2003. La subordination adverbiale en chinois contemporain. Paris: University Paris 7 dissertation]. Lü, Shuxiang (ed.). 2000. Xiàndài hànyǔ bābáicí [Eight hundred words of modern Mandarin]. Beijing: Shangwu. Munaro, Nicola & Cecilia Poletto. 2006. The diachronic path towards particles in the NorthEastern dialects. Nordic Journal of Linguistics 28(2). 247–268. Pan, Victor Junnan. 2007. Interrogation et quantification: le rôle et la fonction des particules et des syntagmes interrogatifs en chinois mandarin. Nantes: University of Nantes dissertation. Pan, Victor Junnan. 2011. Interrogatives et quantification en chinois mandarin: une approche générative. Rennes: Presses Universitaires de Rennes. Pan, Victor Junnan. 2012. Syntactic representation of discourse-configurationality in Mandarin. Paper presented at the 45th Annual Meeting of the Societas Linguistica Europaea, Stockholm University, August 29–September 1, 2012. Paris, Marie-Claude. 1981. Problèmes de syntaxe et de sémantique en linguistique chinoise. Paris: Collège de France.



Chinese SFPs as head-final complementizers 

 77

Paul, Waltraud. 2008. Consistent disharmony: Sentence-final particles in Chinese. Paper presented at the Workshop on Particles, University of Cambridge, October 30–31, 2008. Paul, Waltraud. 2011. Are there embedded root phenomena in Mandarin Chinese? Paper presented at the 44th Annual SLE Meeting, Universidad de la Rioja, Logroño, September 8–11, 2011. Paul, Waltraud. 2014. Why particles are not particular: Sentence-final particles in Chinese as heads of a split CP. Studia Linguistica 68(1). 77–115. Paul, Waltraud. 2015. New perspectives on Chinese syntax. Berlin: de Gruyter Mouton. Paul, Waltraud & John Whitman. 2008. Shi . . . de focus clefts in Mandarin Chinese. The Linguistic Review 25(3/4). 413–451. Petrova, Svetlana. 2017. On the status of the left-peripheral sentence particles inu and ia in Old High German. In Josef Bayer & Volker Struckmeier (eds.), Discourse particles: Formal approaches to their syntax and semantics, 304–331. Berlin: de Gruyter. Rizzi, Luigi. 1997. The fine structure of the left periphery. In Liliane Haegeman (ed.), Elements of grammar, 281–337. Dordrecht: Kluwer. Rizzi, Luigi. 2004. Locality and left periphery. In Adriana Belletti (ed.), Structures and beyond. The cartography of syntactic structures, vol. 3, 104–131. Oxford: Oxford University Press. Struckmeier, Volker. 2014. Ja doch wohl C? Modal particles in German. Studia Linguistica 68(1). 16–48. Tang, Sze-Wing. 1998. Parametrization of features in syntax. Irvine, CA: University of California dissertation. Tang, Ting-chi. 1988. Hanyu cifa jufa lunji [Studies on Chinese morphology and syntax], vol. 1. Taipei: Student Book Co. Whitman, John. 2008. The classification of constituent order generalizations and diachronic explanation. In Jeff Good (ed.), Linguistic universals and language change, 233–252. Oxford: Oxford University Press. Wu, Guo. 2005. The discourse function of the Chinese particle ne in statements. Journal of the Chinese Language Teachers’ Association 40(1). 47–81. Xiong, Zhongru. 2007. Shi…de de goujiàn fēnxī [Syntactic analysis of the construction shi . . . de]. Zhongguo Yuwen 2007, nr. 4, 321–330. Yang-Drocourt, Zhitang. 2007. Parlons chinois. Paris: L’Harmattan. Zhu, Dexi. 1982. Yufa jiangyi [On grammar]. Beijing: Shangwu yinshuguan.

Nathalie Scherf

The syntax of Swedish modal particles In this article, I present a syntactic analysis of the Swedish modal particles ju, väl, nog and visst, which takes the MPs to be two types of deficient adverbs. I argue that the MPs ju and väl are clitic heads in a high functional projection in the IP-domain whereas nog and visst are weak deficient adverbs in a specifier position of a projection below ju and väl. This analysis accounts for the linearization of the MPs, their co-occurrence restrictions, the (im)possibility to occur in the CP-domain, their position with respect to subject and object nouns/pronouns as well as differences between main and subordinate clauses.

1 Introduction The words ju, väl, nog and visst, which are labeled modal particles (MPs) in this article, are frequently analyzed as adverbs, especially modal adverbs, in the literature, cf. Teleman, Hellberg, and Andersson (1999), Beijer (2005), Platzack (2009), but they have also been analyzed as modal particles by Aijmer (1996) and Alm (2012). These four words display features that make them stand out from adverbs both on the level of semantics and syntax. In previous literature, the focus of attention has been on the semantics of the MPs. In this aspect the MPs differ from adverbs by not being a part of the proposition but rather expressing speaker attitudes towards it, cf. Aijmer (1978, 1996). According to Aijmer (1978) the MP nog expresses that the speaker presumes that the proposition is possible, based on some kind of evidence, such as previous experiences, but the speaker is not certain of its truth. The MP visst marks that there is indirect evidence in favor of the proposition and can also be used to mark

Acknowledgements: I would like to thank the audience at the DGfS workshop “The Syntax and Semantics of ­Particles” in Marburg 2014 for valuable comments on this work. I am also very grateful to Marco Coniglio, Markus Egg, Andreas Haida, Johannes Mursell and Sophie Repp as well as anonymous reviewers for the discussions of the topic and for their helpful comments on earlier versions of this paper. Nathalie Scherf: Institut für Anglistik und Amerikanistik, Humboldt-Universität zu Berlin, Berlin, e-mail: [email protected] DOI 10.1515/9783110497151-004



The syntax of Swedish modal particles 

 79

reported speech. The particle ju indicates that the proposition is known to the speaker and the addressee. It is used to remind the hearer of information that is already known. The particle väl expresses uncertainty of the speaker, i.e. the proposition is possible but the speaker is not certain about its truth. Utterances with väl often have a question reading, cf. Aijmer (1996), and in this paper it is translated as a tag question. In this paper I will focus on the syntactic differences between MPs and adverbs. I will present a detailed syntactic description of the particles and discuss how they compare to sentence adverbs. I propose that the Swedish MPs are deficient adverbs of two different types: clitic and weak. This has been proposed for German by Grosz (2005) and Italian by Cardinaletti (2011). However, my proposal differs in some details. In Section 2 the distributional properties of the MPs are described, i.e. the position of the MPs with respect to each other, to adverbs as well as to subjects in the IP-domain. In Section 3 I propose a syntactic analysis of the MPs that draw from the above mentioned analysis by Grosz (2005) and Cardinaletti (2011). The syntactic statuses of the four MPs will be analyzed and an account will be given of their position in main and subordinate clauses, with focus on the correlation between verb movement and the position of the MP as well as the relation between object shift and MPs. In Section 4 additional contrasts between adverbs and MPs are discussed, such as the syntactic difference between MPs and their adverb homonyms and the differences in interaction with information structure. In Section 5 the findings are summarized.

2 Description of the MPs 2.1 Distribution From a crosslinguistic perspective, modal particles show different syntactic properties in different languages. For instance, they are restricted to certain sentence types, e.g. in German ja only occurs in declarative sentences, whereas denn only occurs in interrogatives (Thurmair 1989). The Swedish MPs only occur in declarative sentences, cf. (1).1

1 The following abbreviations are used in glossing examples: cl = clitic (pronoun); def = definite; obj = object; sbj = subject.

80 

 Nathalie Scherf

(1) a. Peter läser ju/ väl/ nog/ visst tidningen i Peter reads ju väl nog visst newspaper.def in köket. kitchen.def ‘Peter reads the newspaper in the kitchen.’ b. *Läser ju/ väl/ nog/ visst Peter tidningen i reads ju väl nog visst Peter newspaper.def in köket? kitchen.def ‘Does Peter read the newspaper in the kitchen?’ c. *Läs ju/ väl/ nog/ visst tidningen i read ju väl nog visst newspaper.def in köket! kitchen.def ‘Read the newspaper in the kitchen!’ MPs are often restricted to certain positions in the clause. For German it has been claimed that MPs are restricted to the IP-domain, cf. Thurmair (1989), whereas Italian and Romanian particles may occur in the IP-and/or the CP-domain, cf. Coniglio and Zegrean (2010). For Swedish I suggest that the MPs fall into two groups: ju and väl pattern with the German particles as they can occur only in the IP-domain, cf. (1a). Nog and visst pattern with a subgroup of the Italian and Romanian particles and can occur in the IP-domain as well as in sentence initial position, i.e. in the CP-domain, cf. (1a) and (2). (2) Nog/ Visst/ *Ju/ *Väl läser Peter tidningen i nog visst ju väl reads Peter newspaper.def in köket. kitchen.def ‘Peter reads the newspaper in the kitchen.’ The two groups ju/väl and nog/visst also display differences within the IPdomain, which only show when the subject of the sentence occurs in the IPinternal position. If the IP-internal subject is a full DP, it cannot precede the MPs ju and väl. Nog and visst, in contrast, can be preceded by a full DP-subject but need not be. Igår såg ju/ väl pojken filmen. (3) a. yesterday saw ju väl boy.def movie.def ‘The boy saw the movie yesterday, as is known./ didn’t he?’ b. *Igår såg pojken ju/ väl filmen. yesterday saw boy.def ju väl movie.def



The syntax of Swedish modal particles 

 81

(4) a. Igår såg nog/ visst pojken filmen. yesterday saw nog visst boy.def movie.def b. Igår såg pojken nog/ visst filmen. yesterday saw boy.def nog visst movie.def ‘I assume/it seems that the boy saw the movie yesterday.’ The data in (3) show the clitic-like properties of ju and väl. They only occur adjacent to the inflected verb in second position. This issue will be addressed more closely in Section 3. With subject pronouns we see a different pattern.2 Unstressed subject pronouns precede or follow the MPs ju and väl, but always precede nog and visst, cf. (5). (5) a. Igår yesterday b. Igår yesterday

såg (han) ju/ väl (han) filmen. saw  he ju väl  he movie.def såg (han) nog/ visst (*han) filmen. saw  he nog visst he movie.def

Stressed subject pronouns follow all the MPs, cf. (6). (6) a. Igår yesterday b. Igår yesterday

såg (*HAN) ju/ väl (HAN) filmen. saw he ju väl  he movie.def såg (*HAN) nog/ visst (HAN) filmen. saw he nog visst  he movie.def

To sum up, in the IP-domain ju and väl can only be preceded by unstressed subject pronouns, whereas nog and visst can be preceded by these pronouns as well as by full DPs. The division into two groups is also present on the level of semantics. The MPs ju and väl both relate to the status of the proposition p, denoted by the sentence in which the MPs occur, in the common ground (CG). Ju marks p as known or presupposed information, i.e. that p is in the CG, and väl expresses that the speaker is biased to add p to the CG and checks with the addressee if p can be added, cf. (7). The sentence in (7a) renders the reading that both speaker and addressee already know that the cake tasted good, i.e. it is in their CG. The sentence in (7b) gives rise to the reading that the speaker thinks that p is true and wants the addressee to confirm this. Subsequently p can be added to the CG. (7) ‘Please take a second piece of the cake!’ a. Den smakar ju väldigt gott. it tastes ju very good. ‘It tastes, as you know, very good.’

2 See Section 3.4 for object pronouns, which display a different pattern yet again.

82 

 Nathalie Scherf

b. Den smakar väl väldigt gott? it tastes väl very good. ‘It taste very good, doesn’t it?’ Nog and visst have two separate meanings depending on the position they take. When they occur in the IP-domain, both particles convey a weak certainty of the speaker about the truth of the proposition, cf. translation in (4). This is a result of the MPs expressing that there is only indirect evidence of some sort for the proposition. Nog indicates that p is an inference drawn from evidence accessible (only) to the speaker, such as experience, cf. (8a). Visst indicates that p is based on indirect evidence accessible to speaker and hearer or reported evidence, cf. (8b). In the CP-domain both particles express the same meaning: a strong ­certainty of the speaker about the truth of the proposition, cf. (8c). When  the MPs occur clause-initially, they carry the default sentence initial stress. Peter är nog/ (# visst) på biblioteket. Det brukar han (8) a. Peter is nog visst at library.def it uses he vara på eftermiddagarna. be on afternoons.def ‘Peter is in the library, I assume. He usually is in the afternoon.’ b. Peter är visst/ (# nog) på biblioteket. Johan sa att Peter is visst nog at library.def John said that han såg honom där. he saw him there ‘Peter is in the library, it seems. John told me that he saw him there.’ c. Nog/Visst är Peter på biblioteket. nog visst is Peter in library.def ‘Peter definitely is in the library.’

2.2 Particle combinations There are further observations that speak for the existence of two groups of particles with differing syntactic statuses, i.e. possible combinations and ­ linear ordering. Two particles of the same type (same syntactic status) cannot combine: – ju väl: The combination of ju and väl results in an unacceptable sentence. (9) *Han har (ju väl)/ (väl ju) redan läst tidningen. he has ju väl väl ju already read newspaper.def



The syntax of Swedish modal particles 

 83

– nog visst: Combining nog and visst results in an unacceptable sentence. (10) *Han har (nog visst)/ (visst nog) redan läst tidningen. he has  nog visst  visst nog already read newspaper.def – ju nog/visst: Ju can be combined with nog or visst. Ju always appears to the left of the other MPs. (11) a. Han har ju nog/ visst redan läst tidningen. he has ju nog visst already read newspaper.def ju nog: ‘As you know, I assume that he has already read the newspaper.’ ju visst: ‘As you know, it seems that he has already read the newspaper.’ b. *Han har nog/ visst ju redan läst tidningen. he has nog visst ju already read newspaper.def – väl nog/visst: Väl can be combined with nog or visst and always appears to the left of nog and visst. (12) a. Han har väl nog/ visst redan läst tidningen. he has väl nog visst already read newspaper.def väl nog: ‘I can assume that he has read the newspaper, can’t I?’ väl visst: ‘It seems that he has read the newspaper, doesn’t it?’ b. *Han har nog/ visst väl redan läst tidningen. he has nog visst väl already read newspaper.def The pattern of combinations of the MPs reflects the two subgroups: ju and väl in one subgroup, nog and visst in the other. Two particles belonging to the same subgroup cannot be combined, but two particles from different groups can. When they are combined, the MPs occur in a fixed linear order in the IP-domain: ju and väl precede nog and visst. Only if nog and visst occur in the CP, can they precede ju and väl, cf. (13). In this example, the meaning of the CP-nog is combined with the question reading induced by väl. The result is a tag question that expresses that the speaker is certain of the truth of the proposition and checks if the addressee agrees with this. boken? (13) Nog har väl Peter köpt nog has väl Peter bought book.def ‘Peter definitely has bought the book, hasn’t he?’ The combinatorial restrictions as well as the ordering in the IP-domain could be due to the semantics of the particles. When looking at MPs from other languages, such as Danish (Hansen and Heltoft 2011) and Japanese (Izutsu and Izutsu 2013), we find that similar restrictions hold, and these are linked primarily to semantics. However, the proposal discussed in the following section will offer a syntactic

84 

 Nathalie Scherf

account for both restrictions, which also explains the syntactic behavior of the MPs discussed in Section 2.1.

3 Syntactic analysis One question frequently addressed in the literature on MPs concerning their syntactic status is whether they are heads or phrases, cf. among others OrmeliusSandblom (1997), Bayer and Obenauer (2011), Jónsson (2017). Reviewing the Swedish data presented above, it becomes clear that they are not one uniform group of words. I propose that the MPs are different sub-types of one syntactic category. Such a sub-type analysis was first proposed by Cardinaletti and Starke (1994) for pronouns in Italian, i.e. a tripartition of pronouns into full, weak and clitic types. This approach has been applied to German MPs by Grosz (2005), who analyzes MPs in Standard German as weak adverbs and the MP dn in Viennese German as a clitic. A similar extension of the analysis has been proposed by Cardinaletti (2011) for Italian MPs. The possibility to apply this type of analysis to Swedish will be discussed in the following sections.

3.1 Deficient elements Cardinaletti (2011) proposes for Italian and German MPs that they are subtypes of adverbs. They have deficient phrase structures compared to the full adverbs. A full, i.e. non-deficient, phrase (of any category) contains the lexical projection (LP), an inflectional (IP), a prosodic (ΣP) and a peripheral projection (CP). Phrases of this type, e.g. strong pronouns and adverbs, can be stressed, modified and coordinated, and they are flexible with regard to their sentence position (IP-or CP-domain). Syntactic deficiency is linked to a reduction of this structure and a loss of some of the features, i.e. the weak type of adverb lacks the C-projection but contains the Σ-projection, and the clitic type lacks the C-projection as well as the Σ-projection encoding prosodic features. In this paper I will not discuss the internal structure of strong and deficient phrases but only adopt the notion of the possibility to create a partition of adverbs into three different categories and compare the features postulated for each type with the features of the Swedish MPs. In Cardinaletti’s (2011) analysis, the weak type of adverbs can carry word stress, and in some cases even sentential stress, cf. Cardinaletti (2011). This possibility to carry word or sentential stress is true for the Swedish MPs nog and visst, but not for ju and väl. Nog is pronounced with a long vowel and hence necessarily carries word stress. Visst can



The syntax of Swedish modal particles 

 85

carry word stress as well as sentence stress.3 The weak type occurs in specifier positions of functional projections in the IP-domain but not in the CP-domain (Cardinaletti 2011). As shown above in (8), nog and visst occur both in the IP-domain and sentence initially, i.e. in a specifier position in the CP-domain. Weak phrases cannot be coordinated or modified. This also applies to nog and visst, cf. (14) and (15). Note, however, that this restriction applies to some sentence adverbials, even those which are full adverb phrases in Cardinaletti’s analysis. The reason for the inability to be coordinated or modified is hence not necessarily linked to structural deficiency, but there might be other, e.g. semantic, reasons for this restriction, i.e. contradictory meanings of the coordinated adverbs4 or that the meaning of the adverb is not modifiable, cf. (14c) and (15c).5 (14) a. *Peter har nog och visst köpt boken. Peter has nog and visst bought book.def b. *Peter har nog och antagligen/ redan köpt boken. Peter has nog and probably/ already bought book.def c. *Peter har definitivt och troligtvis köpt boken. Peter has definitely and presumably bought book.def (15) a. *Peter har väldigt nog köpt boken. Peter has very nog bought book.def Intended reading: ‘I very much assume that Peter has bought the book.’ b. *Peter har väldigt visst köpt boken. Peter has very visst bought book.def Intended reading: ‘It very much seems that Peter has bought the book.’ c. *Peter har väldigt antagligen köpt boken. Peter has very probably bought book.def Intended reading: ‘It is very probable that Peter has bought the book.’

3 Stressed visst: This type of utterance has different context conditions. (i) Peter har VISST köpt boken. Peter has visst bought book.def ‘Peter HAS indeed bought the book.’ This type of visst occurs only in cases of direct contrast between a preceding negative proposition and the sentence containing visst. This usage of visst with sentence stress could possibly be an instance of Verum-focus expressed on the MP, cf. Höhle (1992). 4 For a discussion about semantic restrictions on coordination of adverbs, see Lang (1977). 5 Note that a modification of the equivalent adverbs in German and English is possible. (i) Es ist sehr wahrscheinlich, dass Peter das Buch gekauft hat. it is very probable that Peter the book bought has (ii) It is very probable that Peter bought the book. The reason for these differences is left to further research.

86 

 Nathalie Scherf

The clitic, most deficient, type of adverbs are necessarily deaccented. This phonological feature holds for ju and väl, cf. Aijmer (1978). Cardinaletti (2011) assumes that clitic phrases adjoin to the head of some functional projection, and thus are banned from sentence initial position. This ban holds for both ju and väl, cf. (2b). In this section it was shown that, based on phonological and syntactic criteria, it is possible to classify the Swedish MPs as two different types of deficient elements. However, there are some differences between Cardinaletti’s (2011) and Grosz’s (2005) analyses and the present proposal concerning the derivation of the deficient syntactic structure and with respect to the clause structure, to which we will turn in the following sections.

3.2 Full and deficient adverbs Cardinaletti (2011) claims that Italian and German MPs (which she assumes to be defective versions of adverbs) cannot co-occur with their full adverb counterparts. The reason for this is that the deficient phrases are base generated in the position of the full adverb counterpart but undergo structure deletion. As a result of their deficient structure, they have to move to a higher position in the clause. In Swedish co-occurrence of the deficient type and its full version is possible, cf. (16) with väl and nog, for which adverb counterparts exist. In these examples the first occurrence of väl/nog is unstressed/carries word stress (MP), the second is stressed (full adverb). (16) a. Han vill väl väl. He wants väl good ‘He wants to do good, doesn’t he?’ b. Det får nog vara nog med exempel idag. it may nog be enough with examples today ‘I assume that that will be enough examles for today.’ Taking accent into account, is seems that co-occurrence is possible for MPs and adverbs in German as well. This is also reported by Struckmeier (2014). (17)

Du kannst das SCHON schon machen. (Aber …) you can that MP already make  but ‘You could actually do that now. (But …)’ (Struckmeier 2014: 22)

The stress pattern of the German sentence is contrary to the Swedish examples in (16), but this is not the case for all German MPs when combined with their adverb



The syntax of Swedish modal particles 

 87

counterpart, cf. (18) with stress only on the second occurrence of eben, i.e. the adverb. (18) Er hat es eben EBEN gemacht. he has it MP just.now done ‘He has indeed done it just now.’ On the basis of these data, I conclude that the MPs in Swedish are not subject to a synchronic structure deletion. The possibility of co-occurrence of the MP and its counterpart adverb in two distinct positions in the clause allows for the assumption that the MP is not derived synchronically from the adverb and, as a result thereof, moves to its surface position in the IP, but that the MP is base generated in an IP-position. This is also assumed by Grosz (2005) for the German MPs.

3.3 Phrasal structure of the MPs in the clause In the syntactic account by Cardinaletti (2011) it is assumed that the Italian and German MPs occur in head or specifier positions of functional projections, to which they have moved as a result of their (synchronically derived) deficiency. This analysis cannot be applied to Swedish, as shown above. Grosz (2005) proposes that the deficient phrases are base generated in specifier positions of functional projections in the IP-domain in the sense of Cinque (1999).6 For the weak type, he assumes that it cannot move out of this position. This is a result of its deficient status. In order to compensate for its lack of features, the weak type enters into a local spec-head relation with the head of its hosting functional projection, out of which it cannot move. In example (8) it was shown that the MPs nog and visst can move out of their IP-domain position into the CP-domain. This implies that either they are not deficient elements in the sense of Cardinaletti and Starke (1994), even though they display the distinguishing phonological properties, or that the deficient structure has other effects in Swedish than binding the MPs to their base position in the IP-domain. This question will not be discussed in detail here but will be left open for further research. In Grosz’s (2005) account of the MPs only the clitic particle, base generated in the same position as the weak type, can move out of this position by head-adjunction to the verb. This adjunction compensates for the clitic’s lack of prosodic features. This movement holds for ju and väl as well, as shown in example (2) and will be discussed in detail in Section 3.3.1.

6 Note that Grosz (2005) assumes that the MP denn is base generated in a position that is higher than the highest projection in Cinque’s (1999) hierarchy.

88 

 Nathalie Scherf

I propose to analyze the Swedish MPs as base generated in two separate high functional projections in the IP-domain. Similar assumptions have been made for German by Coniglio (2011), i.e. the German MPs occur in different specifier positions of functional projections in the IP-domain within the hierarchy by Cinque (1999), and by Beijer (2005) for the Swedish MPs ju, väl and nog. However, Beijer assumes that ju, väl, and nog all occur in the same position, i.e. the highest adverb projection in Cinque’s (1999) hierarchy in the IP-domain. In Section 2.2 it was demonstrated that ju or väl can co-occur with nog or visst. Hence, it is not possible that MPs occur in the same position. Instead, I assume two distinct projections, above the adverbs, that host the MPs. I assume that ju and väl occur in the higher of the two projections, nog and visst in the lower, here referred to as MP1 and MP2, see the structure in (19). Ascribing the same syntactic position to ju and väl on the one hand and nog and visst on the other explains why they cannot co-occur and it accounts for their linear ordering in subordinate clauses, as will be shown in Section 3.3.2. Concerning their position in the functional projection, I assume that the MPs ju and väl occur in a head position and nog and visst in a specifier position. This is motivated by ju/väl’s inability to occur in sentence initial specifier positions and the correlation between verb movement and these MPs’ position in main and subordinate clauses. The different statuses of ju/väl and nog/visst also account for their different positions in relation to subjects in the IP-domain. Assuming that the MPs nog and visst occur in the specifier position of MP2 allows for movement to the sentence initial position, i.e. SpecForceP. The reason for this movement, and how the shift of meaning occurs, is still unknown, but it might be driven by some kind of interaction with the illocutionary force operators assumed to be in ForceP, cf. Rizzi (1997). In the next section, I will work out this proposal in detail. 3.3.1 MPs in the matrix clause In examples (3) and (4) above, it was demonstrated that the MPs occur in different positions with respect to subjects. These different positions are only visible in matrix clauses with the subject below the finite verb in second position, or a so called inverted subject. Such a matrix clause is shown in (19). The structure is based on the analyses of the Swedish clause by Platzack (2001) and Holmberg and Platzack (2005), both assuming a split CP. In a main clause with inverted subject, the sentence initial position SpecForceP holds an adverb or another non-subject constituent, the verb surfaces in second position, Force0, and the subject occurs in a position following the verb. According to Holmberg and Platzack (2005), the position of the subject in the matrix clause is flexible, i.e. it occurs in SpecFinP or

The syntax of Swedish modal particles 



 89

SpecIP. This flexibility is visible when adverbs are inserted into the clause.7 If the subject precedes the sentence adverbs, the subject is in SpecFinP, and if the subject follows the adverbs, the subject is in SpecIP, cf. Holmberg and Platzack (2005). Subjects occurring in between two adverbs are assumed to be in SpecIP, with one adverb inserted above IP and one above VP.8 (19) Matrix clause ForceP Adverbs Force’ Nog/visst FinP Verbv ju/väl Subjecti

Fin’ tv

MP1 Ø

MP1’ tju/väl

MP2

nog/visst

MP2’ M0

IP ti

I’ tv

VP ti

V’ tv

Object

As we saw above, the MPs nog and visst can occur sentence initially, in SpecForceP, or in the IP-domain to the left or to the right of an inverted subject. If they are base generated in the specifier position of MP2 above the IP, this is expected. If the subject occurs in SpecIP, it will be preceded by nog and visst in the IP-domain. If the subject occurs in SpecFinP, it will precede these MPs. Ju and väl show a

7 The position of the subject with respect to sentence adverbs seems to be related to the information structure of the sentence, cf. Svenonius (2002). 8 In the structure in (19) the adverb projections have been left out for reasons of space, but they can be inserted above VP and above IP, cf. Platzack (2009).

90 

 Nathalie Scherf

different pattern. They always precede the subject (even in SpecFinP), i.e. the MPs occur directly following the verb in the verb-second (V2-) position. This can be accounted for if we assume that the MPs occur in a surface position above FinP but below Force0. This linearization is only possible if the MP cliticizes onto the verb, and hence moves from its MP1-position with the verb to Force0. Further, a comparison of the positions of the MPs in matrix and subordinate clauses shows that the position of the MPs ju and väl is dependent on verb movement, i.e. an additional argument for a head status of ju and väl. This will be discussed in the following section. 3.3.2 MPs in the subordinated clause In subordinate clauses the MPs, as well as any adverbs, occur to the right of the subject of the clause. (20) Peter säger ... ‘Peter says ...’ a. att Johan ju/ väl/ nog/ visst troligtvis såg that John.sbj ju väl nog visst probably saw Anna igår. Anna.obj yesterday ‘that John MP probably saw Ann yesterday.’ b. *att ju/ väl/ nog/ visst troligtvis Johan såg that ju väl nog visst probably John.sbj saw Anna igår. Anna.obj yesterday c. att Johan (ju/ väl) nog/ visst (*ju/ väl) såg Anna igår. that John.sbj  ju väl nog visst ju väl saw Anna.obj yesterday Following analyses by Holmberg and Platzack (2005) and Platzack (2001, 2009), the structure, shown in (21), of a subordinate clause differs from that of a matrix clause in several aspects: SpecForceP is empty, the complementizer surfaces in Force0, the subject always surfaces in SpecFinP,9 hence preceding any sentence adverbs or MPs, cf. (20), and the verb remains in the VP, following any sentence adverbs. As a result of the lack of verb movement, ju and väl occur in their base position in the IP-domain. They cannot move further left, since there is no verb movement into the C-domain. Hence, they are necessarily preceded by the subject in

9 Holmberg and Platzack (2005: 438) assume that the movement of the subject to SpecFinP is driven by feature checking, and that the complementizer moves from Fin0 to Force0 in order to mark the clause type, i.e. declarative, of the embedded clause.

The syntax of Swedish modal particles 



 91

SpecFinP. The weak adverbs nog and visst also remain in their base position in MP2, below the projection hosting the clitic MPs. The linear ordering of the MPs in the subordinate clause offers an additional argument for two separate projections of MP1 and MP2, cf. (20c). In order for ju and väl to occur above nog and visst in the subordinate clause (i.e. without the possibility to move with the verb), they have to be base generated in a higher position, i.e. in MP1. This is not evident in the main clause, for which the linear ordering of the MPs could be explained by movement of ju and väl alongside the verb to a V2-position. (21) Subordinate clause ForceP Ø

Force’ attc

FinP Fin’

Subjecti tc

MP1 Ø

MP1’ ju/väl

MP2

nog/visst

MP2’ M0

IP ti

I’ Ø

VP ti

V’ Verb

Object

3.4 Interaction with weak and clitic pronouns In (5) and (6) above it was shown that MPs occur in different positions with respect to different pronouns. This was only demonstrated for subject pronouns and for

92 

 Nathalie Scherf

them it holds that all unstressed subject pronouns can precede all MPs. Turning to object pronouns a more fine-grained pattern emerges. There are two reasons for this: First, there are clitic forms as well as weak and strong object pronouns. This is not the case for subject pronouns, which do not have distinct clitic forms, cf. Platzack (2009).10 Second, object pronouns (clitic and weak) can move into a position above the VP by object shift in main clauses if the verb has left the VP (cf. Holmberg’s generalization; Holmberg [1986], among others). Object shift of weak and clitic pronouns might be explained as an instance of deficiency, similar to the MPs. In Italian the obligatory movement of weak and clitic pronouns is linked to deficiency by Cardinaletti and Starke (1994). In main clauses in which object shift is possible, the clitic object pronoun forms precede ju and väl, whereas the weak and stressed forms follow these particles. (22) a. Han gav (’na) ju/väl (*’na) inte (*’na) he gave cl.obj ju väl cl.obj not cl.obj boken. (clitic) book.def b. Han gav (*henne) ju/väl (henne) inte (*henne) he gave her.obj ju väl her.obj not her.obj boken. (weak) book.def c. Han gav (*HENNE) ju/väl (*HENNE) inte (HENNE) he gave her.obj ju väl her.obj not her.obj boken.(strong) book.def ‘He didn’t give her the book, as we know./did he?’ This is expected if the MPs ju and väl are heads. Both the clitic pronoun and the MP cliticize onto the verb and move with it into the CP-domain. Weak pronouns do not cliticize onto the verb, but move out of the VP into a lower position than Force0 (but above MP2, see [25]). As a result, the MPs ju and väl, clitics on the verb in Force0, surface above the weak pronouns. In main clauses with auxiliary

10 The subject pronouns can only take different forms via stress. There are no additional segmental reductions. I analyze subject pronouns as clitics if they precede the MPs. This is only possible for the unstressed pronouns. Such an assumption explains why they can precede the clitic MPs ju and väl, i.e. first the subject pronoun cliticizes onto the verb and then the MP is attached. The same has been proposed for German pronouns by Grosz (2005).



The syntax of Swedish modal particles 

 93

verbs, the lexical verb does not move out of the VP. As a result, object shift is not possible, cf. (23). (23) a. Han har (*’na) inte gett (’na) boken. (clitic) he has cl.obj not given   cl.obj book.def b. Han har (*henne) inte gett (henne) boken. (weak) he has her.obj not given  her.obj book.def ‘He has not given her the book.’ The clitic MPs however, move to a position immediately following the auxiliary verb in such clauses. (24) Då har ( ju) Peter (*ju) inte gett Maria boken. then has ju Peter ju not given Mary book.def This supports the assumption that the MPs are base generated in a position above the VP/AuxP. They can move with the auxiliary and are not blocked by the main verb in V0 as the pronouns are. The MPs nog and visst, being weak MP types, can be preceded by weak and by clitic pronouns, but not by strong ones. The clitic pronouns move with the verb to Force0 and thus precede these MPs. The weak object pronouns move, as a result of object shift, into a high position in the clause preceding nog and visst, presumably above MP2 but below FinP. Only the strong pronouns that do not undergo object shift necessarily follow the MPs (and any sentence adverb if present). (25) a. Han gav (’na) nog/ visst (*’na) inte (*’na) boken. (clitic) he gave cl.obj nog visst cl.obj not cl.obj book.def b. Han gav (henne) nog/ visst (*henne) inte (*henne) boken. (weak) he gave her.obj nog visst her.obj not her.obj book.def c. Han gav (*HENNE) nog/ visst (* HENNE) inte (HENNE) boken. (strong) he gave her.obj nog visst her.obj not her.obj book.def ‘He didn’t give her the book, I assume/it seems.’

4 Distinguishing MPs from adverbs 4.1 Adverbial counterparts For the MPs väl and nog there are full-fledged adverb versions. As we saw in Section 3.2, both types can co-occur, recall (16). The adverb versions display different properties than the MPs, e.g. they occur in a lower position in the clause, see (26a) with nog in the VP. In this position nog can only be interpreted as an adverb. In (26b) nog is in the IP-domain and necessarily is interpreted as a MP.

94 

 Nathalie Scherf

(26) a. Han har ätit nog. he has eaten enough ‘He has eaten enough.’ b. Han har nog ätit. he has nog eaten ‘He has probably eaten.’ In the case of väl, the position in the clause as well as the stress distinguish the MP from the adverb. Only the adverb can be stressed and preceded by MPs or other higher adverbs. In example (27) väl (adverb) follows nog (MP) and is stressed. väl can hence only be interpreted as an adverb (with the adverb meaning well) and not as a MP in this sentence. (27) Han vill nog väl, men det blir fel. he wants nog well but it becomes wrong ‘I assume that he wants to do good, but it turns out wrong.’

4.2 Interaction with information structure The analysis of the MPs as deficient adverbs presented in Section 3 shows that the MPs differ on the level of syntax from sentence adverbs. For ju and väl, assuming they are clitics, it is clear that they have a different syntactic status than full adverbs. For nog and visst, analyzed as weak adverbs, the only possibility to distinguish them from other full sentence adverbs discussed so far is by stress, the MPs only carry word stress. Other aspects, such as coordination and modification do not offer reliable tests to distinguish between MPs and adverbs. An additional aspect that reveals more about the difference between adverbs and MPs is their interaction with information structure. For German it has been claimed that sentence adverbs (e.g. Frey and Pittner 1998) and MPs (cf. Thurmair 1989; Hentschel 1986) mark the border between given and new information in the sentence. A similar claim has been made for Swedish sentence adverbs, cf. Svenonius (2002), i.e. a backgrounded subject precedes the adverb (28a), whereas a focused one follows the adverb, cf. (28b) (B is for “background”, F for “focus”). (28) a. S: ‘What did the boy do yesterday?’ A: #Igår såg antagligen pojkenB filmen. yesterday saw probably boy.def.sbj movie.def.obj A’: Igår såg pojkenB antagligen filmen. yesterday saw boy.def.sbj probably movie.def.obj ‘Yesterday the boy probably saw the movie.’



The syntax of Swedish modal particles 

 95

b. S: ‘Who watched the movie yesterday?’ A: Igår såg antagligen pojkenF filmen. yesterday saw probably boy.def.sbj movie.def.obj A’: #Igår såg pojkenF antagligen filmen. yesterday saw boy.def.sbj probably movie.def.obj ‘Yesterday the boy probably saw the movie.’ This pattern is slightly different for nog and visst. A backgrounded subject can precede the MP, but it does not have to, cf. (29a). A focused subject necessarily follows the MPs, cf. (29b). (29) a. S: ‘What did the boy do yesterday?’ A: Igår såg nog/ visst pojkenB filmen. yesterday saw nog visst boy.def.sbj movie.def.obj A’: Igår såg pojkenB nog/ visst filmen. yesterday saw boy.def.sbj nog visst movie.def.obj ‘I assume/It seems that the boy saw the movie yesterday.’ b. S: ‘Who watched the movie yesterday?’ A: Igår såg nog/ visst pojkenF filmen. yesterday saw nog visst boy.def.sbj movie.def.obj A’: #Igår såg pojkenF nog/ visst filmen. yesterday saw boy.def.sbj nog visst movie.def.obj ‘I assume/It seems that the boy saw the movie yesterday.’ The main difference between sentence adverbs and the MPs with respect to information structure is hence that subjects following MPs are not necessarily interpreted as focused. This speaks in favor of the MPs nog and visst occurring in a position higher than the sentence adverbs. In order to account for the possibility that a backgrounded subject can follow the MPs but necessarily precedes the adverbs, it has to be assumed that only the backgrounded subject moves from SpecIP to another position above the adverbs but below MP2. The question about which position this might be remains open for future work, but note that proposals with two subject positions in the IP-domain – SpecIP and a position above IP (only for topical DPs) – have been made for Swedish by Svenonius (2002) and by Frey and Pittner (1998) for German.11

11 These proposals are based on discussions about the different positions of subjects in relation to adverbs depending on the information structural status of the subject.

96 

 Nathalie Scherf

4.3 Further differences There are some further differences between sentence adverbs and the MPs nog and visst, i.e. the above mentioned shift of meaning that occurs when the MPs move from the IP-domain to sentence initial position, cf. (8). The shift of meaning by shift of position might be related to a possible interaction between the speech act operator in ForceP and the MP, as such sentences are often described as reinforcing the utterances, cf. Aijmer (1996). An elaboration of such an account is left to future research. Note however, that such a shift of meaning is not present with adverbs. Another difference between sentence adverbs and all the MPs is the above, section 2, mentioned restriction of the MPs to declarative sentences. Sentence adverbs are less restricted with respect to the sentence types in which they can occur. A further difference between sentence adverbs and MPs is that sentence adverbs form grammatical answers to questions, but this is not the case with MPs. (30) S: Kommer Anna på besök idag? comes Anna on visit today ‘Will Anna visit us today?’ A: Kanske / antagligen perhaps probably A’: *Ju/ *väl/ *nog/ *visst ju väl nog visst There might be several reasons for this difference. The MPs semantic contribution might be of a type such that they do not constitute felicitous answers. However, the reason might also be syntactic. If ju and väl are clitics, it is expected that they cannot occur on their own. If nog and visst carry only word stress, it is also expected that they cannot occur as the only constituent of a sentence.12 Note that it is unclear which type of nog and visst (IP-domain or SpecForceP) is discussed in (30). Adding material helps us to determine their position in the sentence and shows that both versions of nog can only occur in answers with full sentences, cf. (31a) and (31b). Visst can also occur in fragment answers with the affirmative particle ja, cf. (31c). (31) a. Ja, nog *( gör hon det). yes nog does she it Intended meaning: ‘Yes, of course (she will).’ b. Ja, *( det gör hon) nog. yes it does she nog Meaning: ‘Yes, I assume that she will.’

12 However, see footnote 3 for VISST with sentence stress.



The syntax of Swedish modal particles 

 97

c. Ja, visst (gör hon det). yes visst  does she it ‘Yes, of course (she will).’ d. Ja, *( det gör hon) visst. yes it does she visst Meaning: ‘Yes, the evidence indicates that she will.’ The fragment answer with ja and visst in (31c) can only be analyzed as an elliptical sentence with the sentence initial type of visst. This assumption is motivated by the semantics of such an answer: “ja, visst” can only express strong certainty of the speaker, the meaning ascribed to the CP-version of visst above, not the evidence-meaning expressed by the IP-version. The IP-version of visst can only occur in a full sentence, cf. (31d). Why the fragment answer is not possible for the sentence initial nog is left to future research.

5 Summary In this paper I have proposed to analyze the Swedish MPs as distinct from adverbs and as two subtypes of one category. This analysis is based on the tripartition of word classes proposed by Cardinaletti and Starke (1994). However, the syntactic account of the MPs differs from those proposed for German MPs by Grosz (2005) and for Italian MPs by Cardinaletti (2011). I analyze the Swedish MPs as base generated in two distinct, high functional projections in the IP-domain. This accounts for their combinatorial restrictions and their linear ordering in matrix as well as in subordinate clauses. The two subgroups of MPs differ with respect to their syntactic status. Nog and visst are phrasal, they can carry word stress and move from the IP to a CP-position. Ju and väl, in contrast, are heads. Their position in the clause is dependent on verb movement, i.e. by cliticization onto the verb. In main clauses they occur in a high position, following the verb in Force0. In subordinate clauses they occur in their base position in the IP-domain. The different statuses of the MPs are also evident in relation to pronouns. Only clitic (object) pronouns can precede the MPs ju and väl, whereas clitic and weak object pronouns can precede nog and visst. The difference between sentence adverbs and MPs can be discerned by their position in the clause, especially in relation to the information structure of a sentence. Only sentence adverbs mark the border of backgrounded and focused constituents in the IP-domain. Further, the MPs can also be distinguished from their adverbial homonyms by their position in the clause. The MPs occur in high positions in the IP-domain, whereas the adverbs occur in lower positions in the sentence.

98 

 Nathalie Scherf

References Aijmer, Karin. 1978. Några talaktsfunktioner hos adverb i svenskan [Some speech act functions of Swedish adverbs]. In Olle Josephson, Hans Strand & Margareta Westman (ed.), Förhandlingar vid sammankomst för att dryfta frågor rörande svenskans beskrivning 11 [Descriptions of Swedish 11], 5–20. Stockholm: Stockholm University. Aijmer, Karin. 1996. Swedish modal particles in a contrastive perspective. Language Sciences 18. 393–427. Alm, Maria. 2012. Why not Swedish modal particles? In Johan Brandtler, David Håkansson, Stefan Huber & Eva Klingvall (eds.), Discourse and pragmatics: A festschrift in honor of Valéria Molnár, 29–52. Lund: Lund University. Bayer, Josef & Hans-Georg Obenauer. 2011. Discourse particles, clause structure, and question types. The Linguistic Review 28(4). 449–491. Beijer, Fabian. 2005. On the relative order of adverbs in the I-domain: A study of English and Swedish. Lund: Lund University dissertation. Cardinaletti, Anna. 2011. German and Italian modal particles and clause structure. The Linguistic Review 28(4). 493–531. Cardinaletti, Anna & Michal Starke. 1994. The typology of structural deficiency: On the three grammatical classes. University of Venice Working Papers in Linguistics 4(2). 41–109. Cinque, Guglielmo. 1999. Adverbs and functional heads. A cross-linguistic perspective. Oxford: Oxford University Press. Coniglio, Marco. 2011. Die Syntax der Deutschen Modalpartikeln: Ihre Distribution und Lizenzierung in Haupt-und Nebensätzen. Berlin: Akademie Verlag. Coniglio, Marco & Iulia Zegrean. 2010. Splitting up force evidence from discourse particles. In Lobke Aelbrecht, Liliane Haegeman & Rachel Nye (eds.), Main clause phenomena: New horizons, 229–256. Amsterdam: Benjamins. Frey, Werner & Karin Pittner. 1998. Zur Positionierung von Adverbialen im deutschen Mittelfeld. Linguistische Berichte 176. 489–534. Grosz, Patrick. 2005. dn in Viennese German: The syntax of a clitic version of the discourse particle denn. Wien: Universität Wien MA thesis. Hansen, Erik & Lars Heltoft. 2011. Grammatik over der Danske sprog. Odense: Syddansk Universitetsforlag. Höhle, Tilman N. 1992. Über Verum-Fokus im Deutschen. In Joachim Jacobs (ed.), Information­ sstruktur und Grammatik, 112–141. Opladen: Westdeutscher Verlag. Hentschel, Elke. 1986. Funktion und Geschichte deutscher Partikeln: Ja, doch, halt, und eben. Tübingen: Niemeyer. Holmberg, Anders. 1986. Word order and syntactic features in the Scandinavian languages and English. Stockholm: Stockholm University dissertation. Holmberg, Anders & Christer Platzack. 2005. The Scandinavian languages. In Guglielmo Cinque & Richard S. Kayne (eds.), The Oxford handbook of comparative syntax, 420–448. Oxford: Oxford University Press. Izutsu, Katsunobu & Mitsuko Narita Izutsu. 2013. From discourse markers to modal/final particles. In Liesbeth Degang, Bert Cornillie & Paola Pietrandrea (eds.), Discouse markers and modal particles: Categorization and description, 217–236. Amsterdam: Benjamins. Jónsson, Jóhannes Gísli. 2017. Discourse particles and hvað-exclamatives. In Josef Bayer & Volker Struckmeier (eds.), Discourse particles: Formal approaches to their syntax and semantics, 100–114. Berlin: de Gruyter.



The syntax of Swedish modal particles 

 99

Lang, Ewald. 1977. Semantik der koordinativen Verknüpfung. Berlin: Akademie Verlag. Ormelius-Sandblom, Elisabeth. 1997. Die Modalpartikeln “ja”, “doch” und “schon”: Zu ihrer Syntax, Semantik und Pragmatik. Stockholm: Almqvist & Wiksell. Platzack, Christer. 2001. The vunerable C-domain. Brain and Language 77(3). 364–377. Platzack, Christer. 2009. Den fantastiska grammatiken [The fantastic grammar]. Lund: Språk-och Litteraturcentrum Lunds Universitet. Rizzi, Luigi. 1997. The fine structure of the left periphery. In Liliane Haegeman (ed.), Elements of grammar. Dordrecht: Kluwer. Struckmeier, Volker. 2014. Ja doch wohl C? Modal particles in German as C-related elements. Studia Linguistica 68(1). 16–48. Svenonius, Peter. 2002. Subject positions and the placement of adverbials. In Peter Svenonius (ed.), Subject, explitives and the EPP, 201–242. Oxford: Oxford University Press. Teleman, Ulf, Staffan Hellberg & Erik Andersson. 1999. Svenska Akademiens grammatik, vol. 2: Ord, vol. 4: Satser [The Swedish Academy grammar, vol. 2: words, vol. 4: sentences]. Stockholm: Svenska Akademien. Thurmair, Maria. 1989. Modalpartikeln und ihre Kombinationen. Tübingen: Niemeyer.

Jóhannes Gísli Jónsson

Discourse particles and hvað-exclamatives This paper discusses three discourse particles found in matrix exclamative sentences in Icelandic and their syntactic distribution. Despite some pragmatic differences between them, these particles show uniform behavior in that they are only found in clause-initial position of hvað-exclamatives, a rather surprising fact since Icelandic has many other kinds of exclamatives. This restriction is accounted for by assuming that the discourse particles occupy a high head position in the left periphery from which they select a complement of a particular type. This requirement is satisfied only in hvað-exclamatives because hvað ʻwhatʼ is a general degree word, unlike other items that introduce exclamatives in Icelandic. If correct, this analysis provides further evidence for the head status of discourse particles across languages.

1 Introduction Clausal exclamatives in Icelandic can be introduced by a variety of items, i.e. three wh-words (hvað ‘what/how’, hvernig ‘in what way’, and hvílíkur ‘what a’), clausetyping mikið ‘much’, and various adverbs denoting high degree. These sets of items can also be combined in many ways to form an exclamative (Jónsson 2010). In addition, Icelandic has various discourse particles that occur clause-initially in matrix exclamatives but only in those featuring the wh-item hvað ‘what/how’ (henceforth hvað-exclamatives).1 In this paper, I will discuss the syntactic and pragmatic properties of discourse particles in hvað-exclamatives and develop an analysis to account for their limited syntactic distribution. I will argue that these particles are syntactic heads occupying a high position in the left periphery in which they select a certain kind of ­complement,

1 Whether embedded exclamatives really exist or not is a contentious issue, but see Abels (2004) for much relevant discussion. Acknowledgements: I wish to thank the editors of this volume, Josef Bayer and Volker ­Struckmeier, for all their work on this volume as well as the workshop on which it is based. I am also indebted to two ­anonymous reviewers for constructive feedback which led to various improvements. Jóhannes Gísli Jónsson: Faculty of Icelandic and Comparative Cultural Studies, University of Iceland, Reykjavík, e-mail: [email protected] DOI 10.1515/9783110497151-005



Discourse particles and hvað-exclamatives 

 101

a requirement that is satisfied only in hvað-exclamatives. Therefore, this paper supplies further arguments for the view that discourse particles are functional heads crosslinguistically rather than full phrases (Munaro and Poletto 2003; Bayer and Obenauer 2011; Struckmeier 2014). Furthermore, the data discussed here show that syntax plays a crucial role in the distribution of discourse particles in hvað-exclamatives. The paper is organized as follows. Section 2 introduces the discourse particles to be discussed and describes their basic phonological and pragmatic properties. Section 3 illustrates the syntactic restrictions on the particles, in particular the fact that they are only found in hvað-exclamatives. My analysis of these restrictions is outlined in Section 4, couched within a cartographic model of clausal architecture. The central idea is that hvað, a wh-word denoting degree without any specific value, heads a projection that is different from other exclamatives. Finally, the main points of the paper are summarized in Section 5.

2 The particles 2.1 Integrated vs. independent particles Icelandic has various discourse particles that are used in hvað-exclamatives, e.g. en ‘but’, vá ‘wow’ and æ ‘oh’, as in (1). These three are probably the most common particles in hvað-exclamatives and they will be the focus of this paper. Other particles include guð ‘god’, Jesús ‘Jesus’, jí ‘jii/jy’, mmm, ó ‘oh’, sjitt ‘shit’, úff, and æji ‘ouch’. Since they are syntactically restricted in exactly the same way as the first three, the analysis proposed in Section 4 will carry over to them as well.2 (1) a. En hvað þessi mynd er falleg! but what this picture is beautiful ‘How beautiful this picture is!’ b. Vá hvað þú ert orðinn stór! wow what you are become big ‘How you have grown!’ c. Æ hvað það var leiðinlegt! oh what that was unpleasant ‘What a shame that was!’

2 Please note that all the judgments on the Icelandic examples in this paper are my own unless otherwise noted.

102 

 Jóhannes Gísli Jónsson

Note that hvað does not trigger verb second in exclamatives even though Icelandic is a verb-second language and the same is true for hvernig ‘in what way’ (see Section 3 below). Under the analysis outlined in Section 4, this indicates that both kinds of exclamatives involve a Wh Phrase (WhP), headed by a null complementizer that blocks verb raising. Glossing en as ‘but’ reflects the fact that en is most commonly used as a disjunction in Icelandic. Nevertheless, en expresses shared feelings between speaker and addressee in hvað-exclamatives (see Section 2.2 for further discussion). Thus, it differs from aber ‘but’ in polar exclamatives in German which indicates that something is contrary to expectations (Thurmair 1989: 190–192). The absence of a comma after the particles in (1) signals that they are phonologically integrated into the rest of the clause. This is always the case for en but, as we will discuss shortly, both vá and æ can also be interjections or independent particles set off from the rest of the clause by an intonational break. The integrated particles form a phonological word with hvað, in which case the wh-item hvað has a short unstressed vowel as the second syllable of a phonlogical word. Unstressed hvað has an initial velar stop that is unaspirated just like the velar stop in words like ákveða ʽdecideʼ.3 By contrast, stressed hvað has an aspirated velar stop. Thus, hvað following an integrated particle is pronounced [kvað], which is the unstressed counterpart of [khva:ð]. The pronounciation of the particle en is also affected by a following hvað as the final nasal in (1a) assimilates in place (velar) to the initial stop of hvað. This follows from the fact that en forms a phonological word with hvað. Exclamatives in English are often introduced by interjections like oh, boy, my and wow which are separated from the rest of the clause by a comma intonation (cf. examples like Wow, what this guy can eat!). Both vá and æ can be independent particles set off from the rest of the clause by an intonational break but this is impossible with en.4 This is shown in (2). (2) a. *En, hvað þessi mynd er falleg! but what this picture is beautiful b. Vá, hvað þú ert orðinn stór! wow what you are become big ‘How you have grown!’

3 As the spelling indicates, the initial consonant of hvað (and other words with initial hv-) derives historically from a (velar) fricative. This pronounciation is still preserved dialectally in Iceland. 4 Integrated en also occurs in expressions like En gaman! (lit. ʽbut funʼ) and En þú sniðugur! (lit. ʽbut you cleverʼ) which I take to be truncated exclamatives.



Discourse particles and hvað-exclamatives 

 103

c. Æ, hvað það er óþægilegt! oh what that is inconvenient ‘How inconvenient that is!’ There is a clear pragmatic difference between integrated and independent æ in Icelandic, as the latter conveys a strong negative reaction, almost as if the speaker is suffering physical pain. Thus, æ could also be glossed as ‘ouch’ in (2c). Integrated æ, by contrast, always signals some personal connection between the speaker and the addressee. As a result, it would e.g. be very strange to replace the integrated æ in (1c) with independent æ. Basically, integrated æ is a rather weak expression of sympathy and thus it is appropriate as a reaction to a minor mishap but not as a reaction to really bad news such as somebodyʼs death or a serious accident. The difference between integrated and independent vá is somewhat less conspicuous but it can still be described in similar ways. This can be seen by comparing (1b) and (2b), which contrast only in the use of integrated vs. independent vá. Whereas (1b) is clearly intended as a compliment for the addressee, e.g. a young child that is growing very fast, (2b) is simply an expression of a big surprise with no indication of whether the speaker is happy about this or not. The independent particles have a much freer distribution than the integrated ones. The former are not limited to a particular subtype of exclamatives and they also occur in sentence exclamations: (3) a. Vá, Jón getur talað spænsku! wow John can speak Spanish b. Æ, það var leiðinlegt! oh that was unpleasant Despite superficial similarities, sentence exclamations differ from exclamatives in that they are propositions and have a truth value. This means that one can easily challenge (3a) or (3b) by saying that it is not true. Exclamatives, by contrast, are expressive speech acts and therefore cannot be challenged in this way. A further difference is that exclamatives in Icelandic and many other languages only receive a degree interpretation (Zanuttini and Portner 2003; Castroviejo-Miró 2006; Rett 2008, 2011) but sentence exclamations are not restricted in this way. For instance, (3a) does not have a degree reading; it can only mean that the fact that John can speak Spanish is unexpected. Note also that independent æ is often used in imperatives and wh-questions in which case it indicates annoyance, as exemplified in (4a)–(4b). Æ, farðu nú heim! (4) a. oh go now home ‘For godʼs sake, go home now!’

104 

 Jóhannes Gísli Jónsson

b. Æ, hvað viltu núna? oh what want.you now ‘What the hell do you want now?’ This is in clear contrast to integrated æ which is restricted to hvað-exclamatives like (1c) and always strengthens or reinforces the solidarity between the speaker and the addressee.

2.2 The integrated particles as discourse particles The integrated particles are discourse particles by the criteria laid out by Zimmermann (2011). This is shown by the fact that (i) they do not affect the propositional content of the clause, (ii) they express the speaker’s attitude towards the propositional content of the clause, and (iii) they also say something about the speakerʼs relation to the addressee. The last point entails that the addressee must be known to the speaker and this means e.g. that the addressee cannot be a radio listener. The expressive content of the integrated particles becomes apparent once we examine hvað-exclamatives without these particles. Some relevant examples are provided in (5a)–(5c), which contrast with (1a)–(1c) in that the particles are missing. (5) a. Hvað þessi mynd er falleg! what this picture is beautiful ‘How beautiful this picture is!’ b. Hvað þú ert orðinn stór! what you are become big ‘How you have grown!’ c. Hvað það var leiðinlegt! what that was unpleasant ‘What a shame that was!’ Interestingly, exclamatives like (5a)–(5c) seem to be impossible or at least highly degraded for many speakers, presumably due to the lack of any lexical item expressing the speakerʼs emotional attitude towards the descriptive content of the clause. I find such exclamatives fully acceptable but there is a clear semantic difference between them and hvað-exclamatives with particles. This is shown by the contrast between (5a)–(5c) and (1a)–(1c) as examples like (5a)–(5c) can only convey surprise or noteworthiness, which is the default interpretation of exclamatives across languages (see e.g. Rett 2011 on English). Thus, (5a), unlike



Discourse particles and hvað-exclamatives 

 105

(1a), does not express joy about the beauty of the picture and (5b), unlike (1b), does not express admiration about the size of the hearer. Example (5c) illustrates the same point but it is rather odd pragmatically because it means that the degree of how unpleasant something was surpassed the speakerʼs expectations. Clearly, it is much more natural to express sympathy or commiseration in such cases by using the particle æ, as in (1c). At this juncture, a natural question to ask is how the emotional attitude expressed by en and æ relates to the default interpretation of hvað-exclamatives without particles, surprise or noteworthiness. The answer is that hvað-exclamatives with integrated en and æ are typically used as reactions to new or surprising information.5 Therefore, it would be quite odd to utter (1c) as a reaction to some piece of news that is already known by the speaker. Indeed, this pragmatic condition of immediacy holds of exclamatives in general as Castroviejo-Miró (2006: 182) points out.6 We can now turn to the differences between individual particles. Since vá expresses admiration, it is strange to use it instead of æ in (1c). This is shown in (6a). Conversely, as illustrated in (6b), replacing vá by æ in (1b) is odd since æ typically expresses sympathy or commiseration. (6) a. ?Vá hvað það var leiðinlegt! wow what that was unpleasant b. ?Æ hvað þú ert orðinn stór! oh what you are become big Example (6b) is felicitous in certain contexts, e.g. if the addressee is someone who is growing out of control or he/she is a child that is no longer the cute and well-behaved child that he/she used to be. The particles en and æ are fairly similar but en is more naturally used to express positive feelings like joy or happiness. For this reason, (7b) is more natural than (6b). I also think that en conveys a slightly weaker emotional involvement by the speaker than æ. Thus, (7a) is less pronounced as a commiseration than (1c). (7) a. En hvað það var leiðinlegt! but what that was unpleasant b. En hvað þú ert orðinn stór! but what you are become big

5 This is also true of exclamatives with integrated vá. 6 For an in-depth discussion of the pragmatics of exclamatives from which I have greatly benefited, see Castroviejo-Miró (2008).

106 

 Jóhannes Gísli Jónsson

There is undoubtedly more to say about the pragmatics of the integrated particles but this will have to do here. In the next section, we turn our attention to the syntactic distribution of the particles.

3 Syntactic restrictions on the integrated particles Icelandic has a very rich system of clausal exclamatives (Jónsson 2010). They can be divided into three main classes depending on the content of the left periphery, (i) exclamatives with the clause-typing element mikið ‘much’, (ii) exclamatives with adverbs of high degree, and (iii) exclamatives displaying a wh-item (hvað ‘what’, hvernig ‘in what way’, or hvílíkur ‘what a’). For convenience, these three types will be referred to as M-exclamatives, R-exclamatives and wh-exclamatives, respectively.7 These three types can be combined in three ways. First, M-exclamatives, as in (8a), and R-exclamatives, as in (8b), can be combined as shown in (8c). (8) a. Mikið var þessi leikur spennandi! much was this match exciting ‘How exciting this match was!’ b. Rosalega var þessi leikur spennandi! extremely was this match exciting ‘How extremely exciting this match was!’ c. Mikið rosalega var þessi leikur spennandi! much extremely was this match exciting ‘How extremely exciting this match was!’ Second, as shown in (9a)–(9c), hvað-exclamatives can be combined with exclamatives of the last two kinds but not with M-exclamatives: (9) a. *Mikið hvað þessi leikur var spennandi! much what this match was exciting ‘How extremely exciting this match was!’ b. Rosalega hvað þessi leikur var spennandi! extremely what this match was exciting ‘How extremely exciting this match was!’

7 The term R-exclamative is chosen here because the adverb rosalega ‘extremely’ will be used in this paper to exemplify this kind of exclamative.



Discourse particles and hvað-exclamatives 

 107

c. Mikið rosalega hvað þessi leikur var spennandi! much extremely what this match was exciting ‘How extremely exciting this match was!’ Many speakers do not accept examples like (9b)–(9c) where R-exclamatives combine with hvað-exclamatives. Presumably, this has something to do with the fact that hvað, as a degree word, appears to be completely redundant. Thus, the grammaticality of examples like (9b)–(9c) may depend on hvað not having any degree feature in such cases.8 The discourse particles under discussion are incompatible with M-exclamatives as well as R-exclamatives and any possible combination of these exclamatives, irrespective of word order. This is illustrated below with vá, the pragmatically most appropriate particle to use here, but the other particles behave the same way: (10) a. *Vá mikið var þessi leikur spennandi! wow much was this match exciting b. *Mikið vá var þessi leikur spennandi! much wow was this match exciting (11) a. *Vá rosalega var þessi leikur spennandi! wow extremely was this match exciting b. *Rosalega vá var þessi leikur spennandi! extremely wow was this match exciting (12) a. *Vá rosalega hvað þessi leikur var spennandi! wow extremely what this match was exciting b. *Rosalega vá hvað þessi leikur var spennandi! extremely wow what this match was exciting Since these examples should be pragmatically felicitous, they show quite clearly that the particles have some important syntactic properties that restrict their distribution. As discussed in Section 4, examples (10a)–(10b), (11a)–(11b) and (12a) can be ruled out as violations of the selectional restrictions of the particles as they are not followed by hvað. As for (12b), Section 4 argues that the particles are basegenerated in a position above the landing site for adverbs of high degree like rosalega. This rules out examples like (12b) even if the particle is immediately followed by hvað.

8 In terms of the analysis proposed in section 4, this means that hvað stays in WhP in (9b)–(9c) rather than moving to DegP.

108 

 Jóhannes Gísli Jónsson

A further restriction is that the only wh-item that is compatible with these particles is hvað although Icelandic also has wh-exclamatives based on hvernig ‘in what way’, and hvílíkur ‘what a’. This is exemplified in (13) and (14) below: (13) a. Hvernig Jón plataði alla! how John fooled everybody b. *Vá hvernig Jón plataði alla! wow how John fooled everybody (14) a. Hvílíkur asni hef ég verið! what fool have I been ‘What a fool I have been’ b. *Vá hvílíkur asni hef ég verið! wow what fool have I been The data we have examined in this section show that the particles under discussion are only found in hvað-exclamatives. The particles are incompatible with all other exclamatives, whether they involve clause-typing mikið ‘much’, adverbs of high degree (like rosalega ‘extremely’), the wh-items hvernig ‘in what way’, and hvílíkur ‘what a’, or any possible combination of these elements. As expected, these restrictions do not hold of the independent particles. Thus, (10a), (11a) and (12a) would be grammatical if integrated vá was replaced by independent vá. The same applies to (13b) and (14b).

4 Syntactic analysis 4.1 The basic structure of exclamatives To develop a formal account of the restrictions on the particles under discussion, the syntactic structures associated with different kinds of exclamatives in Icelandic must be explored. An efficient way to do so is to focus on exclamatives that combine as many different elements in the left periphery as possible, as in (15): (15) Mikið rosalega hvað hann var fljótur! much extremely what he was quick ‘How unbelievably quick he was!’ Jónsson (2010) argues that exclamatives like (15) involve a highly articulated structure with three independent phrases in the left periphery. These phrases



Discourse particles and hvað-exclamatives 

 109

occupy the specifier positions of three functional projections: ExclP (Exclamative Phrase), HDegP (High Degree Phrase) and WhP (Wh Phrase): (16) [ExclP Mikið [HDegP rosalega [WhP hvað [TP hann var fljótur! ]]]] much extremely what he was quick As discussed by Jónsson (2010), the order of the left-peripheral elements in clausal exclamatives in Icelandic is completely fixed. Moreover, nothing can intervene between these elements or between hvað and TP. These facts strongly suggest that each item has its own designated position within a limited number of functional projections in the left periphery, as in (16). ExclP in (16) should be seen as a special subcase of ForceP (Rizzi 1997) as it marks clause type (see Westergaard and Vangsnes 2005 and Castroviejo-Miró 2006). HDegP is the landing site for adverbs of high degree, but since these are also evaluative expressions, HDegP is clearly similar to the EvalP of Ambar (2003). More recently, Delfitto and Fiorin (2014) invoke a DegP (Degree Phrase) as the landing site for DPs denoting exceptional degree in nominal exclamatives in Italian. Taking the structure in (16) as a point of departure for our analysis, the first issue to address it how the particles fit into this structure. Since the particles precede hvað in WhP and are not adverbs of high degree, they must be basegenerated in ExclP. This is independently supported by the fact that the particles are not found outside exclamatives in Icelandic. Moreover, as elements of ExclP, the particles distinguish hvað-exclamatives from other exclamatives through syntactic selection (see further below). A popular view of discourse particles holds that they occupy functional space above ForceP (see Haegeman 2014 and references cited there), and thus may select ForceP. The problem with this view is that it would make it difficult to explain why the particles are restricted to hvað-exclamatives. However, these two approaches can be combined by assuming that the particles are in a projection hosting two different features, one relating to clause-typing and the other discourse-related. For convenience, I will continue to refer to that projection as ExclP and its head as Excl. To be more specific, my proposal is that the particles are heads of ExclP rather than specifiers. Being heads rather than full phrases, the particles are capable of selecting a complement of a particular kind (see 4.3 below). Drawing on Bayer and Obenauer (2011), a further argument for their head status is their immobility as well as their phonological shape, i.e. the fact that they are monosyllabic. Bayer and Obenauer (2011) also mention grammaticalization as a diagnostic for the X0 status of discourse particles and this applies to the particles under discussion. Note also that en ‘but’ as a disjunction is clearly a head, so it should be uncontroversial that en as a discourse particle is also a head.

110 

 Jóhannes Gísli Jónsson

Clause-typing mikið is not a discourse particle. Moreover, it cannot fill the specifier position of ExclP when a particle is in Excl, even if the particle is followed by hvað:9 (17) *Mikið vá hvað hann var fljótur! much wow what he was quick ‘How unbelievably quick he was!’ The ungrammaticality of (17) is not surprising as hvað-exclamatives do not combine with mikið, as illustrated in (9a). However, since the presence of a particle in Excl might make a difference, there is no guarantee that (17) is ruled out for the same reason as (9a). Hence, we may have to resort to some syntactic mechanism to prevent ExclP from having an overt head as well as a specifier. One way of deriving the ungrammaticality of (17) is to assume that Excl is endowed with an EPP-feature that can be checked off in two ways: (a) by a particle in Excl, or (b) by mikið if there is no overt element in Excl. Since mikið is only inserted to satisfy an EPP-feature, the result is that a particle in Excl prevents mikið from being licensed in examples like (17).

4.2 The selectional properties of the particles Given the structure in (16), it is natural to claim that particles used in hvaðexclamatives in Icelandic select a WhP.10 This rules out all examples where the particles are immediately followed by anything other than a wh-phrase, including examples like (10a)–(10b), (11a)–(11b) and (12a). Moreover, (12b) is ruled out since the particles are in Excl, a position above HDegP, and thus cannot follow adverbs of high degree. This also derives the ban against recursion of the particles. Stacking these particles, as in (18a)–(18b), is correctly ruled out as the higher particle would not take WhP as its complement. (18) a. *En æ hvað það var leiðinlegt! but oh what that was unpleasant b. *Æ en hvað það var leiðinlegt! oh but what that was unpleasant Note that no such problem arises when the independent particle æ precedes integrated en, as in (19), because independent particles do not require WhP as their complement.

9 Needless to say, (17) is ruled out if mikið is the head of ExclP rather than the specifier. 10 This proposal will be modified in Section 4.3 below.



Discourse particles and hvað-exclamatives 

 111

(19) Æ, en hvað það var leiðinlegt! oh but what that was unpleasant The main empirical challenge for my proposal is how to restrict the relevant particles to hvað-exclamatives, a subclass of wh-exclamatives in Icelandic. As shown in (13b) and (14b), the particles are excluded in wh-exclamatives with hvernig ‘how’, and hvílíkur ‘what a’. We will examine hvernig-exclamatives in Section 4.3 below, but first we will discuss exclamatives with hvílíkur and show how similar they are to R-exclamatives and quite different from hvað-exclamatives as well as hvernig-exclamatives. As discussed by Jónsson (2010), there are good reasons to believe that DPs of the type [hvílíkur + noun] move to HDegP in Icelandic exclamatives. First, hvílíkur denotes high degree, unlike hvað and hvernig, and thus cannot be combined with R-exclamatives:11 (20) * Rosalega hvílíkur asni hef ég verið! extremely what fool have I been ‘What an utter fool I have been’ Second, hvílíkur does not seem to have any wh-features, as shown by the fact that it cannot occur in wh-questions in Icelandic.12 This is exemplified in (21a), which is excluded as a question. By contrast, as shown in (21b), this example is possible as an exclamative: (21) a. * Hvílíkur maður er Jón? what man is John ‘What kind of a man is John?’ b. Hvílíkur maður er Jón! what man is John ‘What a man John is!’ The third reason is that verb second is triggered in exclamatives with hvílíkur, just as in R-exclamatives (and M-exclamatives). This is illustrated in (14a) and (8b). By contrast, there is no verb raising to the left periphery in hvað-exclamatives and hvernigexclamatives, cf. examples like (13a). The conclusion, then, is that exclamatives with DPs hosting hvílíkur are HDegPs and therefore do not fulfill the selectional requirements of the particles.

11 As shown in (9b)–(9c), hvað-exclamatives can be combined with R-exclamatives, at least for some speakers. 12 As discussed by Vangsnes (2008), hvílíkur could be used in wh-questions in Old Icelandic in the sense ʽwhat kind ofʼ.

112 

 Jóhannes Gísli Jónsson

4.3 Distinguishing hvað-exclamatives from hvernig-exclamatives The final issue to resolve is why the particles occur in hvað-exclamatives but not in hvernig-exclamatives. My proposal is that hvað-exclamatives have a unique structure in that WhP is dominated by another projection which hvað moves into to check its degree features. I will label this projection Degree Phrase (DegP). This is an important modification of the structure shown in (17) above. The essential parts of my analysis are sketched in (22), which illustrates the structure of (1a): (22) [ExclP [Excl’ En [DegP hvaði but what

[Deg’ [WhP ti [Wh’ [TP... ]]]]]]]

According to this analysis, hvað-exclamatives are different from all other exclamatives in Icelandic in that they have a DegP and this DegP is selected by the particles. Turning to hvernig-exclamatives, there is no reason to assume that they involve movement to DegP for the simple reason that hvernig is not a degree word and cannot modify gradable adjectives. This is in clear contrast to hvað in exclamatives like (1a)–(1c) where hvað modifies gradable adjectives. Nevertheless, hvernig-exclamatives have a degree reading just like their counterparts in English (see Rett 2008, 2011) even though this reading is not encoded in syntax. This can be seen in (23): (23) Hvernig María lagaði tölvuna! how Mary fixed the.computer This example can only mean that Mary fixed the computer in a way that involves a higher degree of some gradable property (e.g. ingenuity) than the speaker expected. Thus, (23) cannot be felicitously uttered if Mary merely fixed the computer in an unexpected way, e.g. by reciting a magic formula.

4.4 Amount reading vs. kind reading All the examples of hvað-exclamatives we have examined so far involve hvað modifying gradable adjectives like fallegur ‘beautiful’ and stór ‘big’. Additionally, hvað can be an argument, as in (24a). Since exclamatives of this kind express noteworthiness, they are naturally accompanied by vá, as in (24b), but not by en or æ.



Discourse particles and hvað-exclamatives 

 113

(24) a. Hvað hann getur borðað! what he can eat b. Vá hvað hann getur borðað! wow what he can eat As in similar examples in English, (24a) is ambiguous between an amount reading and a kind reading (Zanuttini and Portner 2003 and Rett 2008: 147–149). What is noteworthy in (24a) is either the amount of food that he eats or the kind of food he eats, e.g. if the food is exotic in some way. Interestingly, the kind reading disappears in the presence of vá, as in (24b), and only the amount reading is preserved. Thus, the speaker in (24b) is surprised by the amount of food that he eats but not by the strangeness of the food he eats. This is expected under my analysis because the amount reading, but not the kind reading, arises when hvað denotes something that involves degrees. Thus, the selectional requirements of vá are violated in the kind reading of (24b) because there is no DegP in that case. This contrast between the amount reading and the kind reading is brought out more clearly in examples that unambiguously display only one reading: (25) a. Vá hvað hann getur borðað, þrjú kíló! wow what he can eat three kilos b. *Vá hvað hann getur borðað, hákarl og hrútspunga! wow what he can eat shark and ram’s testacles The measure phrase þrjú kíló ʻthree kilosʼ shows that the intended reading in (25a) is the amount reading and that creates no problems. By contrast, listing strange food items as in (25b) makes the example ungrammatical.

5 Conclusion In this paper, I have discussed clause-initial discourse particles in Icelandic that are integrated into exclamative clauses. These particles have a very limited syntactic distribution as they are only possible in hvað-exclamatives, despite the existence of many other types of exclamatives in Icelandic. Under my analysis, hvað-exclamatives differ from all other exclamatives in Icelandic in that hvað moves to the specifier of DegP, a projection immediately dominated by ExclP. Since the particles are heads of ExclP, they select DegP as their complement and this requirement is only satisfied in hvað-exclamatives. Thus, it is crucial that the particles are functional heads rather than full phrases. To the extent that this analysis is successful, it provides further cross-linguistic evidence for the head status of discourse particles.

114 

 Jóhannes Gísli Jónsson

References Abels, Klaus. 2004. Why surprise-predicates do not embed polar interrogatives. Linguistische Arbeitsberichte 81. 203–221. http://www.uni-leipzig.de/~isruss/reports/abelsLAB81.pdf (accessed 18 May 2016). Ambar, Manuela. 2003. Wh-asymmetries. In Anna-Maria Di Sciullo (ed.), Asymmetry in grammar, 209–249. Amsterdam: Benjamins. Bayer, Josef & Hans-Georg Obenauer. 2011. Discourse particles, clause structure, and question types. The Linguistic Review 28(4). 449–491. Castroviejo-Miró, Elena. 2006. Wh-exclamatives in Catalan. Barcelona: Universitat de Barcelona dissertation. Castroviejo-Miró, Elena. 2008. Deconstructing exclamations. Catalan Journal of Linguistics 7. 41–90. Delfitto, Denis & Gaetano Fiorin. 2014. Exclamatives: Issues of syntax, logical form and interpretation. Lingua 152. 1–20. Haegeman, Liliane. 2014. West Flemish verb‐based discourse markers and the articulation of the speech act layer. Studia Linguistica 68(1). 116–139. Jónsson, Jóhannes Gísli. 2010. Icelandic exclamatives and the structure of the CP layer. Studia Linguistica 64(1). 37–54. Munaro, Nicola & Cecilia Poletto. 2003. Sentential particles and clausal typing in the Veneto dialects. University of Venice Working Papers in Linguistics 13. 127–154. Rett, Jessica. 2008. Degree modification in natural language. New Brunswick, NJ: Rutgers University dissertation. Rett, Jessica. 2011. Exclamatives, degrees and speech acts. Linguistics and Philosophy 34(5). 411–442. Rizzi, Luigi. 1997. The fine structure of the left periphery. In Liliane Haegeman (ed.), Elements of grammar, 281–337. Dordrecht: Kluwer. Struckmeier, Volker. 2014. Ja doch wohl C? Modal particles in German as C‐related elements. Studia Linguistica 68(1). 16–48. Thurmair, Maria. 1989. Modalpartikeln und ihre Kombinationen. Tübingen: Niemeyer. Vangsnes, Øystein A. 2008. ‘What kind of Scandinavian?’ On interrogative noun phrases across North Germanic. Nordic Journal of Linguistics 31(2). 227–251. Westergaard, Marit R. & Øystein A. Vangsnes. 2005. Wh-questions, V2, and the left periphery of three Norwegian dialects. The Journal of Comparative Germanic Linguistics 8(1–2). 117–158. Zanuttini, Raffaella & Paul Portner. 2003. Exclamative clauses: At the syntax-semantics interface. Language 79(1). 39–81. Zimmermann, Malte. 2011. Discourse particles. In Klaus von Heusinger, Claudia Maienborn & Paul Portner (eds.), Semantics. An international handbook of natural language meaning, vol. 2, 2012–2038. Berlin: de Gruyter Mouton.

Hans-Martin Gärtner

Root infinitivals and modal particles. An interim report This interim report revisits and expands on findings by Gärtner (2013) concerning the interaction of German adult root infinitivals (RIs) and modal particles (MPs). The central empirical results concern bare (zu-less) “deontic” RIs and wh-RIs, which restrict occurrence of MPs to subsets of those allowed in imperatives and finite interrogatives, respectively. The specific patterns invite an analysis of the former type by means of Haegeman-style truncation promoting an irrealis operator within a Cinquean hierarchy of functional projections. Additional divergences from imperatives are attributed to a lack of grammaticalized “call on the addressee” leading to constraints on the ordering sources able to interact with covert modality. Wh-RIs are argued to resist epistemic modification at their (extended) propositional level due to internal “non-descriptive” (priority/goal/volition) modality. In addition to a discussion of the absence of polar interrogative and declarative RIs, forays are made into the diagnosis of MP (non-)occurrences within participial and stem-based RI-specimen.

Acknowledgements: For comments and criticism I am very grateful to the audiences at the workshop on “The Syntax and Semantics of Particles” held as part of the annual meeting of the DGfS in Marburg (March 2014), at the LAGB meeting in Oxford (September 2014), and at the workshop on “Redrawing Pragmasemantic Borders” in Groningen (March 2015). Also, I am indebted to three anonymous reviewers for valuable suggestions and criticisms. Special thanks goes to Josef Bayer and Volker Struckmeier for inviting me to Marburg, advising me throughout the process of writing, and editing this volume. Common disclaimers apply. Part of this research was supported by a TÁMOP John-von-Neumann grant (EU/Hungary). Hans-Martin Gärtner: Hungarian Academy of Sciences, Research Institute for Linguistics, Budapest, e-mail: [email protected] DOI 10.1515/9783110497151-006

116 

 Hans-Martin Gärtner

1 Introduction Starting point will be a slightly updated version of the classification of German (adult) root infinitivals (RIs) presented by Gärtner (2013: 203).1 This is given in (1). (1)

Stem −Nom +Nom

−W-Int +W-Int −W-Int +W-Int

   

Infinitive

Participle

−zu

+zu

II

   

()

   

  

( 2) exemplifies cases with the bare infinitive (i−z) (infinitive minus zu) as clausal core. (2b) involves an added nominative DP (+n) and (2c) a fronted W(h)-phrase (+w). (2) a. Hinsetzen (, bitte)! there.sit.inf please ‘Sit down (, please)!’ b. Fußgänger andere Straßenseite benutzen! pedestrians other street.side use.inf ‘Pedestrians are to use the opposite sidewalk.’ c. Wo anfangen? where begin.inf ‘Where to begin?’

〈i−z−n−w〉

〈i−z+n−w〉

〈i−z−n+w〉

Not uncontroversial − whence the marking (ü) − is the theoretical status of root infinitivals based on “zu-infinitives” (i+z) (infinitive plus zu), which are standardly used in exclamations. One example is given in (3). (3) (Skandalös!) So seine Macht zu missbrauchen!  scandalous thus one’s power to misuse.inf ‘(It’s a scandal!) What a misuse of power!’

〈i+z−n−w〉

1 For economy and elegance of expression, “root infinitivals” is used here instead of “non-finite root structures” to render “infinite Hauptsatzstrukturen” (Altmann 1987). The intention is to cover structures with a non-finite verbal core as listed in (1). Thus, “verb-less directives” like Ins Gefängnis mit Hektor Baron! ‘To jail with Hektor Baron!’ (Fries 1983; Jacobs 2008) are not included (but see 2.3.1 for some discussion). Foundational work on German root infinitivals has been carried out by Fries (1983) and Reis (1995, 2003). The issue of modal particles and finiteness is also addressed by Abraham (2017).



Root infinitivals and modal particles. An interim report  

 117

(4) presents cases with a participial (pii) core, typically used as directives. (4b) comes with a nominative DP (+n) while (4a) doesn’t (−n). (4) a. Hereinspaziert!〈pii−n−w〉 here.enter.pcp ‘Do come in!’ b. Freiwillige vorgetreten!〈pii+n−w〉 volunteers forward.step.pcp ‘Volunteers step forward!’ Finally, an example of stem-based (s) “non-inflectionals” is provided in (5). (5) Fasel, fasel.〈s−n−w〉 blab.st blab.st ‘(I’m) Just blabbing.’ According to Bücking and Rau (2013), these come with a peculiar kind of performativity, instrumental in their “depictive” or “simulative” uses in comics and on-line (“chat”) communication. The larger agenda pursued by Gärtner (2013) was to clarify the status of root infinitivals within the class of German sentence types. This involved reviewing their morphosyntactic and semantico-pragmatic properties. Here I will revisit and elaborate on the interaction of RIs and modal particles (MPs), the latter having proven highly useful in establishing finer aspects of the form and function of sentence types (cf., e.g., Thurmair 1989, 2013). In particular, Section 2 will be devoted to bare (“zu-less”) 〈i−z〉-RIs. Most importantly it will be shown that RIs are “impoverished” as compared to full-fledged imperatival and wh-interrogative counterparts in hosting fewer modal particles. Some ways of accounting for this phenomenon will be addressed. Section 2.3.1 discusses and argues against the existence of declarative and polar interrogative RIs. Section 3 briefly surveys modal particles in participial 〈pii〉-RIs and stem-based 〈s〉-RIs. It is concluded here that the set of MPs in the former coincides with the one in 〈i−z±n−w〉-RIs while MPs are presumably banned from the latter.

2 Modal particles in root infinitivals of type 〈I−Z〉 While Altmann (1987: 41) hypothesized that modal particles are entirely absent from root infinitivals, he subsequently revised this judgment (Altmann 1993: 1013), citing the examples in (6) from Thurmair (1989: 65–66).

118 

 Hans-Martin Gärtner

(6) a. Alle mal herhören!〈i–z+n–w〉 all mp here.hear.inf ‘Everyone listen!’ b. Wozu denn hingehen?〈i–z–n+w〉 wherefore mp there.go.inf ‘Why go there?’ Note that the modal particle mal in (6a) goes back to the temporal or frequency adverb einmal ‘sometime’, ‘once’, “mitigating” the force of the request by inviting non-immediate or minimal compliance. Denn in (6b) relates to the homonymous causal/reason conjunction. It can be taken to flag an abstract “inferentiality” relation between the utterance it is used in and the common ground or larger shared store of (contextual) knowledge, with various accompanying pragmatic effects.2 Below we are going to see that 〈i−z−w〉-RIs and 〈i−z+w〉-RIs raise related but sufficiently distinct questions, so we’ll deal with them separately. Section 2.3 discusses and argues against the existence of declarative and polar interrogative RIs.

2.1 〈−w 〉-RIs Gärtner (2013) undertakes a more systematic exploration of the inventory of modal particles in root infinitivals. This builds on work by Coniglio (2006, 2011), who adapts the technique of establishing adverbial hierarchies promoted by Cinque (1999) to the domain of modal particles. In particular, given the affinity of 〈i−z−w〉-RIs and imperatives, a simple comparison of hierarchies has suggested itself, the result of which is shown in (7) and (8).3 (7) MP-hierarchy (IMP-compatible) (cf. Coniglio 2006: 80) doch > halt > DOCH > nur (nicht) > bloß (nicht) > ruhig > {mal,JA}

2 König, Stark, and Requardt (1990) provide a good first orientation regarding the use of modal particles. For further in depth study, one may begin by consulting work by Zimmermann (2011) and Grosz (2016). 3 According to Thurmair (1989: 110–111) DOCH is not an MP but an “Affirmationsadverb,” pace Meibauer (1994: Ch. 4).



Root infinitivals and modal particles. An interim report  

 119

(8) MPs acceptable in 〈i−z±n−w〉-RIs (cf. Gärtner 2013: 224) doch > halt > DOCH > nur (nicht) > bloß (nicht) > ruhig > {mal,JA} Evidence for the diagnosed gap is provided in (9) and (10).4 (9) a. Leg die Bücher doch auf den Tisch!〈imp〉 put.imp the books mp on the table ‘Put the books on the table, will you!’ b. Die Bücher  (* doch) auf den Tisch legen!〈i–z–n–w〉 the books mp on the table put.inf ‘Put the books on the table!’5 Doch in (9) signals that the speaker is generically taking into account and ready to overrule reasons in conflict with the addressee complying with the request. den Tisch halt zum Sperrmüll!〈imp〉 (10) a. (OK,) Bring bring.imp the table mp to.the bulky.waste ‘(OK,) Take the table to the scrap yard, if you must!’ b. (OK,) Den Tisch (* halt) zum Sperrmüll bringen!〈i–z–n–w〉 the table mp to.the bulky.waste bring.inf ‘(OK,) Take the table to the scrap yard!’

4 Counterparts of (9b) improve if doch occurs in combination with another particle like nicht ‘not’ or bitte ‘please’: (i) a. Die Schuhe doch nicht auf den Teppich stellen! the shoes mp not on the carpet put.inf ‘Come on! Don’t put the shoes on the carpet!’ b. Freiwillige doch bitte vortreten! volunteers mp please forward.step.inf ‘Volunteers step forward please!’ (ia) works well as an exclamatory complaint, not as a(n initiating) directive move. These facts can be accounted for by assuming that doch directly attaches to and operates on the particles following it here. The issue of MP combinations (Thurmair 1991; see also Müller 2017) needs to be dealt with separately. 5 This judgment is shared by Fortuin (2003: 36) and van Olmen (2009: 148) for Dutch toch.

120 

 Hans-Martin Gärtner

The use of halt in (10) marks unassailability or inevitability, and thus contributes to the concessive move performed by means of an imperative in (10a).6 In the light of the contrast in (7)/(8), the obvious next step is to ask the question formulated most succinctly by Abraham (1995): Wieso stehen nicht alle Modalpartikel in allen Satzformen? ‘Why aren’t all modal particles found in all sentence types?’. According to one school of thought, grammatical form is decisive. Thus, Thurmair (1993: 39) assumes that “the occurrence of modal particles is solely dependent on the form type of a sentence, i.e., each modal particle subcategorizes for one or more form types” [translation H.M.G.]. Put in its crudest form, as done by the schemata in (11), this seems to boil down to a restatement of the facts.7 (11) a. doch [+IMP] b. halt [+IMP] c. bloß [+IMP]/[+INF] d. ruhig [+IMP]/[+INF] A more systematic picture, however, emerges if one views (11) from the perspective of Han (2000). Here, the following structural bipartition (Han 2000: 128) is proposed for capturing imperatives and their (obligatory) substitutes in negative environments (“negative imperatives,” “prohibitives”), the latter attested, for example, in Italian (infinitive) and Portuguese/Spanish (subjunctive).

6 Concessive moves like the one envisaged here require some conversational back-and-forth exchange, so these are not out-of-the-blue utterances. Thanks to an anonymous reviewer, I realize that the jugdments concerning halt may be more subtle and subject to speaker variation, the latter perhaps even due to (ongoing) language change. Here I record my own intuitions, in particular, the intuition that there is a contrast between imperatives and RIs as well as, within RIs, between doch/halt and the remainder of the MPs. Apparent counterexamples of the kind in (i) (Daniel Gutzmann, p.c.) are most likely reducible to term answers, i.e., discourse ellipsis (see also the discussion in Section 2.3.1): (i) Halt mal länger schlafen! mp mp longer sleep.inf ‘You could sleep longer.’ (i) would be licensed in a situation where someone else complains about tiredness and asks (or implicitly “raises” the question) “What can/shall I do?”. Note that we are dealing again with a particle combination (halt mal). 7 According to Thurmair (1993: 39), meaning is only an additional filter, excluding MP occurrences in context.

Root infinitivals and modal particles. An interim report  



 121

(12) a. Imperatives b. Subjunctives/Infinitivals



CP

CP C'

C' C0 Verbi

directive irrealis

C0

IP ... ti ...

  

irrealis

IP ... Verb ...

If the featural decomposition in (12), indicating a subsumptive relation between imperatives and subjunctives/infinitivals, is on the right track, (11) can be recast as in (13).8 (13) a. doch [+directive] b. halt [+directive] c. bloß [+irrealis] d. ruhig [+irrealis] The situation can be interpreted such that in imperatives, world-to-word direction of fit (d-o-f) is hardwired, while in root infinitivals it is due to “pragmatic inference” (Han 2000: §4.3.2; Reis 1995). At the same time, among the particles in (7) that also occur in declaratives, doch and halt are the ones that go with clear word-to-world d-o-f there (Thurmair 1993: 28). (14) illustrates this. (14) Du lässt mich (#doch)/(#halt) jetzt in Ruhe!〈DEC〉 mp you let me mp now in peace ‘Leave me alone now!’ Insertion of doch or halt into the declarative in (14) makes using it as an indirect request very awkward. Thus, these particles can be taken to “block” d-o-f-reversal in the absence of hardwiring. (13) should therefore be considered to be stating (ingredients of yet to be properly formulated) licensing requirements rather than standard lexical selection information.

8 Labeling the core imperatival feature “directive” is not unproblematic, in particular, if the feature is to be interpreted in the standard Searlean way (cf., e.g., Kaufmann 2012; Wilson and Sperber 1988). For some critical remarks about uncritical usages of terms like “subjunctive” and “irrealis” I refer the reader to van der Auwera and Schalley (2004). I refrain from going more deeply into discussing/criticizing the approach by Han (2000), as this would lead us too far afield.

122 

 Hans-Martin Gärtner

Let us now return to (7)/(8). As is well-known, the Coniglio-Cinque hierarchy is meant to have a phrase structural (“cartographic”) interpretation. The further left a particle (or adverbial) is placed on the hierarchy, the closer to the clausal root is the position of the functional projection(s) it is licensed in. The peripheral “impoverishment” distinguishing 〈i−z±n−w〉-RIs from imperatives strongly suggests that we are dealing with “truncated” structures in the case of these RIs. Such truncation has, of course, earlier been assumed to account for the acquisitional “RI-stage” of child language by Rizzi (1993/94) and Haegeman (1995)9 and it has been appealed to in accounting for the absence of “root transformations” (or “main clauses phenomena”) in various adult language clause types by Haegeman (2010a, 2012). Stated in terms of the latter, certain higher functional projections “cannot be activated” (Haegeman 2012: 127) if they count as interveners in the sense of relativized minimality (Rizzi 2001, 2004) for an operator that has to take wide(st) scope in a particular clause type. A closer look at the (initial segment of the) “universal hierarchy of clausal functional projections” proposed by Cinque (1999: 106) – reproduced in (15) in the enriched form suggested by Coniglio (2006: 68)10 – will enable us to recast (13) within such a model.11 (15) [ frankly Moodspeech act [ fortunately Moodevaluative [ allegedly Moodevidential [ probably Modepistemic [ once T(Past) [ then T(Future) [ perhaps Moodirrealis [ necessarily Modnecessity [ possibly Modpossibility [ willingly Modvolition [ inevitably Modobligation [ cleverly Modability/permission [ usually Asphabitual  [ again Asprepetitive [ often Aspfrequentative . . . Quite strikingly, in this chart irrealis mood separates the “higher” moods and epistemic modality from the “lower” “priority” (cf. Portner 2009) modalities. If we assume that (i) doch and halt are licensed wrt the former and (ii) bloß and ruhig wrt the latter – bloß showing an affinity to necessity and obligation, ruhig to possibility and permission12 – and that (iii) irrealis must, in line with (12b), be the

9 In the current acquisition literature, truncation approaches are highly controversial (cf., e.g., Rau 2005). 10 Cinque (2001, 2004) discusses some of the refinements involved here. 11 It is well-known that cartographic models come with challenges (cf., e.g., Gärtner, Law, and Sabel [2006] on challenges concerning adverb placement). Thus, a satisfactory understanding of “non-activated” projections would be required, massive use of (derivationally introduced) identity functions on the semantic side being more of a last resort rescue manoeuver than a principled approach. For a specific criticism of Coniglio-style assimilation of MPs to adverbs, see Struckmeier (2014). 12 (i) and (ii) are closely related to assumptions made by Grosz (2016).



Root infinitivals and modal particles. An interim report  

 123

outermost operator in 〈i−z±n−w〉-RIs, truncation/deactivation of the higher projections and absence of doch and halt in this kind of root infinitivals follows.13,14 However, even on a superficial understanding of the categories involved in (15), one may doubt that the contrast in (7)/(8) has already been adequately dealt with. In fact, as shown in (16), imperatives pattern with 〈i−z±n−w〉-RIs when MPs are replaced by sentence adverbials.15 müsst (leider)/ (angeblich)/ (vermutlich) hierbleiben. 〈DEC〉 (16) a. Ihr you.pl must unfortunately allegedly presumably here.stay.inf ‘(Unfortunately,)/(Allegedly,)/(Presumably,) You have to stay here.’ b. Bleibt (* leider)/(* angeblich)/(* vermutlich) hier! 〈IMP〉 stay.imp.pl here c. (* Leider)/(* Angeblich)/(* Vermutlich) Hierbleiben!〈i−z−n−w〉 here.stay.inf In fact, the incompatibility of imperatives with evaluative and mirative/evidential adverbials is one of the factors inspiring Jensen (2007: 168) to postulate a truncation account even in the case of imperatives.16 At the same time, absence of an epistemic operator is one of the properties distinguishing imperative from declarative and interrogative sentence mood in the analysis developed by Truckenbrodt (2006a, 2006b). We’ll come back to that point in Section 2.2.

13 Haegeman (2010: §4.2) employs the same mechanism, i.e., operator movement from Moodirrealis, to derive conditional clauses. 14 〈i−z±n−w〉-RIs appear to host “middle field topics” in the sense of Frey (2004) (cf. Gärtner 2013: 213–214). Truncation of (the higher) MoodP-field is compatible with that, as indicated by the possibility for MP doch to occur on either side of the topical DP in imperatives: (i)

Anstatt esi wegzuwerfen, schenk (doch) das Bügeleiseni (doch) instead.of it away.throw.inf give.imp mp the flatiron lieber dem Otto rather the.dat Otto ‘Rather than throwing it away, give the flatiron to Otto, will you!’

The underlying order would have to be doch < TOP, and TOP < doch could be derived by “topic shift.” 15 As shown in (12a), the feature “irrealis” also occurs in imperatives according to Han (2000), but it has to interact with directivity. Truncation accounts of imperatives would have to take this into consideration. Note also, that leider in (16a) does not prevent construal of müssen as a performative modal. Thus, the contrast between (16a) and (16b) has important implications for attempts at reducing imperatives to structures containing covert performative modals (cf. Kaufmann 2012). 16 The exact implementation of this is not entirely clear. The approach is related to the one by Platzack and Rosengren (1998), who pin major peculiarities of imperatives on the absence of FinP.

124 

 Hans-Martin Gärtner

Now, as hinted at by Gärtner (2013, 2014a, 2014b), particular constraints on 〈i−z−w〉-RIs can be derived from the fact that these RIs are “impersonal”: their subjects are closely related to impersonal man ‘one’ (cf. Reis 2003: 187) (modeled by Gärtner [2013, 2014a, 2014b] as arbitrary PRO), and they disallow direct reference to the addressee via second person expressions (Gärtner 2013: 211–212). This property is fully in line with the idea promoted by Truckenbrodt (2006a, 2006b) that structural lack of V0-in-C0 goes together with lack of grammatically encoded “call on the addressee” (cf. Beyssade and Marandin 2006). If, at the same time, MPs like doch and halt signal the need for genuine “common ground checking,” (cf., e.g., Karagjosova 2003; Lindner 1991; Zimmermann 2011), i.e., some kind of call on the addressee to share the speaker’s assessment of conflicting reasons and unassailability, it can be assumed that they require an appropriate (mood related) operator to interact with in performing this function. Structurally, such an approach would follow Jacobs (1991), who worked out the interaction of MPs with illocutionary operators. A variant of this will be briefly discussed in Section 2.2. Independently, the contrast in (10) indicates that RIs, as opposed to imperatives, disallow properly concessive uses. This can be explained along similar lines as the ban on MPs like doch and halt, if one follows Kaufmann (2012: §4.2.3) in assuming that concessive uses of imperatives require the “plugging in” of addressee preferences as modal ordering source. Again, absence vs. presence of the appropriate (mood related) operator can be made responsible for what distinguishes RIs from imperatives here.17

2.2 〈+w 〉-RIs The distribution of MPs in finite vs. non-finite wh-interrogatives found by Gärtner (2013) looks a follows: (17) MP-hierarchy (wh-INT-compatible) (cf. Coniglio 2011: 95) denn > auch > schon > wohl > nur

17 Kaufmann and Kaufmann (2012) derive the incompatibility of ja with imperatives from conflicting presuppositions. Reformulated slightly, ja introduces a “normalcy” assumption that clashes with the preparatory condition on directive speech acts that it is not already clear that the addressee will perform the required action anyway. Doch does not introduce normalcy and is therefore compatible with both declaratives and imperatives. As far as I can see, this analysis of doch is not fine-grained enough to make the further distinction between imperatives and RIs that we are after here.

Root infinitivals and modal particles. An interim report  



 125

(18) MPs acceptable in 〈i−z−n+w〉-RIs (cf. Gärtner 2013: 224) denn > auch > schon > wohl > nur Crucial evidence for the degraded status of wohl in 〈+w〉-RIs comes from contrasts like the one in (19).18 (19) a. Wie man sich wohl verhalten soll in so einer Situation? 〈INT〉 how one self mp behave.inf shall in such a situation ‘How shall one behave in such a situation? (I wonder.)’ b. Wie sich (* wohl) verhalten in so einer Situation?〈i−z−n+w〉 However, unlike the contrast between (7) and (8), the one between (17) and (18) does not lend itself to any neat peripheral truncation.19 So we have to dig deeper right away. A highly relevant explicit approach to wohl has been suggested by Gutzmann (2015):20 For wohl [...] a treatment as a modifier on sentence mood operators seems more plausible, as it is licensed in both declaratives and interrogatives, but not in imperatives. Since the former both involve the epistemic sentence mood operator, while the latter does not, thinking about wohl as modifying the epistemic operator may be the right analysis. (Gutzmann 2015: 256)

(20)

p deont

wohl E (Gutzmann 2015: 141)

18 Speakers who disagree with the judgment in (19b) might take wohl as a pure marker of “reflectivity” (see below) here. For the affinity of wohl with “deliberative” questions, see, for example, Thurmair (1989: 143). 19 Gärtner (2013) has followed Reis (2003) in arguing that German 〈i−z−n+w〉-RIs have CP-status, i.e., they possess a “left periphery” in the sense of Rizzi (1997). As pointed out by Reis (1995: 142), crucial evidence for this comes from long wh-extraction. This fact is consistent with the assumption by, among others, Zimmermann (2004, 2009), Bayer and Obenauer (2011), and Bayer (2012) that MP-licensing in interrogatives involves the CP-layer. Such an assumption can actually further be strengthened by observing that reversals of preferred MP surface orders are possible when the lower MP is pied-piped under wh-movement. This is possible in 〈i−z−n+w〉RIs, too: Wie denn diese Leute nur überzeugen? (i) a. how mp those people mp convince.inf ‘How on earth can one convince those people?’ b. Wie nur diese Leute denn überzeugen? 20 For a slightly more abstract variant of this, see Gutzmann (2017).

126 

 Hans-Martin Gärtner

(20) is modeled on the approach by Truckenbrodt (2006a, 2006b), where wh-movement introduces the (main) epistemic operator (cf. Gärtner 2014a, 2014b). Accordingly, wohl would be incorrectly predicted to be fine in 〈+w〉RIs like (19b). However, 〈+w〉-RIs share “non-descriptive” (priority, buletic, teleological) modality with imperatives (Bhatt 2006; Huntley 1982; Reis 2003), which means they incorporate world-to-word d-o-f at some level. This would allow locating the clash with wohl at a lower (extended) propositional stratum. 21 Within the Cinquean cartographic model in (15), Modepistemic, which is located between the Mood-layer and the non-epistemic modalities, would be a good candidate for the required second epistemic domain. This projection could be assumed to have to be “deactivated” to allow the priority/goal/ volition modalities to determine the “d-o-f orientation” of the (extended) proposition.22

21 As shown in (i), wohl can be retained in directively used V1-interrogatives. (i) Kommst come.2p.ind ‘Come here!’

du wohl her! you mp here

Wohl does not seem to add any epistemic component here. Instead, we may be dealing with a “short-circuited” indirect speech act with an almost grammaticalized particle. At least, this is what has been argued for wrt closely related directive ob-VF-interrogatives: “In ob-imperatives, which are used for strong ‘last call’ commands, the modal particle wohl is obligatory according to Oppenrieder (1989: 191) (cf. also Altmann 1993: 1021 and Kwon 2005: 163)” (Thurmair 2013: 642; translation H.M.G.): deinen Mund hältst! (ii) Ob du wohl gleich if you mp immediately your mouth keep.2p.ind ‘Will you shut up now!’ 22 This comes close to “internal truncation” as proposed by Bianchi (2007) to account for aspectual constraints on Italian wh-infinitives. If one follows Haegeman (2014) in assigning peripheral discourse markers and vocatives to specific functional projections, (many) 〈−w〉-RIs will have to involve internal truncation too. Thus, (prosodically) integrated na, which can be considered a “bonding” particle (Haegeman 2014) (cf. Zifonun, Hoffmann, and Strecker [1997: Ch. C4, §2.16] for detailed discussion and further references), can occur in intial position in (6a) (Na alle mal herhören!). “Address vocatives” like Leute ‘folks’ have to follow na (Na Leute, alle mal herhören! / * Leute, na alle mal herhören!). Using integrated na in its bonding function with bare 〈−w〉-RIs is clearly less felicitous than it is with imperatives as shown in (i). (i) a. Na komm come.imp

her! b. ?? Na here

herkommen! here.come.inf

It is unclear to what extent this effect is due to structural or pragmatic factors. Reluctantly, I have to leave exploration of these matters to another occasion.



Root infinitivals and modal particles. An interim report  

 127

Let us turn to the licensing of denn. Gutzmann (2015: 266) assumes that “the semantics for denn is also able to explain why it can only be used in interrogatives that have the finite verb in C0 [...].” Crucial ingredient of this account is the ability of V0-in-C0 in interrogatives to introduce an additional epistemic operator, hknow, conveying (the Searlean preparatory condition on question acts): “(S assumes) H knows the answer” (Gutzmann 2015: 160; Truckenbrodt 2006b: 398–399). According to Gutzmann (2015: 264–265), denn modifies hknow.23 But again, since 〈+w〉-RIs lack V°-to-C° they should lack hknow too, and, as a consequence, they are incorrectly predicted to be incompatible with denn. The unwelcome result for 〈+w〉-RIs is particularly curious, since other interrogatives without V0-in-C0 seem to observe the predicted ban on denn. This has already been pointed out by Thurmair (1989: 57).24 By way of illustration, let us recapitulate one of the core examples from Truckenbrodt (2004: 333). The following utterance is to be taken in a context where A has asked B about news from someone called Heiner and B has pointed out that s/he hasn’t heard from Heiner for a very long time. Here, V0-in-C0-interrogatives are inadequate as follow-up questions by A, while VF-interrogatives like (21) are fine, but only without denn. (21) A: Ob er (# denn) immer noch kubanische Zigarren mag?〈INT〉 if he mp always still Cuban cigars likes ‘I wonder whether he still likes Cuban cigars.’ Now, 〈+w〉-RIs share pragmatic “reflectivity” with VF-interrogatives in “raising” rather than “asking” a question (cf., e.g., Zimmermann 2013). This makes them selfrather than other-directed (Reis 2003: 190). If we allow ourselves to start from that ingredient and assume affinity of denn and hknow (cf. König 1977: 121) – putting

23 “[W]hat denn does in terms of lexical semantics is [...] adding an additional hearer knowledge condition, namely that the hearer also knows the reason why the speaker raises the question in the first place” (Gutzmann 2015: 265). This function of signaling an (inferable) reason for asking (Ge. Frageanlass) has been pointed out earlier by, e.g., Franck (1980: 223) and Hentschel and Weydt (1983: 268). 24 An anonymous reviewer makes me aware of the fact that there may be speaker variation here. Exceptions like (i) below (Thurmair 1989: 57) arise in “free indirect discourse” reporting on direct questions: (i) Wie ich denn damit zurechtkomme (, wollte er wissen) how I mp there.with cope.1p.ind wanted he know.inf ‘How would I cope with it (, he wanted to know)’ The corpus examples of VF-interrogatives containing denn provided by Kwon (2005: 117) are of this kind too. Eckardt (2015: Ch. 5) provides a recent discussion of particles in such environments.

128 

 Hans-Martin Gärtner

aside the above criticism25 – we are actually able to make sense of cases like (6b). Thus, consider the crude chart in (22) representing the familiar informational parameters of question acts (cf., e.g., Kiefer 1981; Littell, Matthewson, and Peterson 2010; Truckenbrodt 2004): (22)

S

H

a.

+K

+K

rhet

b.

+K

–K

expo

c.

–K

+K

info

d.

–K

–K

refl

Standard information questions, (22c), go together with speaker ignorance (−K) and presumed hearer knowledge (+K). The opposite situation in (22b) gives rise, for example, to “expository questions.” However, where questions are self-addressed, such asymmetries do not (really) make sense, and we are left with (22d) and (22a) for 〈+w〉-RIs. Taking speaker ignorance to be the canonical trigger for question acts, we derive (22d), i.e., reflective questions, as the default for 〈+w〉-RIs.26 And, crucially, if denn requires hearer knowledge, it enforces a switch to something like a self-addressed rhetorical question, i.e., (22a).27 This seems to be exactly the right result. Uttering (6b) conveys that S thinks there is no (compelling) reason to go and examples like Aber woher das Zeug denn nehmen? ‘But where to get this stuff?’ (Reis 1985: 308) signal the conviction that there is nowhere (obvious) the required things can be found.28

25 Zimmermann (2013: 100) accounts for “reflectivity” in terms of the presupposition that S assumes that H lacks certainty wrt answering the question. This is clearly incompatible with hknow (as licensor of denn). 26 “Can’t-find-the-value” effects triggered by insertion of nur (Bayer and Obenauer 2011: §3) are fully compatible with reflectivity in this sense: Wie den Anforderungen nur gerecht werden? ‘How (on earth) can one meet these demands?’ (Gärtner 2013: 214). 27 The pragmatics of 〈+w〉-RIs raises interesting questions about the analysis of inner and public monolog. In particular, the extent to which this can turn into quasi-dialog − with speaker and addressee as one and the same person (cf., e.g., Maynard 1995) − may constitute a parameter responsible for variable acceptability judgments. 28 In a corpus study of denn, Deppermann (2009) provides a number of finite “‘rhetorisch[e]’ denn-Fragen” (Deppermann 2009: 35) and discusses the role of examples like wo soll denn das noch hinführ’n ‘Where is this going to lead?’ (Deppermann 2009: 28) in speech acts like protests and complaints (cf. König 1977: 122). Importantly, the “exclamatory” dimension of such moves (Deppermann 2009: 34) has to be brought into line with the fact that denn is ruled out in wh-exclamatives (cf. Thurmair 1989: 58).



Root infinitivals and modal particles. An interim report  

 129

If the above account is on the right track, the (apparent) non-interaction of denn with finite VF-interrogatives indicated in (21) remains to be explained, but a serious discussion of the resulting complications is beyond the scope of this paper.29

2.3 Declarative and polar interrogative RIs? Let us shift gears slightly and turn to more fundamental questions concerning the inventory of sentence types and moods. While we might be reluctant to speak of non-finite imperatives in cases like (2a)/(2b),30 considering 〈+w〉-RIs non-finite (wh-)interrogatives would seem unproblematic. The latter intuition may be influenced by facts from English, where “wh-infinitives” pattern with finite (bona fide) interrogatives both in root and subordinate environments (cf., e.g., Bhatt 2006). Further light can be shed on this by reflecting upon the status of putative instances of declarative, (23a), and polar interrogative RIs, (23b). (23) a. %Und ich nur noch wegrennen. and I only still away.run.inf ‘And I’m all running away.’ b. Wegrennen(/)? ‘(Shall I/we) Run away?’ (23a) is a counterpart of Dutch (Haegeman 1995: 220; Reuland 1983: 127, fn.18) and French (Nikolaeva 2014) “historical” or “narrative” infinitives, used, among other things, to recount (in vivid terms) the climax of an event.31 (23b) looks like a standard 〈i–z–n–w〉-RI modified prosodically to be used for a question act. Let us deal with the latter case first.

29 For example, it has to be taken into account that in the case of wh-VF-interrogatives, wohl is (almost) obligatory (cf. Oppenrieder 1989: 167). We would thus have to address the more involved issue of multiple MPs (... denn ... wohl ...). And even if we can convince ourselves that examples like Na, wer wird uns denn dabei wohl helfen? ‘Well, (guess) who will help us with this?’ do possess a rhetorical construal, we are obliged to explore the relation to bona fide rhetorical specimen involving the particle schon (cf., e.g., Meibauer 1986: §5.2; 1991). 30 As reported by Mastop (2005: 44), Bolinger (1977: 173) defends an analysis of the English imperative as based on “the bare infinitive with a meaning of hypothesis.” 31 Marking by % is meant to indicate that the construction may be rare and bound to particular regional varieties of German.

130 

 Hans-Martin Gärtner

2.3.1 Polar interrogative RIs? Commenting on examples like (23b), Reis (1995: 115) states that “there don’t seem to be any infinitival structures used as polar questions that are independent from the point of view of sentence grammar” [translation H.M.G.]. And she goes on to explain that “[c]ases like those in [(24)], which obligatorily bear rising intonation, only superficially contradict this. We are either dealing with echo-uses [(24a)] – which are of no interest here – or recognizably discourse dependent elliptical cases [(24b)]” [translation H.M.G.]. (24) a. A: Ihr müßt jetzt weggehen. B: Weggehen(/)? ‘You must go now.’ ‘Go?’ b. A: Was soll man tun? B: Weggehen(/)? Oder dableiben und ...(/)? ‘What should one do?’ ‘Go? Or stay and ...?’ “Instances of initiating occurrences are practically absent, in contrast to what is the case with infinitival wh-clauses and finite interrogatives” [translation H.M.G.].32 If indeed obligatorily reactive types can be dealt with in terms of discourse ellipsis (cf., e.g., Klein 1993 for some discussion), what remains to be focused on are “[s]ome cases of this form [...] [that] cannot be explained this way, since they are doubtlessly initiative, cf. [(25)] (example by I. Lasser, p.c.)” (Reis 2003: 330) (cf. Lasser 2002: 777): (25) [Nurse, entering and inspecting a patient room, asks offeringly]: Na, Betten aufschütteln? prt beds up.shake.inf ‘How about (my) making your beds?/Should I make your beds?’ Such examples are particularly interesting in the current context because “[c]ases like Gehen? [‘(Shall we) Go?’] [...] can be easily shown to be noninterrogatives in terms of sentence grammar in that they do not admit the diagnostic particle denn [...]: *Denn gehen?” (Reis 2003: 161, fn. 7) (cf. Reis 1995: 115). (26b) replicates this finding, showing in addition that the effect cannot be explained by banning modal particles in initial position: (26) a. Soll ich DIEses Fenster (denn) AUFmachen? shall I this window mp open.make.inf

32 Another reactive type can be found among the 15 kinds of “non-sentential utterances” (NSUs) inventorized and analyzed by Ginzburg (2012). (i) instantiates what Ginzburg calls “Fillers” (Ginzburg 2012: 149): (i) A: Ich wollte die ganze Zeit ... B: Wegrennen? ‘All the time I wanted to ...’ ‘Run away?’ Interestingly, initiating types are entirely absent from Ginzburg’s inventory.



Root infinitivals and modal particles. An interim report  

 131

b. DIEses Fenster (* denn) AUFmachen? ‘Shall I open this window?’ As an immediate consequence of Reis’s assessment, we are forced to accept that cases of sluicing (in dialog), (27),33 and wh-phrase initial “verb-less directives” (VLDs) (Fries 1983: §5.2; Jacobs 2008), (28), are interrogatives qua “sentence grammar”: (27) A: Hans kommt nicht. B: Warum denn? ‘Hans isn’t coming’ ‘Why (actually)?’ (28) Wohin denn mit dem Tisch? where mp with the table ‘Where should we put the table?’ According to Müller (2011: §2.3), standard VLDs like In den Keller mit dem Tisch ‘Put the table into the basement!’ actually are (abstract) verb initial “V0-in-C0” clauses. From this it is concluded that they “can be interpreted as imperatives or polar questions” (Müller 2011: 225; translation H.M.G.). Again, however, a ban on denn indicates that VLDs wouldn’t be polar interrogatives qua “sentence grammar” either: (29) (*Denn) In den Keller mit dem Tisch? ‘Shall we put the table into the basement?’ VLDs and 〈i–z〉-RIs thus behave similarly selectively as bases for interrogatives: wh-interrogatives are licensed but polar ones aren’t.34, 35

33 Bayer (2010) and Bayer and Obenauer (2011) provide some discussion of the interaction between wh-movement and particles in the context of sluicing. 34 VLDs and 〈i−z−n−w〉-RIs also coincide wrt the set of modal particles they share. At least, this is what (i) suggests: (i) *Doch/*Halt/Bloß/Ruhig in den Keller mit dem Tisch! Again, particle combinations may blur the picture. Thus, dann halt is acceptable in VLDs: (ii) Dann halt in den Keller mit dem Tisch! ‘Put the table in the basement then (, if you must)!’ The clear contrast between Dann halt nicht ‘Then don’t’ and *Halt nicht − used, for example, as comment on somebody’s rejecting an offer − shows that such combinations with dann have a peculiar status that sets them off from ordinary (single) modal particles. 35 Combined with the perspective on V0-in-C0 defended by Truckenbrodt (2006a, 2006b), the analysis by Müller (2011) makes the correct prediction that wh-VLDs like (28) differ from 〈+w〉-RIs in not being restricted to reflective question acts (cf. Section 2.2).

132 

 Hans-Martin Gärtner

If – to return to the main point of this section – we follow Reis in denying the existence of polar interrogative RIs,36 “[a]s a consequence, the special case illustrated in [(25)] must be accounted for in a different way [...]. Whether or not this is a welcome consequence will have to be left open here” (Reis 2002: 332). To advance our understanding of the matter slightly further, however, I speculate on how such cases fit into a specific pattern of use. It has been observed earlier (Deppermann 2007: §2.8.1.3) that what looks like 〈i−z−n−w〉-RIs figure in “empractic” contexts, i.e., contexts that at core involve an action to be performed. Uhmann (2010) devotes an entire study to such cases in the setting of a surgical operation. The most reduced specimen like Jetzt spülen! ‘Now flush!’ can be considered pure “action triggers,” i.e., within routine settings they function like the pressing of a button on a control board. Bare nominals (Skalpell (bitte)! ‘Scalpel (please)!’) occur with the same function (cf. Uhmann 2010: 52). Now, clearly, the person meant to perform an action can foresee the use of such an action trigger and help procedures (for all kinds of reasons) by actually preempting it. Thus, Jetzt spülen? could be used as an “action trigger preemptor.” This is what I think is going on in cases like (25) as well. And, crucially, if action trigger preemptors (ATPs) live on action triggers (ATs) in the sense of “inverting” them via question intonation, they cannot be structurally interrogative – unless there are interrogative ATs – and the possibility of inserting denn is simply absent. Since ATPs are used in question acts, other particles that ATs might contain will be filtered out pragmatically if they don’t fit: Ruhig anfangen! ‘Begin!’ vs. (# Ruhig) Anfangen? ‘Shall I/we begin?’.37

36 Reis (2002: Appendix A) attempts to derive this from a particular clash between the semantics and pragmatics of answering conditions. The discussion there would seem to have a bearing on the degraded nature of reflective polar root wh-infinitives in English, shown by the contrast between Where to begin? and #Whether to accept the offer?. I’ll have to leave further discussion for another occasion. 37 At least at this stage, it is thus irrelevant for my account whether infinitival ATs are considered a special subtype of ellipsis or full-fledged RIs. The discussion by Uhmann (2010: 51–53) is inconclusive in this respect. Further questions arise in the light of cases like (i) (addressed by a conductor to a BahnCard holder). (i) Punkte sammeln? ‘Would you like to collect credit points?’ While most infinitival ATPs could be reconstructed in terms of “initial material deletion” (cf. De Clercq [2009], Napoli [1982], and Nygård [2014] for Dutch, English, and Norwegian, respectively) by adding soll ich ‘shall I’, here something like wollen Sie ‘do you want (to)’ would be missing. Although this resembles the directive vs. desiderative/optative alternation in 〈i-z-n-w〉-RIs (Reis 1995, 2003), the source AT for (i) would clearly have to differ in flavor from cases like Ach, noch einmal Venedig sehen! ‘Oh, to see Venedig just once again!’. As the contrast in (26) shows, initial material deletion would not be a viable solution for infinitival ATPs, at least not without additional constraints.

Root infinitivals and modal particles. An interim report  



 133

2.3.2 Declarative RIs? In her brief overview over “root infinitives in adult language,” Lasser (2002: 776) states that “RIs are used by adults as declaratives.” This is in direct conflict with the intuitions expressed by Rizzi (1993/94: 375) (for Italian): “What seems to be excluded is a root-infinitival declarative. That is, if I see somebody playing football, I could not describe the scene by uttering [giocare al pallone].” And indeed, most of the examples Lasser presents are not descriptive in this sense.38 But what about “narrative infinitives” (NIs) like (23a)? One method of relating them to bona fide declaratives would be considering them elliptical (am+Infinitive) progressives (cf., e.g., Bhatt and Schmidt 1993), as shown in (30). (30) Und ich bin nur noch am Wegrennen. and I am only still at away.run.inf ‘And I’m all running away.’ What adherents of such an approach would have to explain, though, is why the habitual interpretation allowed for the non-elliptical variant of (30) is ruled out after ellipsis.39 At the same time, the analysis in (30) has the advantage of accounting for the presence of a (first person) nominative (subject) and for divergence from the finding that “[t]he interpretation of independent bare infinitives always involves a modal component” (Reis 1995: 116; translation H.M.G.).40 Now, given the somewhat unclear status of German NIs, I’ll shift over to so-called “nonfinite presentatives” (NFPs) (König and Siemund 2007: 320; Zaefferer 1989: 223), also known as “Mad Magazine Sentences” (Akmajian 1984): (31) Der und Piano spielen?! Niemals. ‘Him play the piano?! Never.’

38 Example (i) is taken from the literature and analyzed as belonging to “comments on on-going actions” (Lasser 2002: 776): (i)

Die Nudeln noch ein bisschen zudecken the pasta still a little cover.inf ‘Now I am covering the pasta for a short while.’ (from Miller 1976, cited in Lasser 2002: 776)

Of course, this would be a good candidate for “left material deletion” of, for example, Ich werde ‘I will’ or Ich sollte ‘I should’. Alternatively this could be a self-addressed AT or the externalization of internal self-talk “mentally activating” an action concept (cf. Fiehler 1994). This rather underexplored dimension of use is likely to be linked to the “suggestive” employment of bare infinitives in advertisements already documented by Fries (1983: §5.6.1). 39 A closely related point is made by Reis (2003: 177–178) concerning the covert modality of 〈+w〉-RIs. 40 Reis (1995) discusses nominatives in RIs in detail (cf. Gärtner 2013: 209–211). The absence of modality in French and Latin narrative infinitives is observed by Nikolaeva (2014: 145, 157).

134 

 Hans-Martin Gärtner

NFPs are canonically “used to present a proposition with the purpose of rejecting it as absurd” (König and Siemund 2007: 319). Importantly, Thurmair (2013: 644) comments on NFPs stating that “the structures noun phrase (with und) and infinitive do not allow modal particles” [translation H.M.G.]. This radical absence of modal particles in NFPs fits in well with the line taken by Reis (2003: 182), who – following Lambrecht (1990) – observes that “the use of the infinitive in [NFPs] is [...] no necessary feature of the construction to begin with.” (32) illustrates this. (32) Der und Pianist?! Niemals. ‘Him a piano player?! Never.’ Crucially, the same holds for German NIs, as shown in (33). (33) %Und ich nur noch sprachlos. and I only still speechless ‘And I’m all speechless.’ Assuming thus that the infinitive makes no essential contribution to NFPs and (German) NIs, we are justified in considering these constructions to lie outside the realm of RIs.41 What’s more, we have reason to believe that there are no declarative RIs.

3 Modal particles in root infinitivals of type 〈PII〉 and 〈S〉 For the sake of completeness I will briefly comment on occurrences of MPs in root infinitivals of type 〈pii〉 and 〈s〉.42 According to Thurmair (2013: 644) “the

41 Grohmann and Etxepare (2003) study modality and truncation in English and Spanish NFPs. Reviewing their findings in the light of our discussion in Sections 2.1 and 2.2 remains to be done. 42 While zu-infinitival clauses occur in a wide range of embeddings, their occurrence in root contexts seems to be limited to exclamations. This makes assuming a separate category of 〈i+z〉RIs doubtful (cf. Gärtner 2013: 223). As dependent clauses, zu-infinitivals have been shown to host modal particles (cf., e.g., Grosz 2014; Rapp and Wöllstein 2013). For root environments, particles found in (non-wh) exclamatives like aber and vielleicht (Thurmair 1989: 49) have to be tested, and it seems to be advisable to come up with adequate finite counterpart constructions first. An exemplary frame is given in (i), but I refrain from carrying out systematic investigation here, except for noting that use of aber seems to improve with supporting auch (> aber auch): (i) a. Dass man einen SO schönen Garten __ DERart verunstaltet! that one a so beautiful garden thus spoils ‘It’s a scandal that they should spoil such a beautiful garden in such a manner!’ b. (Banausen,) Einen SO schönen Garten __ DERart zu verunstalten!



Root infinitivals and modal particles. An interim report  

 135

acceptability of modal particles is restricted: requests with an infinitive [...] allow quite a number of modal particles that are typical for requests [...] requests based on participial forms don’t” [translation H.M.G.]. And, Heinold (2013: 328) remarks that “the few extended items that could be found contained the modal particle mal (Alle mal aufgebrezelt!; [‘Everyone dress up cool!’]).” As far as my own intuitions go, 〈pii±n−w〉-RIs and 〈i−z±n−w〉-RIs allow exactly the same MPs, so (8) applies here as well. The cases in (34), for example, are entirely uncontroversial. (34) (Also Leute,) Ruhig/ JA mitangepackt! well folks mp mp with.on.pack.pcp ‘(Well folks,) Do lend us a hand!’ The relevant aspects of the account for 〈i−z±n−w〉-RIs thus simply carry over. 〈s−n−w〉-RIs, also known as “non-inflectionals” (Ge. Inflektive) (Bücking and Rau 2013; Teuber 1998), raise very interesting questions that can only be discussed very sketchily here. Let us first have a look at one of the more elaborate items from a corpus of online-chat examples (Bücking and Rau 2013: 81). By orthographic convention, 〈s〉-RIs are enclosed by dot-markers (•). B: [...] •Becher hinhalt• (35) A: guten Morgen!!!! •Kaffeeausschenk• good morning coffee.out.pour.st      cup hold.out.st Together with even clearer cases like •dich in den Arm nehm• ‘(I’m) giving you a hug’ these are instances of “doing things with words” in the sense of making things happen in cyberreality. Bücking and Rau (2013: 85; cf. Portner 2007) therefore analyze 〈s〉-RIs as “separate performatives.” Some examples contain mal, but, as indicated by (36) (Bücking and Rau 2013: 72), this can in principle be interpreted as contentful temporal or frequency adverb. Clear cases of mitigating mal require directive uses. (36) •hier mal die fenster aufmach• here the windows open.st ‘(I’m) just opening the windows.’ Interestingly, Brandt et al. (1989: 7) observe that “the unstressed modal particles doch, ja, and schon do not allow performative interpretation of the utterance” [translation H.M.G.]. One of the examples they illustrate this with is shown in (37). (37) a. Ich eröffne ja die Sitzung.〈DEC〉 I open mp the meeting ‘I am opening the meeting, as you can see.’ b. Die Sitzung ist ja eröffnet.〈DEC〉 the meeting is mp opened ‘The meeting has already been opened.’

136 

 Hans-Martin Gärtner

Unfortunately, I have been unable to obtain the relevant data and my impression is that the construction is in some flux, so that different “dialects” supporting different intuitions may exist. However, consider the following continuation of (35), where A has seemingly ignored B’s utterance and action (z marks that the acceptability of the example is unclear and needs to be confirmed).43 (38) A: Na? Will keiner Kaffee? B: z [...] •Becher doch hinhalt• so wants no.one coffee cup mp hold.out.st ‘So, no one wants any coffee?’ / ‘I’m holding up my cup, didn’t you notice?’ If B’s utterance were an acceptable inflective it would have to be descriptive here, i.e., rather than performing the action, B would assert that the action is being performed and that A should have noticed (from the previous utterance). This intuition is fully in line with the fact, that actually there are what appear to be purely descriptive uses of 〈s〉-RIs in the literature, (39) being a case in point (Bücking and Rau 2013: 73).44 (39) •fliegende sektkorken beobacht• flying champagne.corks watch.st Obviously, a full scale study of these matters remains to be done. Nevertheless, it can tentatively be concluded from the preceding discussion that a putative ban on MPs in 〈s−n−w〉-RIs may not be a purely interpretive matter but is likely to require assumptions about their structural “impoverishment” as well.

4 Conclusion This interim report has revisited and expanded on findings by Gärtner (2013) concerning the interaction of German adult root infinitivals and modal particles. The

43 Andreas Pankau (p.c.) rejects the inflectional with inserted doch. 44 It would be interesting to check “non-inflectionals” with modal particles like halt, which Brandt et al. (1989: 6) consider compatible with explicit performativity, together with verbs that can be used in explicit performatives in standard communication. The concession construal of (i) would be a relevant test case. (i) z OK! •euch das Bild halt you.dat the picture mp ‘OK! I (hereby) promise you the picture.’

(hiermit) hereby

versprech• promise.st



Root infinitivals and modal particles. An interim report  

 137

central empirical results are summarized in (8) and (18), repeated here for convenience in (40) and (41). (40) MPs acceptable in 〈i−z±n−w〉-RIs (cf. Gärtner 2013: 224) doch > halt > DOCH > nur (nicht) > bloß (nicht) > ruhig > {mal,JA} (41) MPs acceptable in 〈i−z−n+w〉-RIs (cf. Gärtner 2013: 224) denn > auch > schon > wohl > nur Bare (〈i−z〉) “deontic” RIs and wh-RIs restrict occurrence of MPs to subsets of those allowed in imperatives and finite wh-interrogatives, respectively. As discussed in Section 2.1, the pattern in (40) invites an analysis of 〈i−z±n−w〉-RIs by means of Haegeman-style truncation promoting an irrealis operator within a Cinquean hierarchy of functional projections. Additional divergences from imperatives can be accounted for via lack of grammaticalized “call on the addressee” leading to constraints on the ordering sources able to interact with the covert modal of 〈i−z±n−w〉-RIs. The prima facie identical distribution of MPs in participial 〈pii±n−w〉-RIs (Section 3) invites the conjecture that this account carries over here, mutatis mutandis. The pattern in (41), discussed in Section 2.2, points to a clash between epistemic modifiation and the “non descriptive” (priority/goal/volition) modality located at the (extended) propositional level of 〈i−z−n+w〉-RIs. Occurrences of denn, which is standardly taken to signal presumptions about “hearer knowledge,” is argued to lead to rhetorical question acts with 〈+w〉-RIs due to their quasi monologic uses. A ban on the occurrence of denn, originally observed by Reis (1995), is at the center of an argument developed in Section 2.3.1 against the existence of polar interrogative RIs as formal sentence types of German. Section 2.3.2 then goes on to make a related case against declarative RIs. Finally, Section 3 addresses stem-based “non-inflectionals” (Bücking and Rau 2013), concluding that banning MPs from occurring in this type would require structural in addition to interpretive assumptions as well. Obviously, a lot of work remains to be done both empirically and conceptually. I hope to have shown, however, that studying MPs in root infinitivals has very good potential of leading to (the identification of constraints on) refined theories of sentence types, moods, and speech acts. Abbreviations. AT = action trigger; ATP = action trigger preemptor; dat = dative (gloss); DEC = declarative; d-o-f = direction of fit; hknow = hearer knowledge; imp = imperative; ind = indicative (gloss); inf = infinitive (gloss); INT = interrogative; 〈i−z−n−w〉 = infinitive without zu, without nominative DP, without fronted wh-phrase; 〈i−z+n−w〉 = infinitive without zu, with nominative DP, without

138 

 Hans-Martin Gärtner

fronted wh-phrase; 〈i−z−n+w〉 = infinitive without zu, without nominative DP, with fronted wh-phrase; 〈i+z−n−w〉 = infinitive with zu, without nominative DP, without fronted wh-phrase; MP = modal particle; NFP = nonfinite presentative; NI = narrative infinitive; NSU = non-sentential utterance; pcp = past participle (gloss); pl = plural (gloss); prt = particle (gloss); 〈pii−n−w〉 = past participle, without nominative DP, without fronted wh-phrase; 〈pii+n−w〉 = past participle, with nominative DP, without fronted wh-phrase; RI = root infinitival; 〈s−n−w〉 = verbal stem, without nominative DP, without fronted wh-phrase; st = verbal stem (gloss); VF = verb-final; VLD = verb-less directive; 1p = first person (gloss); 2p = second person (gloss).

References Abraham, Werner. 1995. Wieso stehen nicht alle Modalpartikel in allen Satzformen? Deutsche Sprache 23(2). 124–146. Abraham, Werner. 2017. Discourse marker = discourse particle = thetical = modal particle? A futile comparison. In Josef Bayer & Volker Struckmeier (eds.), Discourse particles: Formal approaches to their syntax and semantics, 241–280. Berlin: de Gruyter. Akmajian, Adrian. 1984. Sentence types and the form-function fit. Natural Language and Linguistic Theory 2(1). 1–23. Altmann, Hans. 1987. Zur Problematik der Konstitution von Satzmodi als Formtypen. In Jörg Meibauer (ed.), Satzmodus zwischen Grammatik und Pragmatik, 22–56. Tübingen: Niemeyer. Altmann, Hans. 1993. Satzmodus. In Joachim Jacobs, Arnim von Stechow, Wolfgang Sternefeld & Theo Vennemann (eds.), Syntax. Ein internationales Handbuch zeitgenössischer Forschung. An international handbook of contemporary reseach. 1006–1029. Berlin: de Gruyter. van der Auwera, Johan & Ewa Schalley. 2004. From optative and subjunctive to irrealis. In Frank Brisard, Michael Meeuwis & Bart Vandenabeele (eds.). Seduction, community, speech: A festschrift for Herman Parret, 87–96. Amsterdam: Benjamins. Bayer, Josef. 2010. Discourse particles in questions. GLOW in Asia 7. 257–273. Bayer, Josef. 2012. From modal particle to interrogative marker: A study of German denn. In Laura Brugé, Anna Cardinaletti, Giuliana Giusti, Nicola Munaro & Cecilia Poletto (eds.), Functional heads, 13–28. Oxford: Oxford University Press. Bayer, Josef & Hans-Georg Obenauer. 2011. Discourse particles, clause structure, and question types. The Linguistic Review 28(4). 449–491. Beyssade, Claire & Jean-Marie Marandin. 2006. The speech act assignment problem revisited: Disentangling speaker’s commitment from speaker’s call on addressee. In Olivier Bonami & Patricia Cabredo Hofherr (eds.), Empirical Issues in Syntax and Semantics 6, 37–68. Paris: CNRS. http://www.cssp.cnrs.fr/eiss6/beyssade-marandin-eiss6.pdf (accessed 29 June 2016).



Root infinitivals and modal particles. An interim report  

 139

Bhatt, Christa & Claudia Maria Schmidt. 1993. Die am + Infinitiv-Konstruktion im Kölnischen und im umgangssprachlichen Standarddeutschen als Aspekt-Phrasen. In Werner Abraham & Josef Bayer (eds.), Dialektsyntax, 71–98. Opladen: Westdeutscher Verlag. Bhatt, Rajesh. 2006. Covert modality in non-finite contexts. Berlin: Mouton de Gruyter. Bianchi, Valentina. 2007. Wh-Infinitives and the licensing of ‘anaphoric tense’. In M. Cecilia Picchi & Alan Pona (eds.), Proceedings of the XXXII Incontro di Grammatica Generativa, 35–47. Alessandria: Edizioni dell’Orso. Bolinger, Dwight. 1977. Meaning and form. London: Longman. Brandt, Margareta, Gabriel Falkenberg, Norbert Fries, Frank Liedtke, Jörg Meibauer, Günther Öhlschläger, Helmut Rehbock & Inger Rosengren. 1989. Die performativen Äußerungen – eine empirische Studie. Sprache und Pragmatik 12. 1–21. Bücking, Sebastian & Jennifer Rau. 2013. German non-inflectional constructions as separate performatives. In Daniel Gutzmann & Hans-Martin Gärtner (eds.), Beyond expressives: Explorations in use-conditional meaning, 59–94. Leiden: Brill. Cinque, Guglielmo. 1999. Adverbs and functional heads. Oxford: Oxford University Press. Cinque, Guglielmo. 2001. ‘Restructuring’ and the order of aspectual and root modal heads. In Guglielmo Cinque & Giampaolo Salvi (eds.), Current studies in Italian syntax, 137–155. Amsterdam: Elsevier. Cinque, Guglielmo. 2004. ‘Restructuring’ and functional structure. In Adriana Belletti (ed.), Structures and beyond: The cartography of syntactic structures, vol. 3, 132–191. Oxford: Oxford University Press. Coniglio, Marco. 2006. German modal particles in the functional structure of IP. University of Venice Working Papers in Linguistics 16. 57–95. Coniglio, Marco. 2011. Die Syntax der deutschen Modalpartikeln. Berlin: Akademie Verlag. De Clercq, Karen. 2009. Fonologische deletie in ja-nee-vraagfragmenten [Phonological deletion in yes-no-question fragments]. Handelingen van de Koninklijke Zuid-Nederlandse Maatschappij voor Taal- en Letterkunde en Geschiedenis 62. 5–14. Deppermann, Arnulf. 2007. Der deontische Infinitiv. In Arnulf Deppermann, Grammatik und Semantik aus gesprächsanalytischer Sicht, 113–209. Berlin: de Gruyter. Deppermann, Arnulf. 2009. Verstehensdefizit als Antwortverpflichtung: Interaktionale Eigenschaften der Modalpartikel denn in Fragen. In Susanne Günthner & Jörg Bücker (eds.), Grammatik im Gespräch. Konstruktionen der Selbst- und Fremdpositionierung, 23–56. Berlin: de Gruyter. Eckardt, Regine. 2015. The semantics of free indirect discourse. Leiden: Brill. Fiehler, Reinhard. 1994. Formen des Sprechens mit sich selbst. In Gisela Brünner & Gabriele Graefen (eds.), Texte und Diskurse. Methoden und Forschungsergebnisse der Funktionalen Pragmatik, 179–198. Opladen: Westdeutscher Verlag. Fortuin, Egbert. 2003. De directieve infinitief en de imperatief in het Nederlands [The directive infinitive and the imperative in Dutch]. Nederlandse Taalkunde 8. 14–43. Franck, Dorothea. 1980. Grammatik und Konversation. Königstein: Scriptor. Frey, Werner. 2004. A medial topic position for German. Linguistische Berichte 198. 153–190. Fries, Norbert. 1983. Syntaktische und semantische Studien zum frei verwendeten Infinitiv. Tübingen: Narr. Gärtner, Hans-Martin. 2013. Infinite Hauptsatzstrukturen. In Jörg Meibauer, Markus Steinbach & Hans Altmann (eds.), Satztypen des Deutschen, 202–231. Berlin: de Gruyter.

140 

 Hans-Martin Gärtner

Gärtner, Hans-Martin. 2014a. On covert modality in German root infinitives. In Robert E. Santana-LaBarge (ed.), Proceedings of the 31st West Coast Conference on Formal Linguistics, 199–206. Somerville, MA: Cascadilla Proceedings Project. Gärtner, Hans-Martin. 2014b. Überlegungen zur versteckten Modalität infiniter Hauptsatzstrukturen. Linguistische Berichte 237. 81–92. Gärtner, Hans-Martin, Paul Law & Joachim Sabel. 2006. Clause structure and adjuncts in Austronesian languages: A critical introductory survey. In Hans-Martin Gärtner, Paul Law & Joachim Sabel (eds.), Clause structure and adjuncts in Austronesian languages, 1–42. Berlin: Mouton de Gruyter. Ginzburg, Jonathan. 2012. The interactive stance. Meaning for conversation. Oxford: Oxford University Press. Grohmann, Kleanthes & Ricardo Etxepare. 2003. Root infinitives: A comparative view. Probus 15(2). 201–236. Grosz, Patrick. 2014. Modal particles in rationale clauses and related constructions. In Werner Abraham & Elisabeth Leiss (eds.), Modes of modality, 263–290. Amsterdam: Benjamins. Grosz, Patrick. 2016. Discourse particles. Ms., University of Tübingen. http://semanticsarchive. net/Archive/GQ4ZDU0N/grosz-SemCom-discourse-particles.pdf (accessed 29 June 2016). Gutzmann, Daniel. 2015. Use-conditional meaning. Studies in multidimensional semantics. Oxford: Oxford University Press. Gutzmann, Daniel. 2017. Modal particles ≠ modal particles (= modal particles). In Josef Bayer & Volker Struckmeier (eds.), Discourse particles: Formal approaches to their syntax and semantics, 144–172. Berlin: de Gruyter. Haegeman, Liliane. 1995. Root infinitives, tense, and truncated structures in Dutch. Language Acquisition 4(3). 205–255. Haegeman, Liliane. 2010a. The internal syntax of adverbial clauses. Lingua 120(3). 628–648. Haegeman, Liliane. 2010b. The movement derivation of conditional clauses. Linguistic Inquiry 41(4). 595–621. Haegeman, Liliane. 2012. The syntax of MCP. Deriving the truncation account. In Lobke Aelbrecht, Liliane Haegeman & Rachel Nye (eds.), Main clause phenomena: New horizons, 113–134. Amsterdam: Benjamins. Haegeman, Liliane. 2014. West Flemish verb-based discourse markers and the articulation of the speech act layer. Studia Linguistica 68(1). 116–139. Han, Chung-hye. 2000. The structure and interpretation of imperatives. Mood and force in universal grammar. New York: Garland. Heinold, Simone. 2013. Eigenschaften von direktiven Partizipien im Deutschen. Deutsche Sprache 4/2013. 313–34. http://www.DSdigital.de/DS.04.2013.335 (accessed 29 June 2016.) Hentschel, Elke & Harald Weydt. 1983. Der pragmatische Mechanismus: denn und eigentlich. In Harald Weydt (ed.), Partikeln und Interaktion, 263–273. Tübingen: Niemeyer. Huntley, Martin. 1982. Imperatives and infinitival questions. In Kevin Tuite, Robinson Schneider & Robert Chametzky (eds.), Papers from the 18th Regional Meeting, Chicago Linguistic Society. Parasession on Non-Declaratives, 93–106. Chicago, IL: Chicago Linguistic Society. Jacobs, Joachim. 1991. On the semantics of modal particles. In Werner Abraham (ed.), Discourse particles, 141–162. Amsterdam: Benjamins. Jacobs, Joachim. 2008. Wozu Konstruktionen? Linguistische Berichte 213. 3–44. Jensen, Britta. 2007. In favour of a truncated imperative clause structure: Evidence from adverbs. Working Papers in Scandinavian Syntax 80. 163–185.



Root infinitivals and modal particles. An interim report  

 141

Karagjosova, Elena. 2003. Modal particles and the common ground. Meaning and functions of German ja, doch, eben/halt and auch. In Peter Kühnlein, Hannes Rieser & Henk Zeevat (eds.), Perspectives on dialogue in the new millenium, 335–349. Amsterdam: Benjamins. Kaufmann, Magdalena. 2012. Interpreting imperatives. Heidelberg: Springer. Kaufmann, Magdalena & Stefan Kaufmann. 2012. Epistemic particles and performativity. Semantics and Linguistic Theory 22. 208–225. DOI: http://dx.doi.org/10.3765/salt. v22i0.2635 (accessed 29 June 2016). Kiefer, Ferenc. 1981. Questions and attitudes. In Wolfgang Klein & Willem Levelt (eds.), Crossing the boundaries in linguistics, 159–176. Dordrecht: Reidel. Klein, Wolfgang. 1993. Ellipse. In Joachim Jacobs, Arnim von Stechow, Wolfgang Sternefeld & Theo Vennemann, (eds.), Syntax. Ein internationales Handbuch zeitgenössischer Forschung. An international handbook of contemporary research. 763–799. Berlin: de Gruyter. König, Ekkehard. 1977. Modalpartikeln in Fragesätzen. In Harald Weydt (ed.), Aspekte der Modalpartikeln, 115–130. Tübingen: Niemeyer. König, Ekkehard & Peter Siemund. 2007. Speech act distinctions in grammar. In Timothy Shopen (ed.), Language typology and syntactic description, vol. 1, 276–324. Cambridge: Cambridge University Press. König, Ekkehard, Detlef Stark & Susanne Requardt. 1990. Adverbien und Partikeln. Ein deutschenglisches Wörterbuch. Heidelberg: Julius Groos Verlag. Kwon, Min-Jae. 2005. Modalpartikeln und Satzmodus. München: Ludwig-MaximiliansUniversität dissertation. Lambrecht, Knud. 1990. “What, me worry?” – ‘Mad Magazine sentences’ revisited. In Kira Hall, Jean-Pierre Koenig, Michael Meacham, Sondra Reinman & Laurel A. Sutton (eds.), Proceedings of the 16th Annual Meeting of the Berkeley Linguistics Society, 215–228. Berkeley, CA: Berkeley Linguistics Society. Lasser, Ingeborg. 2002. The roots of root infinitives: Remarks on infinitival main clauses in adult and child language. Linguistics 40(4). 767–796. Lindner, Karin. 1991. ‘Wir sind ja doch alte Bekannte’. The use of German ja and doch as modal particles. In Werner Abraham (ed.), Discourse Particles, 303–328. Amsterdam: Benjamins. Littell, Patrick, Lisa Matthewson & Tyler Peterson. 2010. On the semantics of conjectural questions. University of British Columbia Working Papers in Linguistics 28. 89–104. Mastop, Rochus Jacobus. 2005. What can you do? Imperative mood in semantic theory. Amsterdam: University of Amsterdam dissertation. Maynard, Senko. 1995. Interrogatives that seek no answers: Exploring the expressiveness of rhetorical interrogatives in Japanese. Linguistics 33(3). 501–530. Meibauer, Jörg. 1986. Rhetorische Fragen. Tübingen: Niemeyer. Meibauer, Jörg. 1991. Existenzimplikaturen bei rhetorischen Fragen. In Marga Reis & Inger Rosengren (eds.), Fragesätze und Fragen, 223–242. Tübingen: Niemeyer. Meibauer, Jörg. 1994. Modaler Kontrast und konzeptuelle Verschiebung. Studien zur Syntax und Semantik deutscher Modalpartikeln. Tübingen: Niemeyer. Miller, Max. 1976. Zur Logik der frühkindlichen Sprachentwicklung: Empirische Untersuchungen und Theoriediskussion. Stuttgart: Klett. Müller, Gereon. 2011. Regeln oder Konstruktionen? Von verblosen Direktiven zur sequentiellen Nominalreduplikation. In Stefan Engelberg, Anke Holler & Kristel Proost (eds.), Sprachliches Wissen zwischen Lexikon und Grammatik, 211–249. Berlin: de Gruyter.

142 

 Hans-Martin Gärtner

Müller, Sonja. 2017. Combining ja and doch: A case of discourse structural iconicity. In Josef Bayer & Volker Struckmeier (eds.), Discourse particles: Formal approaches to their syntax and semantics, 203–240. Berlin: de Gruyter. Napoli, Donna Jo. 1982. Initial material deletion in English. Glossa 16(1). 85–111. Nikolaeva, Irina. 2014. The narrative infinitive construction in French and Latin. In Hans Boas & Francisco Gonzálvez-García (eds.), Romance perspectives on Construction Grammar, 139–179. Amsterdam: Benjamins. Nygård, Mari. 2014. Norwegian discourse ellipses in the left periphery – interacting structural and semantic restrictions. In Tor Åfarli & Brit Mæhlum (eds.), The sociolinguistics of grammar, 171–189. Amsterdam: Benjamins. van Olmen, Daniël. 2009. De imperativische infinitief in het Nederlands [The imperative infinitive in Dutch]. Nederlandse Taalkunde 14(2). 147–170. Oppenrieder, Wilhelm. 1989. Selbständige Verb-Letzt-Sätze: Ihr Platz im Satzmodussystem und ihre intonatorische Kennzeichnung. In Hans Altmann, Anton Batliner & Wilhelm Oppenrieder (eds.), Zur Intonation von Modus und Fokus im Deutschen, 163–244. Tübingen: Niemeyer. Platzack, Christer & Inger Rosengren. 1998. On the subject of imperatives: A minimalist account of the imperative clause. The Journal of Comparative Germanic Linguistics 1(3). 177–224. Portner, Paul. 2007. Instructions for interpretation as separate performatives. In Kerstin Schwabe & Susanne Winkler (eds.), On information structure, meaning, and form, 407–425. Amsterdam: Benjamins. Portner, Paul. 2009. Modality. Oxford: Oxford University Press. Rapp, Irene & Angelika Wöllstein. 2013. Satzwertige zu-Infinitivkonstruktionen. In Jörg Meibauer, Markus Steinbach & Hans Altmann (eds.), Satztypen des Deutschen, 338–355. Berlin: de Gruyter. Rau, Jennifer. 2005. Selbständige Infinitive im Spracherwerb. Tübingen: Eberhard-KarlsUniversität MA thesis. Reis, Marga. 1985. Satzeinleitende Strukturen im Deutschen. In Werner Abraham (ed.), Erklärende Syntax des Deutschen, 271–311. Tübingen: Narr. Reis, Marga. 1995. Über infinite Nominativkonstruktionen im Deutschen. In Olaf Önnerfors (ed.), Festvorträge anläßlich des 60. Geburtstags von Inger Rosengren (Sprache und Pragmatik. Arbeitsberichte Sonderheft), 114–156. Lund: University of Lund. Reis, Marga. 2002. What are we doing with wh-infinitives in German? Georgetown University Working Papers in Theoretical Linguistics 2. 287–341. Reis, Marga. 2003. On the form and interpretation of German wh-infinitives. Journal of Germanic Linguistics 15(2). 155–201. Reuland, Eric. 1983. Governing –ing. Linguistic Inquiry 14(1). 101–136. Rizzi, Luigi. 1993/94. Some notes on linguistic theory and language development: The case of root infinitives. Language Acquisition 3(4). 371–393. Rizzi, Luigi. 1997. The fine structure of the left periphery. In Liliane Haegeman (ed.), Elements of grammar, 281–337. Dordrecht: Kluwer. Rizzi, Luigi. 2001. Relativized minimality effects. In Mark Baltin & Chris Collins (eds.), The handbook of contemporary syntactic theory, 89–110. Oxford: Blackwell. Rizzi, Luigi. 2004. Locality and left periphery. In Adriana Belletti (ed.), Structures and beyond: The cartography of syntactic structures, vol. 3, 223–251. Oxford: Oxford University Press.



Root infinitivals and modal particles. An interim report  

 143

Struckmeier, Volker. 2014. Ja doch wohl C? Modal particles in German as C-related elements. Studia Linguistica 68(1). 16–48. Teuber, Oliver. 1998. fasel beschreib erwähn – Der Inflektiv als Wortform des Deutschen. Germanistische Linguistik 141/142. 7–26. Thurmair, Maria. 1989. Modalpartikeln und ihre Kombinationen. Tübingen: Niemeyer. Thurmair, Maria. 1991. ‘Kombinieren Sie doch nur ruhig auch mal Modalpartikeln!’: Combinatorial regularities for modal particles and their use as an instrument of analysis. Multilingua 10(1–2). 19–42. Thurmair, Maria. 1993. Äußerungsform oder Äußerungsfunktion? Zu den Bedingungen für das Auftreten von Modalpartikeln. Deutsche Sprache 21(1). 22–43. Thurmair, Maria. 2013. Satztyp und Modalpartikeln. In Jörg Meibauer, Markus Steinbach & Hans Altmann (eds.), Satztypen des Deutschen, 627–651. Berlin: de Gruyter. Truckenbrodt, Hubert. 2004. Zur Strukturbedeutung von Interrogativsätzen. Linguistische Berichte 199. 313–350. Truckenbrodt, Hubert. 2006a. On the semantic motivation of syntactic verb movement to C in German. Theoretical Linguistics 32(3). 257–306. Truckenbrodt, Hubert. 2006b. Replies to the comments by Gärtner, Plunze and Zimmermann, Portner, Potts, Reis, and Zaefferer. Theoretical Linguistics 32(3). 387–410. Uhmann, Susanne. 2010. Bitte einmal nachfassen. Professionelles Wissen und seine interaktive Vermittlung. Empraktische freie Infinitive im Operationssaal. In Ulrich Dausendschön-Gay, Christine Domke & Sören Ohlhus (eds.), Wissen in (Inter-)Aktion: Verfahren der Wissensgenerierung in unterschiedlichen Praxisfeldern, 37–69. Berlin: de Gruyter. Wilson, Deirdre & Dan Sperber. 1988. Mood and the analysis of non-declarative sentences. In Jonathan Dancy, Julius Moravcsik & Christopher Taylor, (eds.), Human agency, 77–101. Stanford, CA: Stanford University Press. Zaefferer, Dietmar. 1989. Untersuchung zur strukturellen Bedeutung deutscher Sätze. München: Ludwig-Maximilians-Universität Habilitation thesis. Zifonun, Gisela, Ludger Hoffmann & Bruno Strecker. 1997. Grammatik der deutschen Sprache. Berlin: de Gruyter. Zimmermann, Malte. 2004. Zum Wohl: Diskurspartikeln als Satztypmodifikatoren. Linguistische Berichte 199. 253–286. Zimmermann, Malte. 2009. Discourse particles in the left periphery. In Benjamin Shaer, Philippa Cook, Werner Frey & Claudia Maienborn (eds.), Dislocated elements in discourse, 200–231. Oxford: Routledge. Zimmermann, Malte. 2011. Discourse particles. In Klaus von Heusinger, Claudia Maienborn & Paul Portner (eds.), Semantics. An international handbook of natural language meaning, vol. 2, 2011–2038. Berlin: de Gruyter Mouton. Zimmermann, Malte. 2013. Ob-VL-Interrogativsatz. In Jörg Meibauer, Markus Steinbach & Hans Altmann, (eds.), Satztypen des Deutschen, 84–104. Berlin: de Gruyter.

Daniel Gutzmann

Modal particles ≠ modal particles (= modal particles) Differences between German modal particles and how to deal with them semantically It is an implicit assumption in the literature on German modal particles (MPs) that they behave homogeneously. This assumption, however, has been challenged recently. First, it has been noted that not all MPs exhibit strict speaker orientation and that some can receive a non-speaker oriented reading. Secondly, while some MPs (like wohl) seem to modify the sentence mood directly, others (like ja) have been argued to make an independent contribution. This non-uniform behavior challenges approaches that treat MPs alike. Against the background of a formal, multidimensional framework that analyses MPs as contributing expressive/useconditional content, the paper sketches ways in which the observed heterogeneous behavior can nevertheless be accounted for. Hence, it do not pose a problem for a uniform analysis of how MPs compose with other expressions.

1 Introduction Even though modal particles (MPs) constitute a relatively small and closed class of words, their existence can be viewed as one of the characteristic properties of German. Accordingly, MPs have received a lot of attention in both the traditional German linguistics literature (e.g., Autenrieth 2002; Helbig 1988; Thurmair 1989; Weydt 1969, 1989a) as well as in more theoretically oriented streams of linguistic

Acknowledgements: I would like to thank the audience at the workshop on particles at the ­annual DGfS meeting 2014 for helpful comments and discussion, especially Hans-Martin Gärtner and Manfred Krifka. Huge thanks also go to Robert Henderson, Klaus von Heusinger, and three reviewers, who provided valuable feedback on earlier versions of this paper, as well as to the editors, Josef Bayer and Volker Struckmeier, for all the work they put into this volume. Needless to say, my errors = my errors (=my errors). Daniel Gutzmann: Institut für Deutsche Sprache und Literatur I, Universität zu Köln, Köln, e-mail: [email protected] DOI 10.1515/9783110497151-007

Modal particles ≠ modal particles (= modal particles)  



 145

research (e.g., Bayer and Obenauer 2011; Bayer and Trotzke 2015; Coniglio 2011; Egg 2013; Repp 2013; Struckmeier 2014). Even if there is still some ongoing debate about their syntactic and semantic analysis, a result of the extensive body of research undertaken is that their behavior is well studied and there seems to be a canonical list of properties that most work seems to have settled upon (cf. e.g. Thurmair 1989: 37; Meibauer 1994: 29; Autenrieth 2002: 27). (1)

Characteristic properties of MPs: MPs ... a. are uninflectable. b. cannot receive main stress. c. occur only in the so-called middle field (Germ. Mittelfeld). d. occur commonly before the rheme. e can be combined with each other. f. cannot be coordinated. g. cannot be modified. h. are optional. i. cannot be negated. j. cannot be questioned. k. have sentential scope. l. do not affect truth-conditions. m. are speaker-oriented. n. modify the sentence mood.

Not all of these properties are interesting or unproblematic. The fact that modal particles cannot be inflected, (1a), is one of the necessary properties that makes them particles in the first place, but does not help to distinguish them from other particles like, for ­ otorious instance, focus particles. Regarding (1b), it is well-known that there are some n stressed versions of MPs, most notably DOCH, JA and SCHON, which are analyzed in Egg and Zimmermann 2012, Gutzmann 2010, and Féry 2010, and 2010 respectively.1 (2) A: Malte ist nicht nach Utrecht gefahren. ‘Malte didn’t go to Utrecht’. B: Er ist DOCH nach Utrecht gefahren. he is DOCH to Utrecht gone He DID go to trecht. (Egg and Zimmermann 2012: 226) (3) A: David riecht wie ein Zombie. ‘David smells like a zombie.’

1 See also Abraham (2017).

146 

 Daniel Gutzmann

B: David ist ja ein Zombie. David is ja a zombie ‘David is a Zombie, you should have known that!’ (Gutzmann 2010: 132) (4) A: ‘Do you like natto beans?’ B: Ja, die mag ich SCHON. yes, them like I SCHON ‘Yes, I do like them.’ (Féry 2010) There are also well-known exceptions to the middle-field restriction. Together with a wh-phrase, certain MPs like bloß, nur or schon can occur in the so-called prefield (Germ. Vorfeld), that is, the position that corresponds to CP, spec (Bayer and Obenauer 2011; Bayer and Trotzke 2015).2 (5) [Wo nur] ist der Stolz geblieben? where nur is the pride remained ‘Where on earth did the pride remain?’ (Bayer and Obenauer 2011: 481) Another exception is that some MPs may also occur inside DPs as modifiers in adjectives (see Viesel 2017). (6) a. diese ja nicht gerade einfache Geschichte this ja not exactly simple story ‘this not exactly simple story’ b. der wohl beste Gitarrist der Welt the WOHL best guitarist the world ‘probably the best guitar player of the world’ Regarding the optionality of MPs, (1h), optative clauses are a notable exception in which MPs are rather obligatory (Grosz 2012, 2013), in so far as optative clauses without MPs are judged as rather badly (Grosz 2013: 150). (7) a. Wenn es doch/nur/bloß wärmer wäre! if it DOCH/NUR/BLOsS warmer were ‘If only it were warmer!’ b. #Wenn es wärmer wäre.

2 On the syntax of MPs, see also Egg and Mursell (2017) for German, Scherf (2017) for Swedish and Paul and Pan (2017) for Mandarin Chinese.



Modal particles ≠ modal particles (= modal particles)  

 147

Similar judgments apply to deliberative ob-verb-last questions (Gutzmann 2011b: 80–81, Zimmermann 2013).3 (8) a. Ob sie wohl kommt? if she WOHL comes ‘If she will come, I wonder.’ b. ?Ob sie kommt? These are well-known exceptions to some of what has been assumed to be crucial properties of MPs.4 In this paper, I will add to these exceptions by focusing on the last two properties in (1), that is, speaker orientation (1m) and interaction with sentence mood (1n). First, regarding the former, it shall be noted at the get-go that this is rather unfortunate terminology, since many MPs target the hearer when they appear in interrogatives, as a reviewer rightly pointed out.5 Hence, we should think of this property more as utterance orientation. However, even under this revised notion, it is not the case that all MPs are necessarily utterance-oriented. While utterance orientation seems to be the default, there are also MPs that can receive an embedded interpretation, at least in some embedding contexts. Secondly, even though all MPs interact in intricate ways with the mood of the sentence in which they occur, they do not all do so in the same manner. As has previously been argued (e.g. Gutzmann 2008; Zimmermann 2004a), there are (at least) two ways in which MPs can interact with sentence mood. First, there are MPs that directly modify the sentence mood, just like (1n) suggests. But secondly, there are also MPs that add an independent, mood-like contribution to the sentence. It is only the overall interpretation that makes it seem as if the mood may be modified. My aims in this paper are twofold. First, I want to establish the non-homogeneous behavior of MPs regarding (1m) and (1n) in a bit more detail. These differences between MPs then pose a problem for formal approaches that treat all MPs the same with respect to how they compose with the rest of the sentence in which they occur. One such approach that prima facie is affected by this problem is the multidimensional approach that I presented in Gutzmann 2015, which analyzes MPs as contributing expressive or use-conditional meaning (Kaplan 1999; Potts 2007). My second goal is to show that this approach – which is motivated and

3 For more discussion, see Altmann (1993) and, more recently, Thurmair (2013). 4 For more discussion of the properties commonly ascribed to MPs, see Thurmair (2013). 5 This seems to be true for MPs like denn, which can only occur in interrogatives, or MPs like doch or wohl that occur in declaratives, where they target the speaker, but undergo an “epistemic flip” in interrogatives so that they relate to the hearer’s epistemic state (Zimmermann 2004a, 2004b).

148 

 Daniel Gutzmann

developed independently of the questions raised here – nevertheless is able to implement the observed differences in behavior. After giving a brief introduction into said approach, I make some suggestions about how they may nevertheless be accounted for in the same framework. In order to do so, I will first motivate the general idea of analyzing MPs as expressives in Section 2. After that, I will, in Section 3, provide a very brief introduction into the multidimensional framework, that outlines the basic ingredients, without however motivating why the system is set up as it is. Afterwards, I will discuss the behavior of MPs regarding utterance orientation and speech act modification in Section 4, and then sketch how the heterogeneous data might be implemented in this framework in Section 5. Section 6 concludes and points out some areas for further research on these matters.

2 Modal particles and expressives Many of the characteristic properties of MPs listed in (1) are shared by expressions that contribute expressive instead of descriptive content and, accordingly, MPs have been analyzed as being expressives as well (Gutzmann 2009, 2013; Kratzer 1999, 2004; McCready 2012; Zimmermann 2004a). To see this, let us first check the prototypical properties of expressive meaning. (9) Characteristic properties of expressives (Potts 2007: 166–167) a.  Independence: Expressive content contributes a dimension of meaning that is separate from the regular descriptive content. b.  Nondisplaceability: Expressives predicate something of the utterance situation. c.  Perspective dependence: Expressive content is evaluated from a particular perspective. In general, the perspective is the speaker’s, but there can be deviations if conditions are right. d.  Descriptive ineffability: Speakers are never fully satisfied when they paraphrase expressive content using descriptive, i.e., nonexpressive, terms. e.  Immediacy: Like performatives, expressives achieve their intended act simply by being uttered; they do not offer content so much as inflict it. f.  Repeatability: If a speaker repeatedly uses an expressive item, the effect is generally one of strengthening the emotive content, rather than one of redundancy. Let us go through this list and check in how they apply to MPs as well. Independence targets the fact that expressives (at least non-mixed ones, cf. Gutzmann 2011a;



Modal particles ≠ modal particles (= modal particles)  

 149

McCready 2010), do not affect the truth-conditions of an utterance. For instance, the following sentences are true if Kaplan got promoted, regardless of the negative attitude expressives like damn or bastard express regarding their argument. Ø } { } { { } { bastard } (10) That { } Kaplan got promoted. { damn } } { } { { fucking }

Besides this intuition, the independence of expressive content can be witnessed by the fact they cannot be directly challenged in discourse by a plain denial. An expressive like damn cannot targeted by simple negation.6

(11) A: That damn Kaplan got promoted. B: #No, I like him. The fact that they cannot be interpreted under negation, make them similar to presuppositions. However, with presuppositions, there is an inherent dependency of the descriptive from the presupposed content, while expressives are truly independent. That is, one can agree to the asserted content, while then rejecting the negative attitude regarding Kaplan conveyed by damn. (12) A: That damn Kaplan got promoted. B: Yes, he got promoted, but nevertheless, I think he’s a fine guy. This is not possible for standard cases of presuppositions. (13) A: Kaplan won the race. B: #Yes, he won, but he didn’t participate. Independence also holds for MPs. First, speakers consistently have the very robust intuition that MPs do not effect the truth-conditions of an utterance. That is, the following variants all have the same truth-conditions. Ø } { } { { } { ja } } { } { (14) Kaplan wurde { halt } befördert. } { } { } { { } { doch } eben { }



‘Kaplan got ∅/mp promoted.’

6 Of course, expressive content, just like presuppositions or implicatures, may be challenged by more indirect ways, e.g. by What do you mean by ‘damn’ or Wait – why did you say ‘damn’?.

150 

 Daniel Gutzmann

As is the case for expressives, the contribution of an MP cannot be directly challenged. For instance, the contribution by ja may be rendered requiring that the propositional content “is either already part of shared knowledge, or else can be verified on the spot, given the extra-linguistic evidence” (Kratzer 2004: 126). This cannot be targeted by a plain denial.7 (15) A: Kaplan wurde ja befördert. K. got mp promoted ‘Kaplan got promoted, as you know.’ B: #Nein, das wusste ich nicht. no that knew I not ‘No, I didn’t know that.’ However, it seems possible to challenge the contribution by ja if one first accepts the propositional content, similar to how the expressive worked in (12). (16) A: Kaplan wurde ja befördert. B: Okay, aber das wusste ich nicht. okay but that knew I not ‘Okay, but I didn’t know that.’ Nondisplaceability says that expressive meaning generally cannot be displaced by linguistic means (Potts 2007: 169). That is, even inside linguistically expressed semantic contexts – like speech or attitude reports, modal or conditionalized statements, or reports of past events – expressives cannot be evaluated in that semantic context, but always convey something about the utterance context or the utterance itself.8 For illustration, consider the following examples (cf. Potts 2007: 170). (17) a. That damn Kaplan is late. #I like him. b. Maybe that damn Kaplan is late. (#Then again, maybe I like him.) c. Yesterday, that damn Kaplan was late. #Today, I like him again. In contrast to the expressive content, the descriptive content in (17a)–(17b) gets displaced by the semantic operator. Even if the utterance is true, it does not (necessarily) hold for the utterance situation that Kaplan is late.

7 Again, more indirect ways of challenging the content of ja are impossible. See the previous footnote. 8 That is, displacement in this sense means that the interpretation of an utterance is displaced from the actual utterance situation. It is often assumed to be one of the key features or “design features” of human languages (Hockett 1958: 579), setting them apart from other forms of communication like animal communication (Coleman 2006; Hockett 1958; Hockett and Altmann 1968; Yule 2006). It should not be confused with the syntactic notion of displacement, which is a descriptive notion concerning some movement operations.



Modal particles ≠ modal particles (= modal particles)  

 151

In addition, observe that presuppositions also behave differently. Crucially, presuppositions allow for non-global interpretation. Consider first the following example. (18)

[Context: Kaplan didn’t participate in the race.] David dreamed that Kaplan didn’t win the race.  David dreamed that Kaplan didn’t participate or win in the race. (local)  David dreamed that Kaplan participated in the race and didn’t win. (intermediate)

The context in (18) makes it clear that a global interpretation, in which the presupposition that Kaplan participated in the race project, is not feasible. However, the sentence is nevertheless fine, since it still has both a local and intermediate interpretation of the presupposition. Crucially, this does not hold for expressives and hence a similar example to the one in (19) renders an embedded expressive infelicitous.9 (19) [Context: Neither I nor David feel negatively towards Kaplan or whether he is promoted or not.] #David dreamed that damn Kaplan didn’t get promoted.  David dreamed that it was not the case that he dislikes Kaplan or that he get promoted. (local)  David dreamed that he disliked Kaplan and that Kaplan didn’t get promoted. (intermediate) Similar observations can be made for MPs. That they cannot be negated is actually one of their characteristic properties, as stated in (1i). But just like expressives, they can neither be deferred by modals or past tense. The following examples illustrate this. In addition to ja, I use doch, which marks the propositional content as previously known but inactive, and suggests that it may contradicting some recent discourse issue. (20) a. Kaplan wurde ja nicht befördert. #Das konntest Du nicht wissen. ‘Kaplan ja didn’t get promoted. #You could not have known that.’ b. Vielleicht wurde ja Kaplan befördert. #Aber vielleicht weißt Du das auch noch nicht. ‘Maybe Kaplan ja got promoted. Then again, maybe I didn’t know that.’

9 I used “neither I nor David” in the context to control for possible differences in speaker orientation, which I will discuss later.

152 

 Daniel Gutzmann

c. Gestern wurde Kaplan doch befördert. #Heute hast Du das aber nicht vergessen. ‘Yesterday, Kaplan DOCH got promoted. #But today, you certainly have not forgotten that.’

The reason for this is, arguably, that MPs always take widest scope (1k), and hence can never end up in the scope of negation, or a tense or modal operator. Obviously, nondisplaceability is closely connected to speaker or utterance orientation, since having an embedded interpretation is an instance of linguistic displacement. Speaker orientation is also featured in (1m) of the list of MP-properties and is viewed as a central property of MPs.10 However, and this will of course be one of the two main topics of this paper, the case for speaker orientation is not as clear as the one for displacement by other means. Interestingly, Potts (2007) already equipped the property of speaker orientation with the caveat clause that the perspective from which expressive content is evaluated does not always need to be the speaker if the right conditions hold. Without further contextual support, expressives receive a speaker-oriented interpretation, which is why the following sentence seems odd at least. Kresge believes that that damn Kaplan should be promoted. #I think he’s a (21)  nice guy. However, as Potts makes very clear, the negative attitude displayed by the expressive may be attributed to an individual other than the speaker. Consider, for instance, Kratzer’s example that predates Potts’s by several years. (22)  My father screamed that he would never allow me to marry that bastard Webster. (Kratzer 1999: 6; my emphasis, DG) The most sensible interpretation of (22) is one in which the expressive bastard receives an embedded interpretation. That is, the negative evaluation of Webster is likely to be attributed to the speaker’s father. This reading can be made clear with a continuation like the following. (23) But he always hated him, the more I loved him.

10 A reviewer asked whether one shouldn’t assume that MPs are modifiers of speech acts, following Jacobs’s (1991) approach which is couched in terms of illocutionary semantics. Space limitations do not permit me to give this discussion a proper treatment here, so I refer the reader to relevant discussion in Autenrieth (2002); Gutzmann (2008, 2015); Ormelius-Sandblom (1997); Thurmair (2013). As a quick argument, consider the impossibility of ja in rhetorical question or the impossibility of denn in rising declaratives.



Modal particles ≠ modal particles (= modal particles)  

 153

However, a subject interpretation of (22) is not the only one. Consider a context in which the speaker, after a divorce, feels very negatively, while the speaker’s father was always against the marriage because he think negatively about his daughter and that Webster is too good for her. Then, (22) will again get a speaker-oriented interpretation. The question is how this flexible interpretation can be accounted for. However, to answer this question properly and also tackle the question of non-utterance-oriented MPs, we need to have the details of the formal system in place. Therefore, I will postpone giving examples for the (un)embeddability of MPs for now until the more detailed discussion in Section 4. The next property in (9d), descriptive ineffability, concerns the observation that expressives cannot be perfectly paraphrased by descriptive vocabulary. This is directly reflected by the fact that MPs can hardly be paraphrased at all, as can be witnessed by the huge body of work on the proper translation of MPs.11 However, as already noted by Geurts (2007) in his reply to Potts’s article, this does not seem to be a good criterion to distinguish expressive from non-expressive meaning. It is hard to come up with completely satisfying paraphrases in (descriptive terms), even for many purely descriptive expressions, like the, at, or green (Geurts 2007: 210). The next property on Potts’s (2007) list, immediacy, construes expressive meaning as non-disputable. This contrasts with the descriptive content of an assertion, which is thought to propose an update to the discourse context by adding the propositional content to the common ground (cf., among many others, Farkas and Bruce 2010; Stalnaker 1978). Instead, expressive content is not put onto the discourse table for further discussion, acceptance, or rejection, but directly inflicts an update onto the discourse.12 That is, they impose an update, instead of proposing one, to use AnderBois, Brasoveanu, and Henderson’s (2015) terminology. As for MPs, it is hard to see whether this property applies to them or not. As already shown, it is clear that the contribution made by an MP cannot be accepted, challenged or refuted like ordinary descriptive content can. However, they do not impose an independent update on the common ground either, like

11 The most influential work here certainly is Weydt (1969). For an overview, see LiefländerKoistinen (2004), and for some examples, see, amongst many others, Baunebjerg and Wesemann (1983); Burkhardt (1995); Datcheva (2011); Kärnä (1983); Liefländer-Koistinen (1989); Reiter (1983); Weydt (1989a, 1989b). 12 It is because of this property that, in his previous work, Potts (2005) subsumed expressive content under the broader notion of non-at-issue content. However, as I argued in Gutzmann (2012: Ch. 9.3), the two distinctions between truth-conditional/descriptive vs. use-conditional/ expressive content on the one hand, and at-issue vs. non-at-issue are orthogonal to each other and all four possibilities are attest. The inference from use-conditionality to non-at-issueness is a default inference at best.

154 

 Daniel Gutzmann

damn or the appositives studied by AnderBois, Brasoveanu, and Henderson (2015) may do. Instead, MPs give additional information about how the propositional content relates to the common ground (Bárány 2009; Bross 2012; Egg 2013; Gast 2008; Repp 2013; Zeevat 2005).13 The last remaining property, repeatability, concerns the observation that proper expressives can be repeated without being redundant, instead enforcing the expressive effect. Potts (2007) provides the examples in (24), which contrast which the feeling of redundancy one gets with the descriptive counterpart in (25). (24) a. Damn, I left my keys in the car. b. Damn, I left my damn keys in the car. c. Damn, I left my damn keys in the damn car. (Potts 2007: 182) (25) #I’m angry! I forgot my keys. I’m angry! They are in the car. I’m angry! (Potts 2007: 182) As discussed by Geurts, repeatability is however not a good criterion to distinguish between descriptive and expressive content in general. For starters, it should be noted that some non-expressives are eminently repeatable; an obvious case in point is the entire class of definites, including anaphoric pronouns, indexicals, and names. I am fairly sure that, wherever that bastard Kresge can be reiterated, the name Kresge can be used, too. This makes it [...] doubtful that the notion of repeatability can be sharpened so that it will separate expressive words from non-expressive ones. (Geurts 2007; 213)

In addition to Geurts’s point, there are also expressives that cannot be repeated. First, not all expressives can appear in as many syntactic positions as damn can. Repeating them leads to much more marked constructions. So, while both (26a) and (26b) are perfectly fine, the combination in (26c) is rather odd, while (26d) really feels redundant, and (26e) just is ungrammatical.14 (26) a. That bastard Kaplan got promoted, that bastard. b. Kaplan got promoted, that bastard. c. #That bastard Kaplan got promoted, that bastard.

13 This is why, instead of the traditional term modal particles, the particles in question are often called discourse particles, which seems descriptively more apt, but may suggest to cover other particles with discourse functions as well, which syntactically speaking are not MPs. See the discussion in Abraham (2017). 14 There is compound reading of bastard bastard akin to the famous salad salad (Ghomeshi et al. 2004).

Modal particles ≠ modal particles (= modal particles)  



 155

d. #Kaplan got promoted, that bastard, that bastard. e. *That bastard bastard Kaplan got promoted. In addition, for some interjections, like oops, a repetition evokes a feeling of redundancy as well. (27) #Oops! I forgot my keys. Oops! They are in the car. Oops! (Gutzmann 2013: 47) That is, repeatability does not seem to be a definite criterion for expressive content after all Hence, the fact that MPs usually cannot be repeated at will does not constitute an obstacle for a use-conditional analysis. doch doch } { } { { } { halt halt } (28) *Kaplan wurde { befördert. } } { ja ja } { } { { wohl wohl }



‘Kaplan got MP MP promoted.’

To conclude, MPs share all the clear criteria for expressive meaning, only differing regarding those that do not hold strictly for what we think of as true expressives to begin with. Therefore, it is not a surprise that several suggestions to analyze MPs as contributing expressive meaning have been made (Gutzmann 2009, 2013; Kratzer 1999, 2004; McCready 2012; Zimmermann 2004a). In the following, I will sketch the basic ideas of such an approach and how one might be able to implement the observed flexible non-homogeneous behavior regarding speaker/ utterance orientation and sentence mood modification.

3 A multidimensional approach to MPs The hallmark of multidimensional approaches is that they feature a second meaning dimension for expressive/use-conditional content alongside the ordinary truth-conditional component. The general idea of such a hybrid semantics, as I call the simultaneous employment of truth- and use-conditions for semantic analyses, is rather independent of the actual formalization. However, in order to have a predictive theory, one needs an account of how the use-conditional content of an utterance is composed of the truth- and use-conditional content of its parts. In this section, I will briefly sketch the approach I developed in Gutzmann (2012, 2015), which I call LTU , and which can be regarded as a consequently multidimensional reformulation of the influential family of approaches developed by Potts

156 

 Daniel Gutzmann

(2005) and others (Gutzmann 2011a; McCready 2010). Since I am interested in the application of this approach to MPs, I will directly start using MPs for illustrating how the system works. However, as I note elsewhere (Gutzmann 2013), the term expressive meaning may not be well suited for MPs, since that term may suggest an emotional component, while for most MPs it is not really apparent that they concern the speaker’s emotions. Therefore, I suggest to adopt Recanati’s (2004) terminology and use the term use-conditional meaning as a more general term for conventional non-truth-conditional meaning. The basic idea of all hybrid frameworks is that use-conditional expressions like MPs can access the truth-conditional content in order to take an argument. What is crucial about this kind of expressions, which may be called functional expletive use-conditional items (Gutzmann 2013), is that they do not affect their argument from a purely truth-conditional perspective but rather place a use-conditional proposition into the use-conditional dimension which is independent from the truth-conditional content.15 That is, using an MP in an otherwise purely truthconditional sentence leads to a hybrid content that makes a meaningful contribution to both dimensions of meaning. That is, a simple example like in (29) gets a two-dimensional interpretation along the lines of (30), where the superscripted t and u specific the truth- and use-conditional content respectively.16 (29) Peter schläft ja. ‘Peter sleeps MP.’ (30) a. ||(29)||t = ⟦sleep(peter) : t⟧ b. ||(29)||u = ⟦ja(sleep(peter) : u⟧

The difference between truth-conditional and use-conditional content comes down to a lexical distinction in semantic types. In addition to the usual truthconditional types, LTU introduces a new basic type for use-conditional propositions

15 They are functional, since they need an argument, and expletive, because truth-conditionalwise they are empty. The notion of (certain) expressives being semantic expletives, comes from Cruse (2004; 57), who uses the term if an expression "possess only expressive and no descriptive meaning". This term should not be confused with syntactic expletives which are empty in every semantic dimension. See Gutzmann (2013) for a typology of kinds of use-conditional items. 16  In (30) and the following, bold face ja is the translation of the natural language ja into the intermediate logical language. If ja is interpreted by the interpretation function ⟦⋅⟧, we get its proper meaning, which can be rendered as a function from proposition p onto the set of contexts in which the speaker thinks that the addressee may already have known p. This constitutes the set of contexts in which ja can felicitously be used. The details of the particular use-conditions are hard to pin-down, but perfectly rendered lexical semantics for ja do not matter for the purposes of this paper.

Modal particles ≠ modal particles (= modal particles)  



 157

(which are sets of contexts) and a corresponding clause for the recursive definition of complex use-conditional types.17 (31) Ordinary truth-conditional types a. e, t, s are basic truth-conditional types. (entities, truth values, worlds) b. If σ, τ are truth-conditional types, 〈σ, τ〉 is a truth-conditional type. (32) Use-conditional types a. u is a basic use-conditional type.  (use-conditional proposition) b. If σ is a type and τ is a use-conditional type, 〈σ, τ〉 is a use-conditional type. The way the two meaning dimensions are calculated in LTU is by not using just two, but rather three dimensions.

} } }

(33) t-dimension ♦ s-dimension • u-dimension t-content

u-active content

satisfied u-content

I use the diamond and bullet that separate the three dimension as a notational nod towards the Potts (2005)/McCready (2010) tradition. Officially, though, the semantic representations are just triples. The first dimension, called t-dimension contains the ordinary truth-conditional content of an expression. The third dimension, called u-dimension functions as a store for all complete use-conditions. The second dimension, which I call s-dimension, is where the interaction between truth- and use-conditional content takes place and ensures that the entire semantic derivation proceeds in a compositional fashion.18 Depending on the expression in question, it may contain truthor use-conditional content. This second dimension is actually the most important ingredient of the system that helps to overcome the some of the compositional problems that are associated with the systems of Potts (2005). It helps to mitigate between truth- and use-conditional content that still is not saturated and is able to deal with all varieties of expressives without the introducing special composition rules for each case as it is done in the extensions of Potts’s system (Gutzmann 2011a; McCready 2010). However, since this is not the place to motivate the concrete

17 Following Potts’s (2005) framework, the definition of complex use-conditional types in (32b) does not allow for a mapping from truth- to use-conditional types, which captures the observation that once launched, use-conditional content cannot be brought back to the truthconditional level (cf. Barker, Bernardi, and Shan’s [2010: 111] principle of non-interaction). 18 I use “s” for this dimension for historical reasons, namely that this dimension plays a similar function as the shunting types do in McCready’s (2010) system. It might be better to think of s standing for the “store” that holds the content that is active for the calculation of use-conditional content.

158 

 Daniel Gutzmann

set-up of the framework, I have to refer the reader to Gutzmann (2015), where it is developed and motivated step by step. Crucially, for the purposes of the compositional system, every expression features all three dimensions, even if they may contain rather trivial material. For instance, a simple truth-conditional expression like sleep receives the following 3-dimensional representation in LTU.19 (34) sleep  sleep : 〈e, 〈s, t〉〉♦ sleep : 〈e, 〈s, t〉〉 • U The big U in the third dimension is a dummy element that denotes trivial useconditions that are fulfilled in every context of use (i.e., it denotes the set of all contexts). Note how in (34) the content of the t-dimension is replicated in the second dimension (the s-dimension), in order to make it available for useconditional expression. For instance, an MP like ja, whose 3-dimensional representation in LTU can be given as in (35). What is crucial there is that, in the case of functional expletives like ja, the truth-conditional dimension contains an identity function I on expression of the type that the argument of function in the s-dimension has.20 This models the fact that functional expletives leave the truthconditional content of their argument untouched. (35) ja  I〈s,t〉 ♦ ja : 〈〈s, t〉, 〈u〉〉 • U These 3-dimensional representations are composed according to two composition rules, called multidimensional application and use-conditional elimination respectively. (36) Multidimensional application α1 : 〈σ, τ〉 ♦ α2 : 〈ρ, ν〉 • α3 β1 : σ ♦ β2 : ρ • β3



α1 (β1) : τ ♦ α2 (β2) : ν • α3  β3

MA

19  Note that in the lexicon, we do not need to specify all three dimension for all expressions, since at least one dimension can always be inferred from the other ones. That is, we have at most two lexical dimensions, whereas the compositional system needs three compositional dimensions. This gap between lexicon and semantic composition is bridged by so-called lexical extension rules. They do not only help to keep the lexicon simple, but allow the implementation of some needed restrictions without which the system would be too liberal. I refer the reader to Gutzmann (2015) for the details. 20 An identity function Iσ on type σ is the function that maps every expression of type σ onto itself: Iσ = λxσ.x : 〈σ, σ〉.

Modal particles ≠ modal particles (= modal particles)  



 159

(37) Use-conditional elimination α1 ♦ α2 : u • α3 α1 ♦ α1 • α3  α2 The two arrow diagrams in Fig. 1 illustrate how these two rules compose the three meaning dimensions. The application rule (36) executes functional application in the first two dimensions, while the content in the u-dimensions (all use-conditional propositions) is simply merged by means of the use-conditional conjunction “”, which is interpreted intersectively. The purpose of the elimination rule (37) is to store away use-conditional content that has reached the (use-conditional) propositional level in the s-dimension into the third dimension (where it will be dragged along the derivation). It also refills the s-dimension by replicating the content from the t-dimension in the s-dimension, so that the truth-conditional content thereby becomes available for further applications again.

α

α

α

β

β

β

α

α

α =u

α (β )

α (β )

α

β

(a) Multidimensional application

α

α

α

α

(b) Use-conditional elimination

Fig. 1: Composition with multidimensional application and use-conditional elimination.

Equipped with these rules, a standard lexical introduction rule (Lx) and suitable 3-dimensional semantic representations, we have all the tools set up in order to provide a derivation for (29). In order to save space, I often write the three dimensions vertically, with the t-content at the bottom and the s- and u-content stacked upon it.21

21 I follow Gutzmann (2012) in using the bottom-to-top notation. There it was used to mirror an informal notation in which use-conditional content was written as a dimension “above” the truth-conditional cone. Note, furthermore, that, since ⟦I(α)⟧ = ⟦α⟧ and since ⟦U  φ⟧ = ⟦φ⟧, I do not drag the identity function and all trivial U elements along the entire derivation.

160  (38)

 Daniel Gutzmann Peter •U  peter : e peter : e

Lx

schläft •U  sleep : 〈e, 〈s, t〉〉 sleep : 〈e, 〈s, t〉〉

•U  sleep(peter) : 〈s, t〉 sleep(peter) : 〈s, t〉

Lx

MA

ja •U  ja : 〈〈s, t〉, u〉 I〈s, t〉

sleep(peter) : 〈s, t〉  ja(sleep(peter)) : u • U sleep(peter) : 〈s, t〉  sleep(peter) : 〈s, t〉 • ja(sleep(peter)) : u

Lx

MA UE

Let us have a closer look at the two crucial steps in the derivation of (38). The first one is the multidimensional application (MA) of the 3-dimensional representation of ja to the propositional content that Peter sleeps. (39)

•U  ja : 〈〈s, t〉, u〉 I〈s, t〉

•U  sleep(peter) : 〈s, t〉 sleep(peter) : 〈s, t〉

•U =  ja(sleep(peter)) : u sleep(peter) : 〈s, t〉

What happens here is the following. In the t-dimension at the bottom, the identity function that represents the “empty” truth-conditional dimension of ja, applies to the propositional expression sleep(peter) to return the same as the result. In the s-dimension, the use-conditional contribution of ja applies to sleep(peter). In the u-dimension, we get U  U, which is equivalent to U again. The result is a 3-dimensional representation that features a use-conditional expression of type u in the s-dimension and thus qualifies as input to use-conditional elimination (UE). This is the next important step and also the final step in (38). (40)

•U  ja(sleep(peter)) : u sleep(peter) : 〈s, t〉

UE

• ja(sleep(peter)) : u  sleep(peter) : 〈s, t〉 sleep(peter) : 〈s, t〉

Here, the procedure is as follows. The t-dimension remains unaffected. The expression of type u in the s-dimension, ja(sleep(peter)) : u, is stored into the third dimension, where it intersects via  with the U from the input expression (which is equivalent to just ja(sleep(peter)) : u). In addition, the s-dimension is refilled with the truth-conditional content. From the resulting 3-dimensional representation, we can get the truth- and use-conditional interpretation, by semantically interpreting the first and third dimension respectively.



Modal particles ≠ modal particles (= modal particles)  

 161

Without diving deeper into the details of the sketched framework, note that it, like Potts’s (2005) original system, captures the projection behavior of use-conditional items. In LTU, they cannot be embedded for the simple fact that the use-conditional content generated in the second (s-)dimension is stored into the third (u-)dimension once it reached propositional level and hence escapes higher embedding functions. However, what is crucial about this for the point I want to make in this paper is that, without further assumptions, it treats all MPs indifferently. Hence, we need to think a bit more about how to implement the non-homogeneous behavior regarding embedding and how they interact with sentence mood. Before I will sketch how it may be implemented in LTU, these differences will be discussed in the next section.

4 Heterogeneous behavior In this section, I will add to the set of properties from the list in (1) for which there are exceptions, by looking at the ability of MPs to receive a not speaker/ utterance-oriented, embedded interpretation (1m) and at the different ways they interact with the semantic mood of the sentence (1n). I will start with the embedding cases, before discussing sentence mood modification and the lack thereof.

4.1 Speaker orientation As has recently been discussed, mostly by Döring (2013) and Coniglio (2011), not all MPs behave the same regarding their speaker-orientedness. For instance, ja can hardly be shifted to a non-speaker. This holds for (most) verba dicendi, as well as for evidentials.22 (41) #Yoshi sagte, dass Luigi ja Zelda liebt (but I don’t believe that). Yoshi said that Luigi JA Zelda loves #‘Yoshi said, that (as we know) Luigi loves Zelda, but I don’t believe that.’

22 As a reviewer pointed out, (41) and many of the following examples improve if subjunctive mood is used in the embedded clause. I agree with the reviewer, that this effect, which is systematic, should be taken into account. However, in order to model the interaction between subjunctive and non-utterance-oriented interpretation, one need to capture the distancing effect of subjunctive in the present system, which I cannot do here. But see Potts (2005: 186–193) for a first suggestion for an expressive approach to subjunctive mood in German, which may provide a good starting point for formalizing the observed interaction.

162 

 Daniel Gutzmann

(42) #Laut Yoshi liebt Luigi ja Zelda (but I don’t believe that). according.to Yoshi loves Luigi JA Zelda #‘According to Yoshi, Luigi loves (as we know) Zelda, but I don’t believe that.’ The knowledge ascription of ja invariantly holds for the speaker, which explains why ja is odd in those sentences, due to pragmatic reasons. This is also shown by the fact, that the addition in the parentheses, which try to force an embedded interpretation, does not save the examples, which they would, if such a reading were available. In contrast to ja, an MP like wohl can be attributed to a non-speaker when it occurs in an embedded context or in the scope of evidential constructions. (43) Yoshi hat gesagt, dass Luigi wohl Zelda liebt. Yoshi has said that Luigi WOHL Zelda loves ‘Yoshi said that Luigi probably loves Zelda.’ (44) Laut Yoshi liebt Luigi wohl Zelda. according.to Yoshi loves Luigi WOHL Zelda ‘According to Yoshi, Luigi probably loves Zelda.’ That a non-speaker interpretation is available for these examples is illustrated by the observation that continuations enforcing either one or the other reading are possible. Consider, for instance for (43), the following two continuations. (45) a. Er war sich sicher diesbezüglich. he was himself confident about.that ‘He was confident about that.’ b. Ich bin mir sogar sicher diesbezüglich. I am myself even confident about.that ‘I am even confident about that.’ Space limitations do not permit me to go into the details of which MPs can receive a shifted interpretation and which cannot – I refer the reader to Döring (2013) for a detailed discussion using authentic examples – but for the purposes of this paper, the contrast between ja and wohl suffices to illustrate that there are at least some MPs that can be interpreted embeddedly.

4.2 Interaction with sentence mood In the literature on MPs in German, it is commonly assumed that they interact with sentence mood, as stated by (1n). However, how one should think about this interaction is not always spelled-out in any detail, and it is almost always silently



Modal particles ≠ modal particles (= modal particles)  

 163

assumed that whatever mechanism is used to model that interaction is the same for all MPs. But as has been discussed in different contexts (e.g. Zimmermann 2004a and Gutzmann 2008, 2012), MPs can be divided into two classes according to how they actually interact with sentence mood. First, there are those MPs that can genuinely be called sentence mood modifiers, since they directly alter the semantic mood of the sentence they occur in. In contrast, the other class may be called free modifiers or propositional MPs, because they do not alter the sentence mood directly, but rather apply to the propositional content and add a free and independent use-conditional proposition to the overall use-conditions of a sentence. I will refer to them as mood particles and propositional particles, respectively. Again, ja and wohl serve as examples for each category. For instance, ja is a typical example of a propositional particle. It combines with a proposition and yields an independent use-conditional comment to it without changing the rest of the propositional content. (46) Luigi hat Zelda ja schon immer geliebt. Luigi has Zelda JA already always loved ‘Luigi always loved Zelda (and you may already have known that).’ We already saw a simple derivation involving ja in (38) that illustrated how ja leads to multidimensional content. However, I left out sentence mood from the picture in the previous section. In order to illustrate the difference between propositional and mood particles, we need sentence mood to be represented in the semantic derivation as well though. The basic idea is that mood operators are use-conditional expressions as well that apply to the sentence’s propositional content, in the same way as ja does.23 (47) [ DECL [Luigi loves Zelda ] ] (48) a. ||(47)||t = ⟦love(zelda)(luigi) : t⟧ b. ||(47)||u = ⟦decl(love(zelda)(luigi)) : u⟧

Hence, for a simple declarative involving not only DECL, but also ja, like (46), we get two use-conditions that will end up being conjoined in the third dimension.

23 The details of the semantics of mood do not matter for the present purposes, which is why I just use DECL as a placeholder. For an explicit proposal, see Gutzmann (2015) and, for an alternative modeling, Murray (2014).

164  (49)

 Daniel Gutzmann ⁝ •U  love(zelda)(luigi) : 〈s, t〉 love(zelda)(luigi) : 〈s, t〉

ja •U  ja : 〈〈s, t〉, u〉 I〈s, t〉

•U  ja(love(zelda)(luigi)) : u love(zelda)(luigi) : 〈s, t〉 •U  decl : 〈〈s, t〉, u〉 I 〈s, t〉

Lx

MA

• ja(love(zelda)(luigi)) : u  love(zelda)(luigi) : 〈s, t〉 love(zelda)(luigi) : 〈s, t〉 • ja(love(zelda)(luigi)) : u  decl(love(zelda)(luigi)) : u love(zelda)(luigi) : 〈s, t〉

• ja(love(zelda)(luigi))  decl(love(zelda)(luigi)) : u  love(zelda)(luigi) : 〈s, t〉 love(zelda)(luigi) : 〈s, t〉 The content provided by ja and the assertive mood of the declarative thus independently add use-conditions to the meaning profile of the entire utterance. Taken together, they form complex use-conditions for the entire sentence.24 In contrast to this, wohl is a mood particle, because it can be thought of as directly modifying the sentence mood, for instance, by lowering the knowledge threshold required for felicitous assertions (in case of declaratives), or felicitous answers to a question (in case of interrogatives). immer geliebt. (50) Luigi hat Zelda wohl schon Luigi has Zelda WOHL already always loved ‘Presumably, Luigi always loved Zelda.’ That wohl modifies the mood can be seen by the fact that an utterance containing wohl is licensed in contexts in which a plain assertion or question is not. For instance, if one explicitly mentions one’s uncertainty or lack of evidence,

24 The use-conditional approach to sentence mood sketched here is much simplified version of the analysis of sentence mood I developed in Gutzmann (2015). For the present purposes, it suffices to assume that mood operators are “just there”.



Modal particles ≠ modal particles (= modal particles)  

 165

using a plain declarative is infelicitous, while the addition of wohl rescues the utterance.25 (51) Ich bin mir nicht ganz sicher, aber Luigi liebt I am me not entirely sure but Luigi loves #(wohl) Zelda. WOHL Zelda ‘I am not entirely sure, but Luigi presumably loves Zelda.’ (52) [I know you don’t know Lothar that much,] aber wird ihm diese Wildschweinskulptur #(wohl) gefallen? but will him this wild.boar.sculpture WOHL appeal.to ‘but will this wild-boar sculpture presumably appeal to him?’ Under the current use-conditional view on sentence mood, mood particles pose a problem for the standard Pottsean account of expressives (Potts 2005), since it does not allow for expressions that can take expressive/use-conditional content as their argument. Even if the Pottsean approach can be extended to cover such expressive modifiers as well (Gutzmann 2011a), the sketched multidimensional approach can directly handle such cases with its set of two composition rules. I will come back to this issue when I show how the differences can be accounted for in LTU in the following section.

4.3 Summary Prima facie, the attested heterogeneous behavior poses problems for an expressive, multidimensional analysis. In those systems, including the logic LTU sketched here, it is predicted that all MPs always have wide scope. If some MPs can be embedded and do not take scope at the highest sentential level, how can they be handled by the same system? Furthermore, the difference between propositional and mood particles seems to pose a problem, at least for Potts’s (2005) standard system. Does that mean that a multidimensional expressive approach is not viable?

25 A reviewer asked if this would commit me to conclude that möglicherweise ‘possibly’ or es ist möglich ‘it is possible’ as modify the mood as well, because they would rescue an assertion as well. I would say yes and no. As has recently been argued by Wolf (2014), modal adverbs like possibly modify the strength with which an assertion is made (similar to wohl), while other modal expressions modify the propositional content itself. Even if different, both are strategies to warrant an assertion in view of uncertainty. Similar observations apply to other hedges like sluiced I think.

166 

 Daniel Gutzmann

5 Accounting for differences I think that the answer to this question is no. The tools offered by expressive approaches in general, and LTU in particular, actually allow for enough flexibility to account for these differences in a natural way, without the need for new ad hoc mechanisms that only account for the differences between MPs. To account for the two sources of non-homogeneity amongst MPs, we need to take a closer look at the lexical semantics of the MPs and what the application rules of LTU allow us to compose. In the following, I will show that focusing on these aspects leads us to a natural incorporation of the observed data in the multidimensional framework of LTU.

5.1 “Context shifts” The strong prediction of speaker-orientation made by multidimensional approaches in the Pottsean tradition, including LTU, is not only challenged by the embeddability of some MPs, but also by the apparent shiftability of expressive content as well, as already shown by example (22) above, repeated here as (53) for convenience. (53) My father screamed that he would never allow me to marry that bastard Webster. However, instead of modifying the underlying combinatorics to capture such cases, Harris and Potts (2009) suggest (and tested) that shifted expressives as in (22) are not actually shifted after all. Following Potts 2007, they instead assume that such expressives do not express a speaker attitude, but the attitude of the ­so-called contextual judge (cJ ), which they adopt from Lasersohn’s (2005) work on predicates of personal taste. In most cases, the judge actually is the speaker, so that we get speaker orientation as a default. However, the judge can be shifted to another discourse entity – for instance to the subject of a speech report – if it is both salient enough and makes sense as the source of the attitude. If this is the case, as in (53), where the father and his emotion are not only salient but a subject-oriented interpretation is also pretty reasonable, the father my become the judge. We thus get a virtually embedded interpretation (i.e. the attitude is ascribed to the subject of the matrix clause), while crucially it is still interpreted at matrix level. It just happens to be attributed to the subject of the attitude report who is the contextual judge. This simulates the effect of a shifted interpretation, while the expressive actually still scopes out of the attitude context.



Modal particles ≠ modal particles (= modal particles)  

 167

(54) 〈My father screamed that he would never allow me to marry ___ Webster, cJ [=the father] has a negative attitude towards Webster〉 One argument for implementing the shifting at the discourse level is that expressives can receive a non-speaker-oriented interpretation even if they are syntactically not embedded and the individual to which they are shifted is not mentioned in the sentence at all. (55) Peter really hates Mary’s boyfriend. However, Mary really loves that bastard Webster. An approach to the “shiftability” of expressions along these lines offers us a natural place to account for the differences amongst MPs. The basic idea is that the distinction between shiftable and unshiftable MPs comes down to a distinction of what kind of attitude holder is specific in the lexical semantics of an MP. On the one hand, if an MP attributes an attitude to the contextual judge like expressives do, it is expected to be shiftable in a similar way. On the other hand, if an MP cannot be shifted – ja serves as our example here – it invariantly attributes an attitude to the speaker, and thus the pseudo-shift reading in which the judge happens to be the subject of an attitude verb is not an option for such an MP. That is, a distinction between judge-referring MPs and speaker-referring MPs is a straightforward way to account for the difference in shiftability between ja and wohl.26

5.2 Mood particles Potts’s (2005) approach to expressives does not allow for expressive content to be an argument. However, as I argued elsewhere, we might need to extend his system to also cover what I called expressive modifiers. In particular, we want to account for examples like fucking bastard, in which fucking intensifies the expressive attitude expressed by bastard. The system of LTU already has this option built in, as can be witnessed from the type definitions. This is of importance for the present purposes since, as already mentioned above, I consider mood operators to contribute use-conditional content as well. If mood particles like wohl modify

26 This approach bears some resemblance to Eckardt’s (2015) work on context shifts in indirect speech, where she distinguishes between shiftable context-reference (with a lower case “c”) and unshiftable Context-reference (with a capital “c”).

168 

 Daniel Gutzmann

those operators, they are another instance of expressive modifiers besides those I discussed in Gutzmann (2011a). Given these assumptions, the system of LTU allows one to derive a sentence containing mood particles like wohl. (56)

•U

•U

 wohl : 〈〈〈s, t〉, u〉, 〈〈s, t〉, u〉

 decl : 〈〈s, t〉, u〉

I〈〈s, t〉 〈s, t〉〉

I〈s, t〉

•U

⁝

MA

•U

 wohl (decl) : 〈〈s, t〉, u〉

 love(zelda)(luigi) : 〈s, t〉

I 〈s, t〉

love (zelda)(luigi) : 〈s, t〉 •U

MA

 wohl (decl) (love(zelda)(luigi)) : u love(zelda)(luigi) : 〈s, t〉 • wohl (decl) (love(zelda)(luigi)) : u  love(zelda)(luigi) : 〈s, t〉 love(zelda)(luigi) : 〈s, t〉

That is, the system of LTU easily allows us to have both, mood and propositional particles, and the differences between them boils down to a difference between semantic types.

6 Conclusion Modal particles are often viewed as a homogeneous class. However, at least with respect to the two properties this paper focuses on – speaker orientation and interaction between sentence mood – there is some heterogeneous behavior between MPs that at first sight poses challenges to such approaches. Some MPs are shiftable, while others are not. Some MPs modify the proposition directly, while others modify the sentence mood. However, as I have illustrated, these observed differences can be accounted for by multidimensional approaches that were built in order to capture the expressive-like use-conditional nature of MPs. Regarding shiftability, the differences can be accounted for by a different lexical anchoring of the attitude holder. Either an MP refers to the contextual judge, and thus is predicted to be shiftable, or an MP refers to the speaker and hence is unshiftable. The difference between



Modal particles ≠ modal particles (= modal particles)  

 169

mood and propositional particles can also be analyzed in a system like LTU, by assuming a type difference between those classes of particles. Some topics for further research include the investigation of the syntactic consequences of the discussed semantic heterogeneity, like scoping behavior and conditions for shifted interpretation, as well as the syntactic mechanisms that connect MPs with their higher arguments.

References Abraham, Werner. 2017. Discourse marker = discourse particle = thetical =modal particle? A futile comparison. In Josef Bayer & Volker Struckmeier (eds.), Discourse particles: Formal approaches to their syntax and semantics, 241–280. Berlin: de Gruyter. Altmann, Hans. 1993. Satzmodus. In Joachim Jacobs, Arnim von Stechow, Wolfgang Sternefeld & Theo Vennemann (eds.), Syntax. Ein Internationales Handbuch der zeitgenössischen Forschung. Syntax. An international handbook of contemporary research, 1006–1029. Berlin: de Gruyter. AnderBois, Scott, Adrian Brasoveanu & Robert Henderson. 2015. At-issue proposals and appositive impositions in discourse. Journal of Semantics 32(1). 93–138. doi: http://dx. doi.org/10.1093/jos/fft014 (accessed 13 July 2016). Autenrieth, Tanja. 2002. Heterosemie und Grammatikalisierung bei Modalpartikeln. Eine synchrone und diachrone Studie anhand von eben, halt, echt, einfach, schlicht und glatt. Tübingen: Niemeyer. Bárány, András. 2009. Form and interpretation of the German discourse particles ja, doch and wohl. Wien: University of Vienna MA thesis. Barker, Chris, Raffaella Bernardi & Chung-chieh Shan. 2010. Principles of interdimensional meaning interaction. Proceedings of Semantics and Linguistic Theory 20. 109–121. doi: http://dx.doi.org/10.3765/salt.v20i0.2569 (accessed 13 July 2016). Baunebjerg, Gitte & Monika Wesemann. 1983. Partikelwörterbuch deutsch-dänisch, dänischdeutsch. Ein Arbeitsbericht. In Harald Weydt (ed.), Partikeln und Interaktion, 119–129. Tübingen: Niemeyer. Bayer, Josef & Hans-Georg Obenauer. 2011. Discourse particles, clause structure, and question types. The Linguistic Review 28(4). 449–491. doi: http://dx.doi.org/10.1515/tlir.2011.013 (accessed 13 July 2016). Bayer, Josef & Andreas Trotzke. 2015. The derivation and interpretation of left peripheral discourse particles. In Josef Bayer, Roland Hinterhölzl & Andreas Trotzke (eds.), Discourseoriented syntax, 13–40. Amsterdam: Benjamins. Bross, Fabian. 2012. German modal particles and the common ground. Helikon. A Multidisciplinary Online Journal 2. 182–209. Burkhardt, Armin. 1995. Zur Übersetzbarkeit von Abtönungspartikeln. Am Beispiel von Hoffmannsthals ‘Der Schwierige’. Zeitschrift für germanistische Linguistik 23(2). 172–201. Coleman, John. 2006. Design features of language. In Keith Brown (ed.), Encyclopedia of language & linguistics. Second edition, 471–475. Oxford: Elsevier. doi: http://dx.doi.org/ 10.1016/B0-08-044854-2/04743-X (accessed 13 July 2016).

170 

 Daniel Gutzmann

Coniglio, Marco. 2011. Die Syntax der deutschen Modalpartikeln. Ihre Distribution und Lizenzierung in Haupt- und Nebensätzen. Berlin: Akademie Verlag. Cruse, David Alan. 2004. Meaning in language. An introduction to semantics and pragmatics. Second edition. Oxford: Oxford University Press. Datcheva, Galia. 2011. Mal, wohl und ihre bulgarischen Entsprechungen. Münster: Waxmann. Döring, Sophia. 2013. Modal particles and context shift. In Daniel Gutzmann & Hans-Martin Gärtner (eds.), Beyond expressives. Explorations in use-conditional meaning, 95–123. Leiden: Brill. Eckardt, Regine. 2015. The semantics of free indirect speech. How texts let you read minds and eavesdrop. Leiden: Brill. Egg, Markus. 2013. Discourse particles, common ground, and felicity conditions. In Daniel Gutzmann & Hans-Martin Gärtner (eds.), Beyond expressives. Explorations in use-conditional meaning, 125–149. Leiden: Brill. Egg, Markus & Johannes Mursell. 2017. The syntax and semantics of discourse particles. In Josef Bayer & Volker Struckmeier (eds.), Discourse particles: Formal approaches to their syntax and semantics, 15–48. Berlin: de Gruyter. Egg, Markus & Malte Zimmermann. 2012. Stressed out! Accented discourse particles. The case of DOCH. Proceedings of Sinn und Bedeutung 16. 225–238. Farkas, Donka F. & Kim B. Bruce. 2010. On reacting to assertions and polar questions. Journal of Semantics 27(1). 81–118. doi: http://dx.doi.org/doi:10.1093/jos/ffp010 (accessed 13 July 2016). Féry, Caroline. 2010. Information structure of schon. In Thomas Hanneforth & Gisbert Fanselow (eds.), Language and Logos. A Festschrift for Peter Staudacher, 160–176. Berlin: Akademie Verlag. Gast, Volker. 2008. Modal particles and context updating – the function of German ja, doch, wohl and etwa. In Heinz Vater & Ole Letnes (eds.), Modalverben und Grammatikalisierung, 153–177. Trier: Wissenschaftlicher Verlag Trier. Geurts, Bart. 2007. Really fucking brilliant. Theoretical Linguistics 33(2). 209–214. doi: http: // dx.doi.org/10.1515/TL.2007.013 (accessed 13 July 2016). Ghomeshi, Jila, Ray Jackendoff, Nicole Rosen & Kevin Russell. 2004. Contrastive focus reduplication in English (the salad-salad paper). Natural Language and Linguistic Theory 22(2). 307–357. doi: http://dx.doi.org/10.1023/B:NALA.0000015789.98638.f9 (accessed 13 July 2016). Grosz, Patrick. 2012. On the grammar of optative constructions. Amsterdam: Benjamins. Grosz, Patrick. 2013. Optativsatz. In Jörg Meibauer, Markus Steinbach & Hans Altmann (eds.), Satztypen des Deutschen, 146–170. Berlin: de Gruyter. Gutzmann, Daniel. 2008. On the interaction between modal particles and sentence mood in German. Mainz: University of Mainz MA thesis. Gutzmann, Daniel. 2009. Hybrid semantics for modal particles. Sprache und Datenverarbeitung 33(1–2). 45–59. Gutzmann, Daniel. 2010. Betonte Modalpartikeln und Verumfokus. In Elke Hentschel & Theo Harden (eds.), 40 Jahre Partikelforschung, 119–138. Tübingen: Stauffenburg. Gutzmann, Daniel. 2011a. Expressive modifiers & mixed expressives. In Olivier Bonami & Patricia Cabredo-Hofherr (eds.), Empirical Issues in Syntax and Semantics 8, 123–141. Paris: CNRS. url: http://www.cssp.cnrs.fr/eiss8/gutzmann- eiss8.pdf (accessed 13 July 2016). Gutzmann, Daniel. 2011b. Ob einerwohl recht hat? Zwei Satzmodustheorien für das Deutsche im Vergleich. Deutsche Sprache 39(1). 65–84.



Modal particles ≠ modal particles (= modal particles)  

 171

Gutzmann, Daniel. 2012. Use-conditional meaning. Studies in multidimensional semantics. Frankfurt: University of Frankfurt dissertation. Gutzmann, Daniel. 2013. Expressives and beyond. An introduction to varieties of use-conditional meaning. In Daniel Gutzmann & Hans-Martin Gärtner (eds.), Beyond expressives. Explorations in use-conditional meaning, 1–58. Leiden: Brill. Gutzmann, Daniel. 2015. Use-conditional meaning. Studies in multidimensional semantics. Oxford: Oxford University Press. Harris, Jesse A. & Christopher Potts. 2009. Predicting perspectival orientation for appositives. Proceedings from the Annual Meeting of the Chicago Linguistic Society 45(1). 207– 221. Helbig, Gerhard. 1988. Lexikon deutscher Partikeln. Leipzig: Verlag Enzyklopädie. Hockett, Charles. 1958. A course in modern linguistics. New York: Macmillan. Hockett, Charles & Stuart A. Altmann. 1968. A note on design features. In Thomas A. Sebeok (ed.), Animal communication. Techniques of study and results of research, 61–72. Bloomington, IN: Indiana University Press. Jacobs, Joachim. 1991. On the semantics of modal particles. In Werner Abraham (ed.), Discourse particles, 141–162. Amsterdam: Benjamins. Kaplan, David. 1999. Themeaning of ouch and oops. Explorations in the theory ofmeaning as use. 2004 version. Ms., University of California, Los Angeles. Kärnä, Aino. 1983. Abtönung im Finnischen und im Deutschen. In Harald Weydt (ed.), Partikeln und Interaktion, 85–95. Tübingen: Niemeyer. Kratzer, Angelika. 1999. Beyond ouch and oops: How descriptive and expressive meaning interact. Paper presented at the Cornell Conference on Theories of Context Dependency, Ithaca, NY, March 26, 1999. url: http://semanticsarchive.net/Archive/WEwNGUyO/ (accessed 13 July 2016). Kratzer, Angelika. 2004. Interpreting focus: Presupposed or expressive meanings? A comment on Geurt and van der Sandt. Theoretical Linguistics 30(1). 123–136. doi: http://dx. doi. org/10.1515/thli.2004.002 (accessed 13 July 2016). Lasersohn, Peter. 2005. Context dependence, disagreement, and predicates of personal taste. Linguistics and Philosophy 28(6). 643–686. doi: http://dx.doi.org/10.1007/s10988005-0596-x (accessed 13 July 2016). Liefländer-Koistinen, Luise. 1989. Zum deutschen doch und finnischen -han. Beobachtungen zur Übersetzbarkeit der deutschen Partikel. In Harald Weydt (ed.), Sprechen mit Partikeln, 185–195. Berlin: de Gruyter. Liefländer-Koistinen, Luise. 2004. Modalpartikeln als Übersetzungsproblem. In Harald Kittel, Armin Paul Frank, Norbert Greiner, et al. (eds.), Übersetzung – Translation – Traduction. Ein internationales Handbuch zur Übersetzungsforschung / An international encyclopedia of translation studies / Encyclopédie internationale de la recherche sur la traduction. Vol. 1, 550–554. Berlin: de Gruyter. McCready, Eric. 2010.Varieties of conventional implicature. Semantics and Pragmatics 3(8). 1–57. doi: http://dx.doi.org/10.3765/sp.3.8 (accessed 13 July 2016). McCready, Eric. 2012. Formal approaches to particle meaning. Language and Linguistics Compass 6(12). 777–795. doi: http://dx.doi.org/10.1002/lnc3.360 (accessed 13 July 2016). Meibauer, Jörg. 1994. Modaler Kontrast und konzeptuelle Verschiebung. Studien zur Syntax und Semantik deutscher Modalpartikeln. Tübingen: Niemeyer. Murray, Sarah E. 2014. Varieties of update. Semantics and Pragmatics 7. doi: http://dx.doi. org/10.3765/sp.7.2 (accessed 13 July 2016).

172 

 Daniel Gutzmann

Ormelius-Sandblom, Elisabet. 1997. Die Modalpartikeln ja, doch, schon. Zu ihrer Syntax, Semantik und Pragmatik. Stockholm: Almqvist & Wiksell. Paul, Waltraud & Victor Pan. 2017. What you see is what you get: Chinese sentence-final particles as head-final complementisers. In Josef Bayer & Volker Struckmeier (eds.), Discourse particles: Formal approaches to their syntax and semantics, 49–77. Berlin: de Gruyter. Potts, Christopher. 2005. The logic of conventional implicature. Oxford: Oxford University Press. Potts, Christopher. 2007. The expressive dimension. Theoretical Linguistics 33(2). 165–197. doi: http://dx.doi.org/10.1515/TL.2007.011. (accessed 13 July 2016). Recanati, François. 2004. Pragmatics and semantics. In Laurence R. Horn & Gregory Ward (eds.), The handbook of pragmatics, 442–462. Oxford: Blackwell. Reiter, Norbert. 1983. Pragmatische Norm und fakultative Setzung von Partikeln, untersucht an Übersetzungen aus dem Russischen. In Harald Weydt (ed.), Partikeln und Interaktion, 96–105. Tübingen: Niemeyer. Repp, Sophie. 2013. Common ground management: Modal particles, illocutionary negation and verum. In Daniel Gutzann & Hans-Martin Gärtner (eds.), Beyond expressives. Explorations in use-conditional meaning, 231–274. Leiden: Brill. Scherf, Nathalie. 2017. The syntax of Swedish modal particles. In Josef Bayer & Volker Struckmeier (eds.), Discourse particles: Formal approaches to their syntax and semantics, 78–99. Berlin: de Gruyter. Stalnaker, Robert. 1978. Assertion. In Peter Cole (ed.), Syntax and semantics 9: Pragmatics, 315–332. New York: Academic Press. Struckmeier, Volker. 2014. Ja doch wohl C? Modal particles in German as C-related elements. Studia Linguistica 68(1). 16–48. Thurmair, Maria. 1989. Modalpartikeln und ihre Kombinationen. Tübingen: Niemeyer. Thurmair, Maria. 2013. Satztyp und Modapartikeln. In Jörg Meibauer, Markus Steinbach & Hans Altmann (eds.), Satztypen des Deutschen, 627–651. Berlin: de Gruyter. Viesel, Yvonne. 2017. Discourse particles ‘embedded’: German ja in adjectival phrases. In Josef Bayer & Volker Struckmeier (eds.), Discourse particles: Formal approaches to their syntax and semantics, 173–202. Berlin: de Gruyter. Weydt, Harald. 1969. Abtönungspartikel. Die deutschen Modalwörter und ihre französischen Entsprechungen. Vol. 4. Bad Homburg: Gehlen. Weydt, Harald, (ed.) 1989a. Sprechen mit Partikeln. Berlin and New York: de Gruyter. Weydt, Harald. 1989b. Was soll der Übersetzer mit deutschen Partikeln machen? – Nachts schlafen die Ratten doch als Beispiel. In Andrzej Katny (ed.), Studien zur kontrastiven Linguistik und literarischen Übersetzung, 235–252. Frankfurt: Lang. Wolf, Lavi. 2014. Degrees of Assertion. Be’ er Scheva: Ben-Gurion University of the Negev dissertation. Yule, George. 2006. The study of language. 3rd ed. Cambridge: Cambridge University Press. Zeevat, Henk. 2005. A dynamic approach to discourse particles. In Kerstin Fischer (ed.), Discourse particles, 133–147. Amsterdam: Elsevier. Zimmermann, Malte. 2004a. Discourse particles in the left periphery. In Benjamin Shaer, Werner Frey & Claudia Maienborn (eds.), Proceedings of the Dislocated ElementsWorkshop, ZAS Berlin, November 2003. (ZAS Papers in Linguistics 35), 543–566. Berlin: ZAS. Zimmermann, Malte. 2004b. Zum Wohl. Diskurspartikeln als Satztypmodifikatoren. Linguistische Berichte 199. 253–286. Zimmermann, Malte. 2013. Ob-VL-Interrogativsätze. In Jörg Meibauer, Markus Steinbach & Hans Altmann (eds.), Satztypen des Deutschen, 84–104. Berlin: de Gruyter.

Yvonne Viesel

Discourse particles “embedded”: German ja in adjectival phrases The article is concerned with the German discourse particle ja (lit. ‘yes’) in adjectival phrases. Data shows that ja may appear embedded even if its host construction does not comprise an independent illocutionary force domain, e.g. in restrictive adnominal modifiers in indefinite DPs with specific interpretation. As in main clauses, ja signals that information in its scope supplements related information, thus being indirectly Relevant to the Question under Discussion. Corresponding to this discourse function, corpus evidence suggests that the particle requires a clear indication of focus in its scope even in non-restrictive modifiers which are associated with root properties.

1 Introduction With respect to its approximately 20 discourse particles (DiPs)1 (cf. ­Zimmermann 2011: 2015), derived from homonymous counterparts in different word classes – e.g. adverbs and focus particles (FoPs), German displays some peculiar ­properties. DiPs are a root phenomenon and interact with the system of illocutionary force, as reflected in their clause-type sensitivity, cf. (1).2 Accordingly DiPs occur sentence-peripherally instead of sentence-medially in most other languages (cf. Abraham 2017).

1 Discourse, or modal particles, should not be confused with discourse markers or linkers (cf. Abraham 2017). Underscoring their discourse-structuring function while avoiding confusion regarding the notion of modality, the former term is used here (cf. the discussion in Grosz [2016a]). 2 The examples in this paper are formatted uniformly. For the sake of readability, glosses are not provided for larger contexts. The following abbreviations are used: DAT = dative; GEN = genitive; SBJV = subjunctive mood. Acknowledgements: For invaluable advice and discussion, I thank Josef Bayer, Maribel Romero, María Biezma, Ellen Brandner, Nicole Dehé, Giuliana Giusti, Tatjana Scheffler, Angelika Kratzer, and two anonymous reviewers. Errors are, of course, my own. Yvonne Viesel: Fachbereich Sprachwissenschaft, Universität Konstanz, Konstanz, e-mail: [email protected] DOI 10.1515/9783110497151-008

174 

 Yvonne Viesel

(1) a. Wann ist (denn / * ja) das Paket bei Dir when is DENN JA the parcel at you.dat angekommen? arrived ‘When did you get the parcel?’ b. Der Absender hat esi / das Paketi ( ja / * denn) letzte the sender has it the parcel ja denn last Woche ti aufgegeben. week posted ‘The sender posted it/the parcel last week.’ DiPs convey information about the interlocutors’ mental states and “in most cases [operate] on semantic categories (sentence-type operators, speech-act operators) that are structurally and semantically associated with the periphery of the clause (Rizzi 1997)” (Zimmermann 2011: 2034). Unsurprisingly, they are mainly encountered in matrix clauses, but also in peripheral as opposed to central adverbial clauses and in appositive relative clauses (RCs), i.e. in such clauses as have been claimed to constitute separate illocutionary force domains (cf. Hinterhölzl and Krifka 2013; Coniglio 2008; Haegeman 2002). Contrary to standard assumption though, they appear in propositional complements of factive verbs3 (cf. Hinterhölzl and Krifka 2013; Kwon 2005), as simple corpus searches for dass ja ‘that JA’ show (2), as well as in prenominal adjectival and participial modifiers (3): (2)  man sollte auch berücksichtigen, dass ja zugleich one should also consider that JA simultaneously die Löhne steigen the wages rise ‘it should also be taken into account that wages rise at the same time’ (Die Zeit, 06/21/2012)4 (3) diese ja nicht gerade einfache Geschichte this JA not exactly simple story ‘this not exactly simple story’ (Die Zeit, 10/14/1983)

3 The presupposition in (2) that it is true that the wages rise is triggered by berücksichtigen ‘take into account’. 4 Data from the newspaper Die Zeit can easily be accessed by searching for word strings online. These and most of the many internet findings discussed in the following will hopefully stay available to the interested reader for a long time to reveal their original contexts.



Discourse particles “embedded”: German ja in adjectival phrases  

 175

In structures such as (2) and (3), the particle takes local scope over a propositional part expressed within the embedded CP or DP. That is, by adding ja (lit. ‘yes’) in its function as a discourse particle, the speaker in (2) signals that it is uncontroversial that the wages rise, but not that it is uncontroversial that this should be taken into account. Similarly in (3), the story being ‘not exactly simple’ is uncontroversial, but is made salient nevertheless because it is significant to what is actually being discussed, often in the at-issue dimension of the utterance. The following discussion focuses on the DiP ja as one of the most frequent DiPs in German (Thurmair 1989: 208) and is mainly based on a corpus search for ja in DPs with adjectival modification, conducted in four corpora of the DWDS (Digitales Wörterbuch der deutschen Sprache; cf. Klein and Geyken 2010). After a brief introduction to the syntax and semantics of DiPs at the level of CP and of the DiP ja in particular (Section 2), a more detailed outline of the properties and problems associated with constructions as in (3) is given in Section 4. The empirical picture is more complicated than previously recognized (Section 5), suggesting that different factors, especially information structure, play a role for the grammatical or felicitous employment of ja in adjectival phrases (APs). The article closes with a summary of the results.

2 The DiP ja at the level of CP This section provides basic syntactic and semantic assumptions on CPs with DiPs in general before focusing on the DiP ja and its discourse function. For reasons to be developed in Section 3, the analyses for DiPs at the level of CP cannot straightaway be expanded to syntactically embedded contexts whereas the meaning and function of the DiPs prove to be identical in different environments.

2.1 German DiPs in general German DiPs are most likely functional heads (Bayer and Obenauer 2011; Struckmeier 2014) and are immobile in a pre-VP/vP position, where they may be preceded by information-structurally relevant elements, such as scrambled phrases (1b) or frame setting adverbials (cf. Krifka 2008): (4) Er hat gestern ja die Tür gestrichen. he has yesterday JA the door painted ‘He painted the door yesterday.’

176 

 Yvonne Viesel

DiPs are sensitive to clause type but basegenerated in their scope position above the lexical layer of the clause, its propositional core. Different suggestions have been made concerning the connection between DiPs and the C-system of the clause, in which clause-type and speaker attitude are encoded. Zimmermann (2009) accounts in terms of LF-raising for the effect that the particle wohl (roughly ‘I suppose’) has on the illocutionary force of an utterance. Moving covertly to ForceP, wohl modifies the strength of speaker commitment and determines the epistemic reference point of an assertion (Zimmermann 2009: 20). But such an approach cannot be applied to cases such as (2), in which the particle takes local scope within an embedded clause. If DiPs were LF-raised to the locus of speaker attitude in the matrix CP, global scope would be the only option available, and (2) would be falsely predicted to have the same interpretation as (5), where it is uncontroversial that it should be taken into account that the wages rise: (5) man sollte ja auch berücksichtigen, dass zugleich die Löhne steigen ‘it [JA] should also be taken into account that wages rise at the same time’ Likewise, the difference between the two versions of the rhetorical question in (6) cannot be captured by an LF-raising account: (6)

Wo glaubst du (schon), dass man hier nachts where believe you SCHON that one here at-night um 3 Uhr (schon) Benzin bekommt? at 3 o’clock SCHON gasoline gets ‘Where do you believe that one can get gasoline here at 3 o’clock in the night?’ (Bayer, Häussler, and Bader 2016: 623)

Informally, in (6) the speaker rejects the proposition p expressed in the embedded clause when the latter contains the DiP, whereas schon in the matrix clause indicates that the speaker dismisses the addressee’s believes about p. Rejecting LF-movement, Bayer and Obenauer (2011) advocate a featuresharing approach, in line with current minimalist research. They adopt a model by Pesetsky and Torrego (2007), who unlike Chomsky (2001) allow for probes with interpretable features to agree with goals bearing the corresponding uninterpretable feature. The mechanism as applied to DiPs is illustrated in (7) for the interrogative particle denn in a version of (1a) with the direct object having undergone scrambling: (7) a. [FinP/ForceP Wh Force0/Fin0iQ, iQForce [ ] [(TopP) ... [PrtP Prt0uQForce [ ] [(AdvP*) [VP/vP ... ]]]]] (Bayer and Obenauer 2011: 464)



Discourse particles “embedded”: German ja in adjectival phrases  



b. [FinP/ForceP Wannj istiQ, iQForce [3] [(TopP) das Paketi [PrtP dennuQForce [3] [tj [VP ti bei Dir angekommen]]]]]

 177

The feature-sharing account attributes the distribution of DiPs to features in DiPs that correspond to specific clause types. Assuming a Force domain split into a speaker-attitude and a (lower) clause-typing projection (cf. Haegeman 2010; Haegeman and Hill 2013), feature agreement is also able to explain how the expressive meaning components of the DiPs, revealing what attitude speaker or hearer hold towards a proposition according to the speaker, can be accessed from the left periphery. The reason why the probe has to bear the interpretable feature in Bayer and Obenauer’s (2011) analysis is that the illocutionary Force of an utterance is primary to the DiP.5 The same holds for the pragmatic focus-background partition of a sentence, cf. Egg and Mursell (2017) for a feature-sharing mechanism that includes the focused constituent. At least some DiPs, just like FoPs, seem to be conventionally focus-sensitive in the sense of Beaver and Clark (2008) (cf. Grosz 2016b), i.e. they associate with the part of a sentence that provides an augmentation of the context. Since the background of a proposition contains presupposed content, an instruction (the discourse function of even a truth-conditionally nonrelevant DiP) on how to relate information to the Common Ground6 (CG) can only affect what is focused.

2.2 The discourse function of ja In accordance with ja as a response particle, ja in its function as a DiP at the level of CP is not licit in interrogatives (1a). In declaratives, ja signals that a proposition p is uncontroversial, or, put differently, that p is part of the CG (cf. Jacobs 1991; Kaufmann 2010; Hinterhölzl and Krifka 2013, among others). This “semantic core” function (Zimmermann 2011: 2012) amounts to a “relatively unspecific meaning”7. The question whether p has to be discourse-old indeed, or at least convey information that is evident or part of the shared knowledge of speaker and

5 It is not independent under the assumption that DiPs co-determine the illocutionary Force of an utterance, e.g. by facilitating a rhetoric reading for cases such as (6), but the right illocutionary type is a prerequisite for DiPs. 6 Cf. Roberts (2012) for the formal discourse model, structured around the “Question under Discussion”, that is assumed in the following, and for a formal definition of the CG. 7 Thurmair (1989: 208): “relativ unspezifische Bedeutung”.

178 

 Yvonne Viesel

hearer, has been subject to debate (cf., e.g., Thurmair 1989; Kratzer 1999, 2004; Karagjosova 2004) (cf. Krifka [2013] for the response particle ja as an anaphoric element). It is safe to say that information under ja is often discourse-old, and repeating an utterance without the particle labeling it as a reminder would sound redundant. But it is possible to mark new information as uncontroversial with ja to prompt the hearer to put p to the CG without further debate, to take it as significant for another utterance that is actually Relevant to the Question under Discussion (QuD) in the discourse – “Relevant” in the sense of corresponding to the QuD (cf. Roberts 2012; Hinterhölzl and Krifka 2013). Intuitively speaking, ja is therefore used to (re‑)establish CG. Hinterhölzl and Krifka (2013) define subordinate clauses able to host DiPs as “background assertions”, insisting that this information is asserted and does not need to be accommodated, but serves as background for the evaluation of the matrix construction, unlike presuppositions when they are not fulfilled in the CG. Similarly, what is special about ja-declaratives is not so much whether the information is known but rather that it is background information at the discourse level. As such, an utterance with ja can supplement following arguments or previous ones, as in (8) (cf. Brausse 1986), help to identify whatever else is really at issue (9)/(10), or function as a discourse-opener (11)/(12), but it is not a complete discourse move by itself (12): (8)

Wir gehen dieses Jahr wieder in den Musikantenstadl. Wir stehen ja nicht nur auf Jazz. we stand JA not only on jazz ‘We go to [a folk music show] again this year. We’re not only into jazz.’

(9)

Das nötige Geld hätte ich ja, the needed money had.sbjv I JA aber mir fehlt die Zeit. ‘I have the money needed but I lack the time.’ (König and Requardt 1991: 65)

(10) Anna: ‘I have not heard of lonesome Cedrick in ten years. What about you?’ Bela: Also Cedrick, Cedrick hat ja vor fünf Jahren well C. C. has JA before five years Didi geheiratet, D. married und jetzt haben sie schon zwei Kinder und wohnen in einer Kommune. ‘Well Cedrick, five years ago (you know/see,) Cedrick married Didi and now they already have two children and live in a commune.’ (Viesel 2015: 420)



Discourse particles “embedded”: German ja in adjectival phrases  

 179

(11) Du hast ja ’n Loch im Ärmel. you have JA  a hole in.the sleeve ‘You’ve got a hole in your sleeve!’ (Kratzer 1999: 1) (12) At the bus stop, Anna meets a new neighbor, Bela, who is holding by the leash what is obviously his dog. a. Conversation A: Bela: Das ist mein Hund. ‘This is my dog.’ Anna: Süß. ‘Cute.’ b. Conversation B: Bela: Das ist ja mein Hund. Anna: ...? Bela: i. #  Ø ii. ü Braucht er eine Busfahrkarte? ‘Does it need a bus ticket?’ (Viesel 2015: 420) A proposition conveying uncontroversial, given information would pragmatically make little sense unless it is important for the evaluation of other propositions. In (12a), Anna can take Bela’s utterance as Relevant in itself and react to the introduction of the dog by making the next discourse move. Employing ja in (12b), however, Bela signals that the dog being his is a fact of some import, but, crucially, not his main point. Anna will expect Bela to say or ask something based on his initial statement, and only if he does (ii), the utterance, or actually his use of ja in the utterance, proves felicitous. As (12) shows, information being already known, or obviously uncontroversial, is not sufficient for the felicitous employment of ja. The speaker in (10) even uses ja with brand-new information in a clause that might be understood as the speaker’s main point if it did not contain ja. Using ja, Anna is less likely to interrupt, knowing that Bela wants to say more than that Cedrick married Didi five years ago. A similar case could be (9). Interestingly, e.g. to convince hearers in a debate, speakers can use ja strategically, but may, by doing so, also flout the Gricean Maxims of Relevance or Quality. Speakers might try to sneak pieces of information onto the CG that are more controversial than they purport to be. On the other hand, critical hearers can cast doubt on the main argument by calling background information into question (13a). The data in (13) are from the conciliation talks of the violently disputed Stuttgart 21 railroad station project. In (13b), the second speaker enquires

180 

 Yvonne Viesel

precisely about the incomprehensible information in the ja-parenthetical, which the first speaker might not have wished to discuss in depth.8 (13) a. Michael Holzhey: denn das heißt ja, wenn das X maximal eine 9 for that means JA if the X maximally a 9 wäre – dass Sie 9 Prozent Tape(?)fehlerdifferenz haben [...] were ‘for that means, if X were maximally a 9, that you have 9 percent tape error difference’ Dr. Heiner Geißler: das ist eine Pressemeldung? [...] Ist die Meldung richtig? ‘That is a press report? Is the report correct?’ b. Michael Holzhey: Jaja – am Tag – wir reden über den Tag; um auf einen Stundentakt zu kommen, wohlbemerkt – das ist ja immer noch nicht eine Achsenfunktion that is JA always still not an axis.function – sind wir noch weit davon entfernt. ‘Yes yes – per day – we are talking about a day; to reach hourly intervals, please note – that is still not an axis function yet – we are still far from it.’ Dr. Heiner Geißler: Das war gerade wieder: „Achsenfunktion“ – was ist denn das? ‘That just was again: “axis function” – what is that?’ (http://stuttgart21.wikiwam.de/Wortprotokoll_der_Schlichtung_04.11.2010, 08/29/2014)

8 It cannot be judged from (13) if the speaker using ja was insincere, intending to sneak false or useless information to the CG, or if the clarity of his argument just suffered from the complexity of the matter. Interestingly, the verbatim minutes of the Stuttgart 21 mediation talks are currently evaluated with regard to argumentation and persuasion strategies (http://www.visargue.unikonstanz.de/index.php?lang=en, 03/18/2015).



Discourse particles “embedded”: German ja in adjectival phrases  

 181

3 “Embedded” ja and independent Force domains “Embedded” ja may appear in interrogatives. In (14), the addressee’s ideas are evaluated as (uncontroversially) ‘pretty neat’, which is the background for what is at-issue and controversial in the sentence. The QuD is how the hearer would ‘get around the enormous power of the market’ with his ideas, and the contrast between the neat ideas and the enormous power is plausibly expressed intonationally. The felicity/relevance condition outlined in Section 2.2 are clearly fulfilled. (14) Wie kommen Sie mit Ihren ja ganz hübschen Ideen your JA quite neat ideas an dieser gewaltigen Kraft des Marktes vorbei? ‘How do you get around this enormous power of the market with your pretty neat ideas?’ (Die Zeit, 10/21/2004) Structurally, the AP containing ja is presumably clausal, constituting a local Force domain which the particle is dependent upon, cf. Coniglio (2008), Hinterhölzl and Krifka (2013) (with reference to Haegeman [2002] on Force in peripheral as opposed to central adverbial clauses and the resulting consequences for particle licensing). Neither a feature-sharing nor an LF-raising (cf. Zimmermann 2004) approach are designed to capture the local scope of the DiP, even under the assumption of a CP split into a speaker-attitude and a (lower) clause-typing projection (cf. Haegeman 2010), considering the grammatical appearance of ja in questions. Sections 3.1 and 3.2 address the issues of sentential APs and participle constructions, and syntactically embedded yet semantically independent Force domains.

3.1 Prenominal modification and clause structure Treating prenominal adjectival and participial constructions alike, Struckmeier (2010: 673) analyzes “complex attributive structures” as sentential (like e.g. Fanselow 1986; Toman 1986; Kayne 1994). In (15a) and (15b), ja is indeed embedded within the attributive phrase and does not depend on adjacency to the determiner of the containing DP. The particle may be preceded by scrambled constituents or adverbial elements as at the level of CP.

182 

 Yvonne Viesel

(15) a. Will nicht die Hochschulpolitik einmal mehr auf dem Holzweg enden, ist eine rasche Umsetzung der [AP1 skizzierten] und sketched and [AP2 von vielen ja weitgehend akzeptierten] by many JA far.going accepted Reformvorschläge erforderlich. ‘If higher-education policy does not want to find itself once more barked up the wrong tree, a quick realization is necessary of the reform proposals sketched and largely accepted by many.’ (Die Zeit, 02/21/1997) b. Hier muß ich mich begnügen, auf die [AP1 von mir konsultierte], the by me consulted [AP2 im übrigen ja sehr bekannte] Literatur  in.the rest JA very known literature zu verweisen ‘Here I have to content myself with referring to the literature consulted by me, which, besides, is very well-known’ (Ernst Troeltsch, Die Soziallehren der christlichen Kirchen und Gruppen, p. 248, Tübingen, 1912) Omitting the details, such as the precise locus of Force, Struckmeier (2010: 685) assumes the same “functional layers above the lexical level of argument projection” in prenominal attributive phrases as at the level of CP. The internal structure of the (extended) AP2 in (15a) is then roughly as in (16), with the DiP occupying a dedicated position above the propositional core of the construction: (16) [TopP von vieleni [PrtP ja [TP [AP (lexical layer) ti [weitgehend [A akzeptiert]]] [ T0 ]]]]9 Mainly concerned with complex APs, Struckmeier arrives at the conclusion that, in minimalist terms, prenominal modifiers are separate phases, which are embedded in DPs as subordinate clauses are embedded in CPs, with the AP suffix as the phasal head equivalent to the complementizer. Neither adnominal modifiers nor subordinate clauses refer autonomously, but “specify a matrix

9 In participle constructions, T0 contains the present or past participle suffix overtly, while in APs “these minimal T heads lack tense, aspect and φ-features, and consist of the (universal) EPP feature and the information structure feature set that is needed to derive the scrambled orders” (Struckmeier 2010: 685). Whether the label “TP” is appropriate in the absence of tense features is of no concern for the present purpose.



Discourse particles “embedded”: German ja in adjectival phrases  

 183

structure” (Struckmeier 2010: 689). By contrast, Giusti (2014), exploring simple APs, concludes that “APs are not phases” and do not make “reference to properties” but only attribute properties to objects. According to Giuliana Giusti (p.c.), APs are generally non-propositional in the absence of further structure but complex APs might be (reduced) predications of a property onto a noun. As the data show (cf. Section 4.1 below), ja is not often found in simple APs, i.e. with just a bare, positive, non-negated, lexically non-complex10 adjective without any arguments or modifiers such as degree elements. If (mostly syntactically overt) structural complexity as such is a defining factor of APs with ja, the distribution of ja could be due to the availability of two distinct, more and less extended adjectival projections (cf. Section 4), possibly comparable to finite and non-finite clauses.

3.2 Local Force domains According to Jacobs (1986: 108), an “NP node can absorb the semantic scope of the particle if and only if the latter is part of an appositive attribute’ (17a), which, ‘contrary to a restrictive attribute [17b], expresses its own assertive illocution” [translation Y.V.):11 (17) a. Diese ja schon oft erhobenen Vorwürfe hat er  these JA already often raised reproaches has he wiederholt. repeated ‘He repeated these reproaches, which had been made frequently already.’ b. *Ja schon oft erhobene Vorwürfe hat er wiederholt. ‘He repeated reproaches which had been made frequently already.’ (Jacobs 1986: 107)

10 Adjectives such as wunderschön (lit. ‘wonderbeautiful’) are taken to be lexically complex as they consist of a simple adjective, here schön ‘beautiful’, and a specification to (usually the degree of) this property. 11 “[17] schließlich zeigt, daß der semantische Bereich der Partikel genau dann durch eine NP-Knoten absorbiert werden kann, wenn sie Bestandteil der Formulierung eines appositiven Attributs ist. Die Erklärung dafür ist, daß appositive im Gegensatz zu restriktiven Attributen eigene assertive IIlokutionen zum Ausdruck bringen” (Jacobs 1986: 108).

184 

 Yvonne Viesel

Jacob’s (1986) observation is confirmed by evidence from the related domain of RCs. According to standard assumption, DiPs occur in appositive, i.e. nonrestrictive, RCs (18a), but not in restrictive ones (18b), (19), due to the lack of an independent Force domain in the latter (Coniglio 2011; Hinterhölzl and Krifka 2013): (18) a. Tracy, die ja in Syracuse wohnt, wird kommen. Tracy who JA in Syracuse lives will come ‘Tracy, who lives in Syracuse, will come.’ b. Eine Kollegin, die (*ja) in Syracuse wohnt, wird kommen. ‘A colleague who lives in Syracuse will come.’ (Kratzer 1999: 5) (19) *Die Firma sucht einen Angestellten, der ja immer the firm looks-for an employee who JA always pünktlich ist. punctual is ‘The firm is looking for an employee who is always on time.’ (Zimmermann 2004: 32) Potts classifies appositive RCs as conventional implicatures (CIs). In his analysis, “the same semantic force as a main clause assertion” is assigned to “all CI-content” (Potts 2005: 68), though, including non-clausal structures (cf. McCready [2010] on types of CI, even lexemes of mixed, CI and at-issue, type). Regarding the scope of non-restrictive expressive adjectives, he observes a “deviation from the expected isomorphism between the syntax and the semantics” (Potts 2005: 18) in that such an adjective does not take the noun as its argument. It rather applies to an “entity level term” that is not necessarily “a surface structure constituent”. In Potts’ terms, the nominal-internal friggin’ in (20) takes a full clause as its argument, as “the speaker disapproves of the fact that his bike tire is flat again”: (20) My friggin’ bike tire is flat again! (Potts 2005: 18) In sum, DiPs in RCs and semantic research on the notion of CI show that assertive Force cannot be equated to a certain clause type. The non-restrictive APs with which the present paper is concerned, however, invariably specify the referent of the containing DP, unlike expressives as in (20), according to Potts’ analysis. Their structural properties as apparently required by the DiP ja are discussed in greater detail in the next sections, where it will be shown that ja can occur in restrictive modifiers of indefinite DPs with a specific reading.



Discourse particles “embedded”: German ja in adjectival phrases  

 185

4 Empirical observations Each of the following sections presents an unexpected piece of empirical evidence followed by a discussion of the questions raised. Section 4.1 deals with the semantically unexpected “complexity requirement” of prenominal modifiers. Since a comprehensive analysis of all aspects pertaining to this puzzling matter is impossible at this point, my necessarily speculative explanatory account is postponed until Section 5, where more data and information on related phenomena are presented. Section 4.2 deals with ja in restrictive modifiers, encountered in indefinite DPs with specific reference contrary to all expectations. In the absence of appositive material, the discussion focuses on an adequate distinction between the at-issue and implicated meaning components, again delaying more structural considerations until later.

4.1 Complexity in prenominal modifiers under ja Corpus data suggest that bare adjectives under ja are strongly dispreferred.12 Only four of 120 instances of the corpus search (3.3%) feature DPs with a single, simple AP consisting of ja and one non-complex adjective (negated neither by nicht ‘not’ nor the negational prefix un-, no further adverbial or degree modifiers). Among ten instances (8.3%) featuring the adjective bekannt ‘known’ non-restrictively, there is only one simple AP of the form ja bekannt. Rare cases like (21a) improve greatly with heavy stress on the adjective. (21) a. der ja lange Schritt zur Hürde the JA long step to.the hurdle (Ralph Hoke; Otto Schmith, Grundlagen und Methodik der Leichtathletik, p. 70, Leipzig, 1937)

12 This claim applies to the DiP ja, not to ja in a use roughly meaning ‘even’, which can be used to conjoin virtually any categorically identical material ordered climactically (cf. Kwon 2005: 31), thus also adjectives: (i) a. ein außergewöhnliches, ja singuläres Zwischenspiel ‘an extraordinary, even unique interlude’ (Die Zeit, 07/17/2014) b. ein Rätsel, ja fast ein Unfug ‘a riddle, even almost a nonsense’ (Die Zeit, 07/24/2014)

186 

 Yvonne Viesel

b. mein ja brand-neues / *(nicht / leider) neues Auto my JA brand-new not unfortunately) new car c. die ja *(sehr) nötige Kritik the JA   very necessary criticism (Jahresberichte für Deutsche Geschichte, 6. Jahrgang 1930–1932, p. 162) d. der ja wohl sichere Erfolg the JA WOHL certain success (Berliner Tageblatt, 03/04/1904) e. in einer ja nicht ganz unheiklen Angelegenheit in a JA not entirely non-delicate affair (Die Zeit, 03/13/1964) Generally, APs under ja display varying degrees of greater complexity, featuring a wide range of material such as (stacked) degree and different adverbial modifiers, adjectival arguments, (lower) DiPs, and (double) negation (21c)–(21e). Out of the blue, constructions like (21a)–(21c) in their simplest forms are judged more or less deviant. Utterances like (22a), but not (22b), have been rejected by native speakers when presented in isolation: (22) a. *Gestern habe ich meinen ja kranken Nachbarn getroffen. b. Gestern habe ich meinen ja schwerkranken Nachbarn yesterday have I my JA seriously-sick neighbor getroffen. met ‘Yesterday I met my (seriously) sick neighbor.’ (adapted from Trotzke and Viesel 2014: 1) On the semantic side, if an AP is non-restrictive, it is assumed to project an assertive, propositional CI compatible with the use of ja. Differences in acceptability would not be expected on grounds of the co-occurrence of, e.g., grading particles. Any DiP just scopes over the conventionally implicated proposition p: (23) a. CI (p): The step to the hurdle is long (ja(p), cf. [21a]) b. CI (p): The success is certain (ja(p), wohl(p), cf. [21d]) (23b) is an analysis where the meaning of the two particles applies simultaneously, as suggested by Lindner (1991) for ja and doch (cf. Müller 2017). An alternative, compositional account for ja and wohl by Zimmermann (2011: 2030) is adapted as in (24): (24) ja(wohl(The success is certain)) = Speaker assumes the weakened proposition that the success is relatively certain to be uncontroversial.



Discourse particles “embedded”: German ja in adjectival phrases  

 187

In combination with other DiPs, ja strictly occupies the highest position. But even if the surface order of the DiPs is merely reflective of “their attachment at different syntactic and semantic levels” (Zimmermann 2011: 2013) and without any impact on the compositional derivation of the utterance meaning, (21d) still patterns with the complex cases syntactically and, arguably, prosodically. At any rate, if structural complexity as such plays a role for the occurrence of ja, any analysis must take into account that ja is in principle available in seemingly simple APs as in (21a), albeit not as readily as in complex ones. This could mean that two structures of different complexity are available for superficially equally simple instances, cf. Section 3.1, depending on pragmatic, lexical, or, again, prosodic circumstances. It might be possible to discern a crucial projection or feature present in all complex, but only some non-complex APs. To summarize, it has not been studied yet what complexity does for licensing ja in APs, but it seems as if, contrary to semantic expectations, not all non-restrictive APs can contain this DiP, the problematic cases featuring bare adjectives under ja. This issue cannot be resolved here, but some observations on major influences on the target constructions are offered in Section 4, after addressing the second empirical issue, ja in restrictive modifiers.

4.2 Indefinite DPs and specificity With respect to the licensing of DiPs, APs and participial constructions pattern similarly to RCs (as shown in Section 3.2), depending on their status as restrictive or non-restrictive modifiers. This dichotomy is too crude to capture all relevant facts. First, the particular semantics of DiPs differ. For instance, wohl, simply expressing some uncertainty on the part of the speaker, is licit in the restrictive RC in (18b), contrary to ja: (25) Eine Kollegin, die wohl in Syracuse wohnt, wird kommen. Furthermore, cases like (19) improve when the DP has a unique referent in the discourse. Under the most natural reading of (26), the discourse revolves around an average-sized firm and the speaker hardly intends to merely assert that with Mr. K., the firm gets an employee. (26) Mit Herrn K. bekommt die Firma einen Angestellten, With Mr. K. gets the firm an employee der ja immer pünktlich ist. who JA always punctual is ‘With Mr. K., the firm gets an employee who is always punctual.’

188 

 Yvonne Viesel

Similar observations are made in the adjectival domain: (27) Und wer die Bedingungen studiert, unter denen sie arbeiten, der staunt nicht minder: ob der institutionellen Unterstützung für eine ja nicht immer einfache Musik. a JA not always easy music ‘And those who study the conditions under which they [jazz musicians in Switzerland] work are no less astonished: at the institutional support for a music which is not always easy.’ (Die Zeit, 01/27/2006) If the AP in (27) were non-restrictive, the speaker would assert (at-issue) that the institutional support for any music was astonishing, and implicate conventionally via the appositive modifier that (any) music was not always easy. This is the reading for some non-restrictive APs in indefinite DPs when the property denoted is inherent to each member of the set denoted by the head noun. Since a chance victory is always possible, the DP in (28a) is an example for “conceptual nonrestrictiveness” (Fabricius-Hansen 2009: 90) in this sense. It may be left out while leaving the at-issue content of the utterance intact (28b): (28) a. [(irgend)] ein ja immer möglicher Zufallssieg some a JA always possible chance.victory ‘some chance victory, which is always possible’ b. ein [...] Zufallssieg ohne Vorbereitung hierzu verdient nicht die Ehre, die allein dem wahren Kämpfer zukommt ‘a chance victory without preparation does not deserve the honor which befits the true fighter alone’ (Ralph Hoke; Otto Schmith, Grundlagen und Methodik der Leichtathletik, p. 177, Leipzig, 1937) Such a reading is neither plausible for (27), nor for (26). Just like the RC in (26), the AP in (27) does not convey mere additional information on the entity denoted by the head noun, but information without which the DP would be quite uninformative, i.e. the AP cannot be omitted without destroying the intended DP reference. This is even clearer in (21e), repeated in (29a), from an article on the sensitive issue of income control to ensure social housing for eligible beneficiaries. Without AP, the parenthetical is essentially devoid of all meaning (29b): (29) a. Dieser erstaunliche Sinneswandel in einem Zeitraum von nur wenigen Jahren – und in einer ja nicht ganz unheiklen Angelegenheit – ist das sozialpolitisch Bemerkenswerte an dem Gesetzentwurf



Discourse particles “embedded”: German ja in adjectival phrases  

 189

‘This amazing change of mind in a period of only a few years – and in an affair that is not entirely non-delicate – is what is sociopolitically remarkable in the bill’ (Die Zeit, 03/13/1964) b. ‘This amazing change of mind in a period of only a few years # – and in an affair – is what is sociopolitically remarkable in the bill’ The modifiers in (27) and (29) are conceptually restrictive. Intersecting the denotations of AP and NP yields the sets of music which is not always easy (27), and of not entirely non-delicate affairs (29), which are smaller than the sets denoted by the NPs alone. That is not the case in (28). What distinguishes (26), (27) and (29) from the cases in which ja is illicit in restrictive modifiers (17b)/(18b)/(19) is that there is one salient affair, employee or music in the discourse that fits the description. On first sight, the DPs in question seem to be epithetical along the lines of Potts (2005: 19): anaphoric elements which are “syntactically integrated (often as argument nominals)” and come “with some added punch”, the lexical content that is conventionally implicated. Potts’ example, involving a more local antecedent, illustrates the pronominal character of such expressions: (30) Every Democrat advocating [a proposal for reform]1 says [the stupid thing]1 is worthwhile. (Potts 2005: 19) In Donnellan’s (1966) terms, definite descriptions can be referential and serve “to get the addressee to realize which entity is being spoken about”, or attributive and used to say “something about whoever or whatever fits the description” (Abbott 2004: 128). With respect to the cases at hand, the DPs with restrictive modifiers allowing for ja would be referential, or epithetical, whereas the DPs that are attributive cannot contain ja as there is no conventionally implicated content (for the parallels between the specific-non-specific distinction in indefinite DPs and the referential-attributive distinction in definite DPs, cf. Abbott [2004: 146]). However, analyzing the indefinite DPs in (26), (27) and (29) accordingly does not yield the right results. (27), for instance, would basically receive the same analysis as if the pertinent DP were definite and the AP non-restrictive (31): (31) Und wer die Bedingungen studiert, unter denen sie arbeiten, der staunt nicht minder: ob der institutionellen Unterstützung für diese ja nicht immer einfache Musik. this JA not always easy music

190 

 Yvonne Viesel

As an epithet, the DP might be argued to contribute nothing but reference to the at-issue meaning of the utterance, so that the at-issue and CI-meaning of (21) would be roughly divided as in (32): (32) at-issue (p):  And those who study the conditions under which they work are no less astonished: at the institutional support for this music / jazz. CI (p): Jazz is a music which is not always easy. To assume that the at-issue information is just the same in (27) and (31) would mean to ignore the difference between a demonstrative DP and the indefinite one chosen by the speaker. The difference is subtle in (27), yet the speaker is not astonished at the institutional support for jazz in particular, but for such a difficult kind of music, in general. The same holds for the utterances in (29) and, most clearly, (26), where the modifiers are a vital a part of what is asserted (in (26), that the firm gets an employee of the kind which is always on time). Shifting the intensional meaning of the entire DP to the dimension of CI, leaving behind the referential component only, would in (26) amount to the tautology that with Mr. K., the firm gets Mr. K. That is not the intended reading, and therefore the non-specific descriptive parts of the indefinite descriptions must remain in the at-issue dimension for a de dicto reading, as exemplified for (26) in (33): (33) at-issue (p): With Mr. K., the firm gets an employee who is always punctual. This analysis captures the difference between definite and indefinite descriptions regarding the data: In contrast to (31), where just the reference to Jazz is at-issue and the lexical meaning of both AP and NP are conventionally implicated, (27) is a more general assertion about a kind that is newly introduced in the discourse (de dicto reading). The de re reading, the anaphoric relation to a discourse antecedent, which is lexically expressed in definite descriptions, is not targeted and hence not expressed by use of a definite or demonstrative determiner. The same holds for (26) and (29).

5 Some observations on DiPs and information structure This section sums up the conclusions to be drawn from the previous observations and suggests areas for further research. Section 5.1 provides an overview of the



Discourse particles “embedded”: German ja in adjectival phrases  

 191

issues arising from empirical facts that have hitherto gone unnoticed, to the effect that crucial influences have been underestimated. Section 5.2 deals with information structure and the findings with ja in the adjectival domain. The evidence adds up to the conclusion that speakers might choose to include a “non-at-issue” (Arndt Riester, p.c.), or “secondary focus”, the structural or prosodic prominence of which indicate a separate layer of information.

5.1 The pieces of the puzzle So far, a wide range of data and theoretical observations have been presented that do not seem to be inherently connected. Major orthogonal factors might have far-reaching consequences and change the view on other, apparently unrelated aspects. These issues have been mentioned above and will be reconsidered below: i. the clause-medial position of DiPs ii. the occurrence of DiPs in embedded contexts other than non-restrictive modifiers iii. focus sensitivity iv. the discourse function of ja to label propositions as background information v. the semantic scope of the particle vi. the locus of the local Force domain vii. the tendency of APs with ja to internal complexity viii. reference Regarding (i), notice that DiPs and FoPs always c-command the base position of the constituent in their nuclear scope (cf. Reis 2005). If DiPs are to be analyzed as focus sensitive, then for the cases discussed in Section 2.1, the semantic nuclear scope of DiPs as operators at the level of CP is the focus part of a structured proposition, its background is the restrictor (cf. Partee 1999), and restrictor and nuclear scope together constitute the scope (iii). When only a subpart of VP is focused, presupposed material is preferably scrambled around the DiP: (34) [‘What has daddy gone to buy on a Sunday?’] a. (?) Sonntags kauft1 doch [VP der Papa [Foc  on.Sundays buys DOCH the daddy Brötchen] t1]. rolls

immer frische always fresh

192 

 Yvonne Viesel

b. Sonntags kauft1 [der Papa]2 doch [VP t2 [Foc immer frische Brötchen] t1]. ‘On Sundays, daddy always buys fresh rolls.’ The special relation between operators and their nuclei (on the tendency to “closeness” of FoPs and their nuclei, cf. Reis [2005]; for DiPs and information structure, cf. Grosz [2016b]) needs to be examined with respect to DiPs in embedded contexts, particularly in view of (vii). Concerning (ii), “embedded” ja has been shown to feature in factive clauses (2) and restrictive modifiers (Section 4.2) of indefinite DPs with specific reference, and wohl has been implemented in a restrictive RC even without specific reference (25). Reference (viii) has proven to be an orthogonal factor, featuring in different ways in definite DPs, indefinite DPs with specific reference and restrictive modifiers, and indefinites with non-restrictive modifiers (28). Crucially, the occurrence of ja in embedded contexts other than non-restrictive modifiers requires a closer look at issues (v) and (vi). Accounts that rely solely on non-restrictive modifiers with respect to Force and scope have to exclude DiPs with intensional, “direct modification adjectives” in the sense of Cinque (2005) (cf. Coniglio 2008: 101), cases with a DiP in DPs without adjectival modification,13 and cases like (37) where the nuclear scope of the DiP seems to be the quantifier:14 (35) a. unser – nun ja leider ehemaliger – Pfarrer our now JA regrettably former reverend Karl-Heinz Stoffels Karl-Heinz Stoffels (www.unsichtbare-freunde.de/html/body_25__jahre_forderverein.html, 11/01/2014) b. die (ja nur vermeintliche) Unerreichbarkeit the  JA only putative elusiveless (http://www.lovetalk.de/archiv-allgemeine-themen/40267-i-cant-helpfalling-in-love-4.html, 08/29/2014)

13 The DP in (36) comments on the adequacy of the word Auflösung ‘dissolution’ for a discoursegiven event. It is acceptable in its context and a specimen of spontaneous spoken German by native singer Andreas Spechtl. Luckily, the register is preserved in the literal quotation. 14 Methodological questions cannot be discussed in depth here, but complementary internet searches yield data from written language quite different than in the text types most common to corpora. A blog entry or contribution to a topic discussed in an internet forum is closer to “the dialogical, interactional aspect” of DiPs (an anonymous reviewer) than strictly structured prose or news reports, and the register and style are more informal and spontaneous.



Discourse particles “embedded”: German ja in adjectival phrases  

 193

(36) [die Krise,] diese ja auch fast Auflösung der Band  the crisis this JA AUCH nearly dissolution the.gen band ‘the crisis, this (uncontroversially) near dissolution of the band.’ (http://www.arte.tv/de/ja-panik-sind-die-laessigsten-poptheoretiker-derberliner-indieszene/7825136,CmC=7822214.html, 08/29/2014) (37) ein ja leider nur paar Bilder (wordplay)15 a JA unfortunately only few pictures ‘an (uncontroversially) unfortunately only few pictures’ (http://www.gespensterweb.de/wbb2/print.php?threadid=3031&page=7&si d=dfea43227089cbe3aaeba9e6cd9fc4d4, 08/29/2014) Even without a detailed analysis of the meaning components that are at-issue and not at-issue (the scope of ja in nominal descriptions) in (35) to (37), it is self-evident that “embedded” ja interacts with non-at-issue dimensions of meaning less specifically than expected. To capture ja in factive clauses, though, which convey presupposed information, requires either to drop the affiliation of this DiP with asserted meaning, reducing it to an affiliation with factive (Grosz 2016a) or veridical (Jacobs 2015) environments, or alternatively, to assume that the content of a factive clause can have a discourse status independent from the embedding clause, be asserted, so to speak, pragmatically (cf. Viesel 2015). This issue cannot be resolved here, but note that the content of the factive clause in (2), repeated in (38), can be the target of negation, i.e. a second speaker’s objection that ‘this is wrong’ can easily be taken to refer to the wages rising (cf. Potts [2015] for tests on the meaning-dimensional status of information). (38) man sollte auch berücksichtigen, dass ja zugleich one should also consider that JA simultaneously die Löhne steigen the wages rise ‘it should also be taken into account that wages rise at the same time’ (Die Zeit, 06/21/2012)

15 The author of (37), judging from the context omitted for reasons of space, is a hobby photographer complaining at length about the meagre results of some trip before providing, ‘as promised’, pictures, but ‘(regrettably, of course) only a few’. Cues like animated smileys and the general style suggest that the native speaker chooses the somewhat funny construction humorously, thereby pushing the borders of grammaticality in a telling way.

194 

 Yvonne Viesel

5.2 DiPs and focus If DiPs can be analyzed like FoPs, DiPs are operators on a nuclear scope. The nature of the nuclear scope, however, is uncertain not only in cases of “embedded” DiPs, but also where FoPs associate with elements in the background of a sentence (speaker B in [39a]) or within a larger focus constituent (allfocus in [39b]). (39) a. A: Eva only gave xerox copies to the GRADUATE STUDENTS. B: No, PETER only gave copies to the graduate students. (Partee 1999: 216) b. [F There’s only a month till CHRISTMAS now]. (Vallduví and Zacharski 1994: 9) Both (39a) and (39b) contain the FoP only, which is associated with a nucleus that does not carry nuclear stress and is not identical with the informational sentence focus. In her discussion of “deaccented focus”, Partee (1999) points out work by Bartels (1997) and Krifka (1997) allowing for “acoustic correlates of focus” in second-occurrence focus (Partee 1999: 228–229). Following Krifka (1997), this is the case in “improper second-occurrence focus”, where there is no completely segmentally identical antecedent in the discourse.16 In cases such as (39a), however, complete segmental identity is given, hence the “proper second-occurrence focus” is deaccented for reasons of economy (Partee 1999: 228) or plain lack of “informativeness (or interestingness)” (Vallduví and Zacharski 1994: 13). With respect to ja in embedded contexts, no completely identical discourse antecedents are involved. Propositions under ja convey additional information in light of which to evaluate what is asserted/at-issue. Such background information in the sense of Hinterhölzl and Krifka (2013) is asserted “as an aside” and functions as a supplement in the sense of Potts (2005: 33). Note that in this sense, ‘background information’ is indirectly related to the QuD and sometimes made salient – “evoked” in the sense of Potts (2015) – even though discourse-given. It needs to be distinguished from “backgrounded information” as in presuppositions, which have led Potts (2005) to the formulation of the “antibackgrounding requirement” of CIs. To avoid confusion caused by the concept of “antibackgrounded background information”, embedded ja-constructions might be called “informationally autonomous” in accordance with their independent, supplementary role in a complex sentence. The notion of “informational autonomy” has been introduced independently by Jacobs

16 For this very reason, the nuclear scope of only in (39b), a month, does not seem quite so deaccented as graduate students in (39a).



Discourse particles “embedded”: German ja in adjectival phrases  

 195

(1999), who does not consider DPs with prenominal modifiers, but as the term suggests, some of his findings might prove useful for understanding embedded ja. In Jacobs’ (1999: 57) terms, informational autonomy is central to “the theory of phrasal and compound stress” and has consequences for “the phenomenon of focus ambiguity” and the theories of feature projection and extraction sites. Jacobs makes a distinction between (non)autonomy and (non)stressability. The latter results either from lexical properties of “small words” like nonstressable etwas ‘something’ in (40a), where the head sehen ‘to see’ is stressed despite being informationally nonautonomous, or from “contextual predictability”,17 as in (40b), where the informationally autonomous predicate are rusty can be deaccented, e.g. in answering the question Why does the water from the tap come out brown?. (40) a. [[ etwas]-str [ sehen]-ia] something   see ‘see something’ b. [[The pipes] [are rusty]-str, +ia] (Jacobs 1999: 59) Informational autonomy itself is defined as “a relation between heads and their sister constituents” (Jacobs 1999: 56). If a “head” requires a separate step of semantic processing, it is autonomous, if the head and its sister are “semantically processed at one fell swoop, by referring holistically” to a concept, the head is non-autonomous. This is intuitively clear e.g. in the minimal pair weißblau ‘white-blue’ and dunkelblau ‘dark blue’, with weißblau involving two steps of semantic processing, but dunkelblau only one, referring to one property. For the time being, the status of simple and complex APs and the modified NP as heads need not be of concern. Note that the non-stressable argument etwas in (40a) is intuitively uninteresting and deaccenting of the predictable are rusty in (40b) is very similar to deaccenting as in (39a), which has been claimed to be caused by “uninformativeness”. Without getting into too much detail, a more general and intuitive notion of “informational autonomy” is therefore adopted to describe the status of both modifiers and nominal descriptions with respect to each other and taking into account plausible stress patterns.

17 An anonymous reviewer points out that the predicate of a thetic sentence like (40b) is expected to be deaccented, cf. Rochemont (2013: 55). According with Rochemont, speakers may choose to present a new predicate as given, especially if the hearer can be expected to accommodate it, a view that seems compatible with Jacobs’.

196 

 Yvonne Viesel

Consider the phrase stress scenario in the subject DPs in (41): (41) a. Ein GROßer und ein KLEIner MANN (stehen an der a tall and a small man  stand at the Bushaltestelle.) bus.stop Der GROße (Mann) (geht weg.) the tall  man  walks away. b. Ein GROßer MANN (steht allein an der Bushaltestelle.) a tall man  stands alone at the bus.stop Der große MANN (geht weg.) the tall man  walks away. The first subject DP in (41a) contains two contrastively focused adjectives, and, comparatively less prominent, the head noun, receiving pitch accent by virtue of comprising its own intermediate intonation phrase (on stress assignment in German DPs with adjectival modifiers cf. Roberts [2012: 53–56]; FabriciusHansen [2009]). The second subject DP contains a noun that is deaccented due to discourse-givenness and a restrictive adjective carrying contrastive stress, which correlates with the evocation and exclusion of alternatives in the discourse. Both the adjective and noun in the first subject DP in (41b) receive pitch accents. They are equally informative and the intersection of their denoted sets being the set of large men. Stress on the adjective also seems justified because the AP is non-restrictive and thus a CI, i.e. antibackgrounded. Again, the AP can be dubbed informationally autonomous in an intuitive sense, but it also carries stress, which is consistent with Jacobs’ (1999) definition. The second subject DP in (41b) is puzzling, however. The adjective is nonrestrictive in this context, since only one referent has been introduced contextually and fulfills the uniqueness presupposition of the definite determiner (cf. Heim 1991). All the properties of a CI seem to be there semantically, but of course, it is deaccented because of a completely segmentally identical antecedent, which renders it uninformative. However, for cases of deaccented secondary-focus occurrences as in (39a), it might be argued that a structured proposition has entirely be copied as one “special sort of ‘economy’ anaphor” (cf. Partee 1999: 228). This anaphor is processed exactly as described by Jacobs for informationally non-autonomous heads, “in one fell swoop”. But in the case of the deaccented adjective in (41b), it is not the necessary background of an assertion that is copied, but only the CI in addition to the noun, despite the fact that it could have been left out for all that is at-issue. Should we have encountered a “backgrounded CI” after all? What sense does such a construction make pragmatically, and what is the semantic difference to the non-restrictive AP in the first utterance in (41b)?



Discourse particles “embedded”: German ja in adjectival phrases  

 197

Let us resort back to the extended notion of informational autonomy. If the adjective is unstressed, it should be informationally non-autonomous. “Nonautonomy (or integration) produces tight, wordlike structures that allow features to climb the tree even when they come from a non-head.” (Jacobs 1999: 57) The adjective is copied in union with the noun because the concept of a tall man is perceived as an information unit. This distinction between two types of non-restrictive adjectives is able to explain why bare adjectives under ja are dispreferred. An extra feature, stress, is apparently associated with APs that are understood as CIs and as asserting a property, while an informationally non-autonomous structure can express a property as just one feature of one lexically complex concept. This second option of bare APs corresponds to the syntactic behavior of non-complex APs as described by Giusti (2014): APs are not phases, but mere attributions onto a noun in one DP phase. The simpler structure is predicted to be the default case, and this default case does presumably not license DiPs. As soon as APs are structurally more complex, however, they seem to be informationally autonomous by default. A DP like dieser ja nicht kleine Laden ‘this JA not small store’ sounds better than dieser ja große Laden ‘this JA big store’. The difference is that the complex AP carries pitch accent on nicht or kleine even in the absence of ja, whereas große might be deaccented and could be processed non-autonomously, “in one fell swoop” with the noun. Phonological research might reveal different intonational patterns to be compatible with ja in different contexts.18 The syntactic and semantic details of embedded ja in connection with information structure must be left for further research. An in-depth analysis will also have to clarify how the referent comes to be in the scope of a DiP embedded in an AP. Of course, FoPs may appear lower than where they take scope when associated with their nuclear scope in ad-focus position (cf. Bayer 1996; Taglicht 1984; Partee 1999):

18 Written data, as an anonymous reviewer points out, is of course not a good source for research on intonation. But some contexts are indicative of enhanced prominence, for instance in the case of contrastive focus. Such conditions apparently allow for DiPs in simple APs: (i) N  ur fiel hier nicht den ja kleinen Schneidezähnen, sondern den riesigen Eckzähnen die umgestaltende Rolle zu [...] ‘Only here, the role of transforming did not devolve to the [JA] small incisors, but to the huge canine teeth’  (Wilhelm Branca, Die menschenähnlichen Zähne aus dem Bohnerz der Schwäbischen Alb, p. 119, Stuttgart, 1898)  he interplay of intonation and information structure with respect to embedded DiPs deserves T far more attention.

198 

 Yvonne Viesel

(42) We were (only1) forced to (only2) study (only3) syntax. The first and second positions for only are distinct scope positions and result in two unambiguous readings: only1 yields the interpretation that nothing was mandatory but the study of syntax, only2, that it was mandatory to study nothing but syntax. The third position for only is immediately preceding the focused constituent, leading to scope ambiguity.

6 Conclusion The data presented above demonstrate that DiPs in embedded contexts are affected by influences such as restrictiveness, reference and “informational autonomy”. Especially the role of information structure and its phonological and syntactic reflexes deserves closer attention, both at the level of complex sentences and inside the embedded ja-constructions (cf. Hinterhölzl and Krifka [2013] for the role of focus, marked overtly by a FoP or in the form of contrastive focus, in licensing DiPs in dependent clauses). Embedded ja seems to depend on enhanced prominence more than on a clear-cut scope domain inherently endowed with assertive Force. Speakers may use ja whenever information is supplemented independently from the matrix construction. Crucially, CIs in general are not necessarily clausal constructions either, but the information conveyed is always expressed overtly, i.e. “conventionally”, “‘by virtue of the meaning of’ the words [a speaker] chooses” (Potts 2005: 211). In order for ja to occur e.g. in the nominal domain, a nuclear scope in the form of a heavily stressed bare adjective seems to suffice for the DiP to associate with and take higher scope over the predication onto the noun. Heavy stress and intervening DiPs, FoPs, (double) negation and adverbs are observed to contribute to the licensing of ja with bare adjectives or exceptionally even other categories such as nouns, possibly by co-indicating the independent function of the modified material or by supplying independent meaning lexically. The claims made in this exemplary discussion of the DiP ja in embedded constructions evidently need to be refined in future research. The empirical findings are challenging, and the theoretical arguments provided above are intended to offer fresh perspectives on an intriguing interface issue. It has become clear that the distribution of DiPs interacts with factors like illocutionary force, (non-) restrictiveness, and specificity, but these notions play different roles in different contexts that themselves affect the felicity conditions and required informationstructural properties of utterances. Therefore, the account is particularly devoted to the promotion of an integrative approach, to the great advantage that insights



Discourse particles “embedded”: German ja in adjectival phrases  

 199

from different fields feed into a principled account. At the moment, we can be optimistic that such an account can be given.

References Abbott, Barbara. 2004. Definiteness and indefiniteness. In Laurence Horn & Gregory Ward (eds.), The handbook of pragmatics, 122–149. Oxford: Blackwell. Abraham, Werner. 2017. Discourse marker = discourse particle = thetical = modal particle? A futile comparison. In Josef Bayer & Volker Struckmeier (eds.), Discourse particles: Formal approaches to their syntax and semantics, 241–280. Berlin: de Gruyter. Bartels, Christine. 1997. Acoustic correlates of “second occurrence” focus: Toward an experimental investigation. In Hans Kamp & Barbara Partee (eds.), Context-dependence in the analysis of linguistic meaning: Proceedings of the workshops in Prague and Bad Teinach, 11–30. Stuttgart: Universität Stuttgart. Bayer, Josef. 1996. Directionality and Logical Form: On the scope of focusing particles and wh-in-situ. Dordrecht: Kluwer. Bayer, Josef & Hans-Georg Obenauer. 2011. Discourse particles, clause structure, and question types. The Linguistic Review 28(4). 449–491. Bayer, Josef, Jana Häussler & Markus Bader. 2016. A new diagnostic for cyclic wh-movement: Discourse particles in German questions. Linguistic Inquiry 47(4). 591–629. Beaver, David I. & Brady Z. Clark, 2008. Sense and sensitivity: How focus determines meaning. Malden, MA: Wiley-Blackwell. Brausse, Ursula. 1986. Zum Problem der sogenannten Polyfunktionalität von Modalpartikeln: ‘Ja’ und ‘eben’ als Argumentationssignale. Zeitschrift für Phonetik, Sprachwissenschaft und Kommunikationsforschung 39(2). 206–223. Chomsky, Noam. 2001. Derivation by phase. In Michael Kenstowicz (ed.), Ken Hale: A life in language, 1–52. Cambridge, MA: MIT Press. Cinque, Guglielmo. 2005. The dual source of adjectives and phrasal movement in the Romance DP. Ms., Università Ca’ Foscari Venezia. Coniglio, Marco. 2008. Die Syntax der deutschen Modalpartikeln: Ihre Distribution und Lizensierung in Haupt- und Nebensätzen. Venezia: Università Ca’ Foscari dissertation. Coniglio, Marco. 2011. Die Syntax der deutschen Modalpartikeln: Ihre Distribution und Lizensierung in Haupt- und Nebensätzen. Berlin: Akademie Verlag. Donnellan, Keith. 1966. Reference and definite descriptions. Philosophical Review 75(3). 281–304. Egg, Markus & Johannes Mursell. 2017. The syntax and semantics of discourse particles. In Josef Bayer & Volker Struckmeier (eds.), Discourse particles: Formal approaches to their syntax and semantics, 15–48. Berlin: de Gruyter. Fabricius-Hansen, Catherine. 2009. Überlegungen zur pränominalen Nicht-Restriktivität. Linguistische Berichte, Sonderheft 16. 89–112. Fanselow, Gisbert. 1986. On the sentential nature of prenominal adjectives in German. Folia Linguistica 20(3–4). 341–380. Giusti, Giuliana. 2014. Are APs phases? Paper presented at the University of Konstanz, Department of Linguistics, July 1, 2014.

200 

 Yvonne Viesel

Grosz, Patrick. 2016a. Discourse particles. Ms., University of Tübingen. http://semanticsarchive.net/Archive/GQ4ZDU0N/grosz-SemCom-discourse-particles.pdf (accessed 7 June 2016). Grosz, Patrick. 2016b. Information structure and discourse particles. In Caroline Féry & Shinichiro Ishihara, (eds.), The Oxford handbook of information structure. Oxford: Oxford University Press. Haegeman, Liliane. 2002. Anchoring to speaker, adverbial clauses and the structure of CP. Georgetown University Working Papers in Theoretical Linguistics 2. 117–180. Haegeman, Liliane. 2010. West Flemish verb-based discourse markers and the articulation of the speech act layer. Ms., University of Ghent, http://ling.auf.net/lingbuzz/001027 (accessed 7 June 2016). Haegeman, Liliane & Virginia Hill. 2013. The syntacticization of discourse. In Raffaella Folli, Christina Sevdali & Robert Truswell, (eds.), Syntax and its limits, 370–390. Oxford: Oxford University Press. Heim, Irene. 1991. Artikel und Definitheit. In Arnim von Stechow & Dieter Wunderlich (eds.), Semantik – Ein internationales Handbuch zeitgenössischer Forschung, 487–535. Berlin: Walter de Gruyter. Hinterhölzl, Roland & Manfred Krifka. 2013. Modal particles in adverbial and adnominal clauses. Ms., Università Ca’Foscari Venezia; Humboldt-Universität Berlin. Jacobs, Joachim. 1986. Abtönungsmittel als Illokutionstypmodifikatoren. Groninger Arbeiten zur germanistischen Linguistik 27. 100–111. Jacobs, Joachim. 1991. On the semantics of modal particles. In Werner Abraham (ed.), Discourse particles, 141–162. Amsterdam: Benjamins. Jacobs, Joachim. 1999. Informational autonomy. In Peter Bosch & Rob van der Sandt (eds.), Focus: Linguistic, cognitive, and computational perspectives, 56–81. Cambridge: Cambridge University Press. Jacobs, Joachim. 2015. Deutsche Modalpartikeln als Hauptsatzphänomene. Paper presented at the workshop “Haupt- und Nebensatz an der Syntax/Pragmatik-Schnittstelle”, University of Konstanz, June 4–5, 2015. Karagjosova, Elena. 2004. The meaning and function of German modal particles. Saarbrücken: Saarland University dissertation. Kaufmann, Stefan. 2010. Strong and weak presupposition: German ‘ja’ under quantifiers. Ms., Northwestern University. http://homepages.uconn.edu/~stk12004/Papers/ja.pdf (accessed 7 June 2016). Kayne, Richard. 1994. The antisymmetry of syntax. Cambridge, MA: MIT Press. Klein, Wolfgang & Alexander Geyken. 2010. Das Digitale Wörterbuch der Deutschen Sprache (DWDS). Lexicographica: International Annual for Lexicography 26. 79–96. König, Ekkehard & Susanne Requardt. 1991. A relevance-theoretic approach to the analysis of modal particles in German. Multilingua 10(1–2). 63–77. Kratzer, Angelika. 1999. Beyond ‘oops’ and ‘ouch’: How descriptive and expressive meaning interact. Paper presented at the Cornell Conference on Theories of Context Dependency, Ithaca, NY, March 26, 1999. http://semanticsarchive.net/Archive/WEwNGUyO/ (accessed 7 June 2016). Kratzer, Angelika. 2004. Interpreting focus: presupposed or expressive meanings? A comment on Geurts and van der Sandt. Theoretical Linguistics 30. 123–136. Krifka, Manfred. 1997. Focus and/or context: A second look at second occurrence expressions. In Hans Kamp & Barbara Partee (eds.), Context-dependence in the analysis of linguistic



Discourse particles “embedded”: German ja in adjectival phrases  

 201

meaning: Proceedings of the workshops in Prague and Bad Teinach, 253–275. Stuttgart: Universität Stuttgart. Krifka, Manfred. 2008. Basic notions of information structure. Acta Linguistica Hungarica 55. 243–276. Krifka, Manfred. 2013. Response particles as propositional anaphors. Semantics and Linguistic Theory 23. 1–18. DOI: http://dx.doi.org/10.3765/salt.v23i0.2676 (accessed 7 June 2016). Kwon, Min-Jae. 2005. Modalpartikeln und Satzmodus: Untersuchungen zur Syntax, Semantik und Pragmatik der deutschen Modalpartikeln. München: Ludwig-Maximilians-Universität dissertation. Lindner, Karin. 1991. ‘Wir sind ja doch alte Bekannte’: The use of German ja and doch as modal particles. In Werner Abraham (ed.), Discourse particles, 163–201. Amsterdam: Benjamins. McCready, Eric. 2010. Varieties of conventional implicature. Semantics and Pragmatics 3(8). 1–57. DOI: http://dx.doi.org/10.3765/sp.3.8 (accessed 7 June 2016). Müller, Sonja. 2017. Combining ja and doch: A case of discourse structural iconicity. In Josef Bayer & Volker Struckmeier (eds.), Discourse particles: Formal approaches to their syntax and semantics, 203–240. Berlin: de Gruyter. Partee, Barbara. 1999. Focus, quantification, and semantics-pragmatics issues. In Peter Bosch & Rob van der Sandt (eds.), Focus: Linguistic, cognitive, and computational perspectives, 213–231. Cambridge: Cambridge University Press. Pesetsky, David & Esther Torrego. 2007. The syntax of valuation and the interpretability of features. In Simin Karimi, Vida Samiian & Wendy K. Wilkins (eds.), Phrasal and clausal architecture: Syntactic derivation and interpretation, 262–294. Amsterdam: Benjamins. Potts, Christopher. 2005. The logic of conventional implicatures. Oxford: Oxford University Press. Potts, Christopher. 2015. Presupposition and implicature. In Shalom Lappin & Chris Fox (eds.), The handbook of contemporary semantic theory. Second Edition, 168–202. Oxford: WileyBlackwell. Reis, Marga. 2005. On the syntax of so-called focus particles in German: A reply to Büring and Hartmann 2001. Natural Language and Linguistic Theory 23(2). 459–483. Rizzi, Luigi. 1997. The fine structure of the left periphery. In Liliane Haegeman (ed.), Elements of grammar, 281–337. Dordrecht: Kluwer. Roberts, Craige. 2012. Information structure in discourse: Towards an integrated formal theory of pragmatics. Semantics and Pragmatics 5(6). 1–69. DOI: http://dx.doi.org/10.3765/ sp.5.6 (accessed 7 June 2016). Rochemont, Michael. 2013. Discourse new, F-marking, and normal stress. Lingua 136. 38–62. Struckmeier, Volker. 2010. Attributive constructions, scrambling in the AP, and referential types. Lingua 120(3). 673–692. Struckmeier, Volker. 2014. ‘Ja doch wohl’ C? Modal particles in German as C-related elements. Studia Linguistica 68(1). 16–48. Taglicht, Josef. 1984. Message and emphasis: On focus and scope in English (English Language Series 15). London: Longman. Thurmair, Maria. 1989. Modalpartikeln und ihre Kombinationen. Tübingen: Niemeyer. Toman, Jindřich. 1986. A (word-)syntax for participles. Linguistische Berichte 111. 367–408. Trotzke, Andreas & Yvonne Viesel. 2014. Discourse particles in the German DP. Paper presented at the workshop: Discourse Particles: Cross-linguistic Perspectives, Schloss Freudental, January 24–25, 2014.

202 

 Yvonne Viesel

Vallduví, Enric & Ron Zacharski. 1994. Accenting phenomena, association with focus, and the recursiveness of focus-ground. In Paul Dekker & Martin Stokhof (eds.), Proceedings of the 9th Amsterdam Colloquium, 683–702. Amsterdam: ILLC. Viesel, Yvonne. 2015. Discourse structure and syntactic embedding: The German discourse particle ‘ja’. In Thomas Brochhagen, Floris Roelofsen & Nadine Theiler (eds.), Proceedings of the 20th Amsterdam Colloquium, 418–427. Amsterdam: ILLC. Zimmermann, Malte. 2004. Zum Wohl: Diskurspartikeln als Satzmodifikatoren. Linguistische Berichte 199. 253–286. Zimmermann, Malte. 2009. Discourse particles in the left periphery. In Benjamin Shaer, Philippa Cook, Werner Frey & Claudia Maienborn (eds.), Dislocated elements in discourse, 200–231. Oxford: Routledge. Zimmermann, Malte. 2011. Discourse particles. In Klaus von Heusinger, Claudia Maienborn & Paul Portner (eds.), Semantics. An international handbook of natural language meaning, vol. 2, 2011–2038. Berlin: de Gruyter Mouton.

Sonja Müller

Combining ja and doch: A case of discourse structural iconicity This paper examines the combined use of the modal particles (MPs) ja and doch. Starting off with the assumption from the literature that the order in MP-sequences is fixed (and should be ja doch in this case), it challenges this claim by referring to corpus data which reveal hits for the reversed order doch ja. By making out three linguistic contexts in which this order can be found, it holds the view that a difference in markedness is at play here. An analysis is developed which explains the difference between the unmarked ja doch and the marked doch ja by building a) on the effect assertions have on the discourse context (modelled within the discourse model by Farkas and Bruce [2010]) and b) on the MPs’ discursive contribution (by relying on the MP-description by Diewald [2007] which gets incorporated into the above model). The main thesis is that the MP-orders mirror communicative requirements which are claimed to be independently and generally held in conversations.

1 Introduction Subject of this paper is the combination of modal particles (MPs) as in (1). (1) a. Sabrina ist ja doch im Urlaub. Sabrina is MP1 MP2 on holiday ‘But remember, Sabrina is on holiday.’1 b. Wer will denn wohl Wasserflugzeug fliegen? Who wants MP1 MP2 seaplane fly ‘Who can it be who wants to fly by seaplane?’ c. Bestell doch ruhig den Hummer! Order MP1 MP2 the lobster ‘Don’t hesitate to order the lobster!’

1 Note that this is just a rough translation as MPs cannot easily be translated into other languages (cf. Schubiger 1965; Burkhardt 1995; Masi 1996, among others). The same objection holds for all MP-sentences which I will translate into English in this article. I believe, it is sometimes even better not to translate the MPs at all. Sonja Müller: Germanistik – Linguistik, Bergische Universität Wuppertal, Wuppertal, ­e-mail: [email protected] DOI 10.1515/9783110497151-009

204 

 Sonja Müller

The possibility to combine MPs is among those properties which are typically referred to when MPs get characterised as a class (beside the missing possibility to inflect, their optionality, occurrence in the middle field, absence of referential meaning etc.) (cf. Thurmair 2013: 628–630; Müller 2014b: Ch. 2; among others). A well-known assumption is that sequences of this type are subject to restrictions. This paper will look at the observation that certain orders are preferred to others.2 Whereas the orders in (1) are perfectly acceptable, the reversed orders in (2) are clearly worse in those sentences.3, 4 (2) a. *Sabrina ist doch ja im Urlaub. b. *Wer will wohl denn Wasserflugzeug fliegen? c. *Bestell ruhig doch den Hummer! In this paper, I will look at one particular MP-sequence, namely the combination of ja and doch, cf. (3). (3)

Konrad ist ja doch / *doch ja verreist. Konrad is MP1 MP2 MP2 MP1 out of town ‘But as you know, Konrad is out of town.’ (Doherty 1987: 114)

I hold the view that the (allegedly) fixed order of ja and doch is an iconic reflex in grammar. Based on new data, I argue that ja doch is indeed the unmarked order. However, linguistic contexts can be found in which the reversed (marked – even if highly marked) order doch ja is attested.

2 Note that there is also the restriction on the combinability of MPs in principle which has priority to the restriction on the relative order in that it decides on the combinations whose order can be looked at. In the literature on MPs, there is agreement that syntactic and semantic/ pragmatic (in)compatibilities play a role here (cf. Dahl 1988: 218, 222–223; Thurmair 1989: 205, 1991: 20, 25–27). My study starts after the compatibility of ja and doch has been established in assertive contexts. 3 I am using the asterisk for reasons of simplicity here. I believe that all one can say is that the sentences in (2) are judged worse than those in (1). 4 Due to limited space, I cannot go into detailed descriptions of existing approaches. The following reference to a number of other pieces of work must suffice at this point: Approaches concerned with the ordering restrictions range from mere classifications (cf. Helbig and Kötz 1981), the formulation of descriptive generalisations (cf. Thurmair 1989, 1991), semantic/ pragmatic criteria, e.g., assertive force (cf. Doherty 1985, 1987), illocutionary weight (cf. Abraham 1995), syntactic conditions, e.g., scope relations (Ormelius-Sandblom 1997; Rinas 2006), input conditions (Doherty 1985; Rinas 2006) and information structural criteria (de Vriendt, Vanderweghe, and van de Craen 1991) to phonological (Lindner 1991) and historical (Abraham 1995) argumentations (cf. Müller 2014a: 170–172, 2014b: 87–89).



Combining ja and doch: A case of discourse structural iconicity 

 205

I suggest a discourse-semantic analysis which integrates the MP-description by Diewald (2007) into the formal discourse model by Farkas and Bruce (2010). My analysis traces the difference in markedness between the two orders of ja and doch back to discourse structural requirements which are assumed to exist in communication independently. In order to illustrate my argumentation, the paper will proceed as follows: In Section 2, I will give the common illocutionary contexts for the combination ja doch, namely assertive clause types, and I will introduce the more controversial assumption about three linguistic contexts in which the reversed order doch ja is attested in corpus data. Section 3 will introduce the formal discourse model by Farkas and Bruce (2010). The components (3.1) as well as their assumptions on canonical reactions to speech acts will be sketched (3.2). In Section 4, I will present my view on the contribution of MPs to discourse which is based on the MP-description by Diewald (2007). I will incorporate her basic ideas into the discourse model presented in Section 3. Equipped with these tools (a model which captures the effect of assertions on the context + an MP-characterisation), Section 5 represents my actual analysis of the difference in markedness between ja doch and doch ja. Section 5.1 brings up the question regarding the scope-interpretation of MP-combinations. In Section 5.2, I will explain in which way I consider the unmarked order to be motivated iconically in discourse structural terms. Section 5.3 will clarify how I account for the marked order doch ja. Section 6 discusses the status of the order doch ja. Section 7 summarises the argumentation.

2 The distribution of ja doch and doch ja It is well known that particular MPs can only combine in certain illocutionary types. As far as ja and doch are concerned, their domain of combination are assertive contexts. Looking at independent sentences, these can be [–wh], V2 sentences of the type in (1a). However, there are also less canonical assertive clause types such as independent V1-clauses or non-local wo-clauses as occur in (4) and (5). (4) [‘The prize will be awarded to citizens who have gained special services. The mayor of Bürstadt Alfons Haag found a lot of entertainment there.’] Ist Deckenbach ja doch weit über die dörfliche Grenze Is Deckenbach MP1 MP2 far over the village borders hinaus als „Kerwe-Opa“ bekannt […]. beyond as Kerwe-grandpa known […].

206 

 Sonja Müller

‘Because Deckenbach is known as “Kerwe-grandpa” far beyond the village borders.’ (M09/DEZ.97728 Mannheimer Morgen, 09.12.2009; modified by S.M.) (5) [‘At first I was a bit choked by the discussion that suddenly occurred, but now I would accept the hints and take them seriously, whether I asked for them or not.’] Wo ja doch eine Menge Arbeit beim Anlegen und der As MP1 MP2 an amount work with creating and the Pflege in das Portal gesteckt wurde. care in the portal put was ‘As a lot of work was put into the portal when creating and maintaining it.’ (http://de.wikipedia.org/wiki/Portal_Diskussion:Musik_der_DDR/Archiv/ Archiv_II; accessed 14.04.2013) The two particles also combine in certain sub-clauses which are classified as assertive, such as causal and concessive adverbial clauses, as in (6) and (7),5 or certain that-complement clauses, as in (8). (6)

da ist mir jedesmal zum heulen zumute wenn ich then is me everytime to.the cry feel when I das sehe da ich ja doch so leidenschaftlich koche. that see because I MP1 MP2 so passionately cook ‘Every time I see that I feel like crying since I cook that passionately.’ (DECOW2012-C00X1M: 41947270)

(7)

Die Jungs rocken auch richtig amtlich ab, obwohl The boys rock also really officially off although Bielefeld ja doch nicht  so wirklich Rock’n’Roll ist. Bielefeld MP1 MP2 not so really  Rock’n’Roll is ‘The boys also rock real hard although Bielefeld is not really Rock’n’Roll.’ (DECOW2012-03X: 50083124)

5 DECOW is a German web corpus contained in COW (Corpora from the Web), a collection of linguistically processed web corpora, cf. Schäfer and Bildhauer (2012).



Combining ja and doch: A case of discourse structural iconicity 

 207

(8) a. Wilkes betonte, dass das ja doch das Ziel der Wilkes stressed that this MP1 MP2 the goal of-the meisten Arbeitssuchenden sei. most job seekers be ‘Wilkes stressed that this was the goal of most job seekers.’ (M02/JAN.03085 Mannheimer Morgen, 14.01.2002) b. Er sinnierte darüber, dass Raucher ja doch gesund He pondered on that smokers MP1 MP2 healthy lebende Menschen seien: living people be ‘He pondered that smokers were people who live healthy lives:’ (RHZ08/FEB.01892 Rhein-Zeitung, 02.02.2008) The general assumption in the literature on MPs is that the order in MP-combinations is fixed. For ja and doch, the judgements are that ja has to precede doch; cf. (9) and (10) as well as the judgements between *, ?? and ? in Jacobs (1991: 157), Meibauer (1994: 101), Ormelius-Sandblom (1997: 93), and Rinas (2007: 431). (9)

Konrad ist ja doch / *doch ja verreist. Konrad is MP1 MP2 MP2 MP1 out of town ‘Konrad is out of town.’ (Doherty 1987: 114)

(10) Er hat ja doch / *doch ja getanzt. He has MP1 MP2 MP2 MP1 danced ‘He danced.’ (Struckmeier 2014: 20) However, in contrast to these standard assumptions made in the literature on MPs, the analysis of authentic data reveals that the reversed order doch ja is actually attested. From my point of view, one can sort the hits very well according to the contexts in which this order occurs. The three classes I assume are a) evaluations, b) epistemically modalised sentences and c) modal causal sub-clauses. The first class comprises utterances with which the speaker gives some qualitative judgement of an issue, in (11), for example, that something is typical.6

6 For more examples see Section 5.3 as well as Müller (2014a: 176, 200–202).

208 

 Sonja Müller

(11) Das ist doch ja wieder typisch. That is MP2 MP1 again typical ‘That’s typical again.’ (A “nerd” runs amok because of frustration on wife and teacher.) http://webcache.googleusercontent.com/search?q=cache: V4RzlhWolBoJ:identi.ca/notice/66906092+%22doch+ja%22&cd=825&h l=de&ct=clnk&gl=de&source=www.google.­de, WebAsCorpus; accessed 21.09.2011) The second class is constituted by epistemically modalised sentences. In those contexts, one finds adverbs such as vielleicht as in (12), epistemically interpreted modal verbs such as müsste in (13) or tag questions such as oder + question mark, for example. These are all means to code a restricted speaker commitment and as (14) shows, they can also occur in combination (here: adverb eigentlich + epistemic modal verb sollte + modal particle/adverb wohl + discourse marker denke ich). (12) [‘I believe this will backfire.’] Vielleicht geht es Dir da doch ja so ähnlich wie mir? Maybe goes it you there MP2 MP1 so similar as me ‘Maybe you feel like me?’ (DECOW2012-00:B00: 553854471) (13) [‘So both Pcbs are fine.’]  Dann müßte es doch ja am Monitor dran liegen oder  Then had it MP2 MP1 on.the monitor up be or flipperdok? flipperdok? ‘It would have to be up to the monitor then, wouldn’t it flipperdok?’ (DECOW2012-01:B01: 1100889124) (14) Eigentlich sollte sich ein solches Gerät bei der Einstellung Actually should self a such device in the setting „Sendersuche: automatisch“ doch ja wohl den besten,  tuning: automatically” MP2 MP1 probably the best aber empfangbaren Sender nehmen, so denke  ich. but receivable dispatcher take, so think I  ‘Actually, in the setting “tuning: automatically”, such a device should probably choose the best, but receivable dispatcher, I think.’ (http://www.gopal-navigator.de/archive/index.php/t-5689.html; accessed 21.07.2014)



Combining ja and doch: A case of discourse structural iconicity 

 209

The third class in which the reversed order doch ja can be found is constituted by epistemically or illocutionarily interpreted causal (sub-)clauses. These are causal clauses which do not give reasons for certain issues but they explain why the speaker makes a certain assumption (cf. [15]), or as in (16) why he performs a certain speech act: In this example, he makes the suggestion to tear the building down and then provides a reason why he suggests that: false ceilings are already missing. (15) Am 28.02.09 dürfte er dann noch einige junge Zuschauer On 28/02/09 may he then else some young viewers diese möchten doch  ja weniger haben, denn garantiert less have because these want MP2 MP1 surely DSDS sehen, da geht es ja um „Alles oder Nichts“ DSDS watch, there goes it MP about all or nothing bei den 15 DSDS Sternchen. for the 15 DSDS starlets ‘On 28/02/09, he will probably have even less young viewers because they certainly want to watch DSDS. All or nothing is at stake for the 15 DSDS starlets.’ (http://www.deutschlands-superstar.de/2009/02/27/wetten-dass-gegendsds-die-entscheidung/ accessed 09/06/12 – taken from Müller 2014a: 176) (16) [‘The Capa-Haus in Jahnallee 61 is still in danger of acute collapse. The historical building is now secured in order to prevent it from collapsing. [...] To ultimately save the Capa-Haus, it would have to be sanified completely by tightening the ceiling and repairing those parts that were damaged by the fire. The agency further reports that the instigation of those works falls to the owner. Posted: 01.02.2012 12:31 anonymous’] Einfach abreißen! Just down.tear ‘Just tear it down!’ denn es fehlen doch ja schon sämtliche zwischen because it lacks MP2 MP1 already all false decken ceilings ‘Because all false ceilings are already missing.’ (http://www.leipzig-fernsehen.de/default.aspx?ID=5846&showNews=1107962 – taken from Müller 2014a: 9)

210 

 Sonja Müller

Because of such examples,7 in the following, I will argue that it is not the case that ja doch is the grammatical order and doch ja the ungrammatical order, but that ja doch is the “unmarked” order and doch ja the “marked” order. This is why the sequence doch ja can always be substituted by ja doch, but not the other way round. This becomes evident from examples as in (9) and (10), for which I share the judgements from the literature. Based on these empirical generalisations, the two research questions I attempt to answer in the course of this paper are the following: a) Why is ja doch the unmarked order which can occur in all assertive contexts? b) Why can the order be reversed only in the three contexts named above (evaluations, epistemically modalised sentences, modal causal clauses)? The following section will introduce two ingredients my analysis in Section 5 will be based upon: the formal discourse model by Farkas and Bruce (2010), which captures the effect assertions have on the discourse context, and a description of the effect ja and doch have in isolation (proposed by Diewald in a similar fashion in a number of pieces of work) on which I will base the interpretation of the combination.

3 The discourse model (Farkas and Bruce 2010) The discourse model the analysis to be presented in Section 5 refers to was suggested by Farkas and Bruce (2010). It comprises four components.

7 Most of the data comes from web sources. However, I also found hits in DeReKo (which also contains web data) and DWDS. Old data is also attested (Projekt Gutenberg, zeno.org). My searches have led to approximately 150 hits. As the reviewers expressed some scepticism about some of the examples, I will discuss the status of the doch ja-hits and their sources in Section 6 in more detail. At this point, I only want to emphasise that ja doch outnumbers doch ja without doubt very clearly. As I collected the data from various sources, it is not possible to sensibly contrast the frequencies of the two orders as I cannot look through all ja doch-sequences. However, I do not think that this is the decisive question regarding my claim for the necessity to consider the reversed order. My point is not about arguing that speakers prefer doch ja to ja doch in any context. What I claim is that one can find systematic patterns in the doch ja-utterances. I, therefore, reject excluding this order on the whole and suggest an account of the “good” order which opens up the possibility for including the reversed one.



Combining ja and doch: A case of discourse structural iconicity 

 211

3.1 Common ground, discourse commitment set and the table A central component in every discourse model is the common ground (cg) which is typically modelled by a set of propositions, cf. (17). (17) cg = {p1, p2, p3, ..., pn} According to Farkas and Bruce, the cg contains the discourse partners’ consciously shared public discourse commitments. These are the propositions to which the discourse partners have committed themselves in public, which they agree upon and which they mutually know that they agree upon. Apart from these consciously shared discourse commitments, speaker and hearer also have individual systems of discourse commitments. A so-called discourse commitment set of interlocutor X (DCX) contains for each participant the propositions to which he has publicly committed himself in the course of the conversation. The last component relevant for my analysis is the table. It saves the open questions of the conversation, or, put differently, what is at issue at a given point in the conversation. As long as elements are placed on the table, there are topics which need to be sorted out. When an assertion is uttered, the components are involved in the following way: Before the assertion is made, a context state (K1) as in (18) exists.8 (18) K1: initial context state DCA Common Ground s1

Table

DCB

Projected Set ps1 = {s1}

The discourse commitment sets of A and B as well as the table are empty, the cg has a particular state (the state it has reached in the conversation up to that point). Against this contextual background, discourse participant A utters the assertion in (19) which results in the context state K2 in (20). (19) The linguist trusts her dentist.

8 Cf. Section 3.2 on the discourse component projected set (ps).

212 

 Sonja Müller

(20) K2: A asserted relative to K1: The linguist trusts her dentist. DCA

Table

p

〈The linguist trusts her dentist [D]; {p}〉

Common Ground s2 = s1

DCB

Projected Set ps2 = {s1 ∪ {p}}

The effect of introducing (19) into the context is that p is added to A’s discourse commitment set. According to Farkas and Bruce, the syntactic structure of the sentence and its denotation are put on the table. In contrast to DCA, the cg does not change; its new state is identical to the previous one. Before discourse participant B does not show a reaction of the sort in (21) (cf. Farkas and Bruce 2010: 99), p remains a contribution by speaker A to which he publicly committed himself. (21) Uh huh/sure/right/you bet/yup. (or non-verbal behaviour such as nodding) From these conditions, the authors derive that assertions can open up an issue by placing an element on the table. To assert a sentence with propositional content p then means to open up the issue whether p.9 If interlocutor B accepts p by some sort of confirming reaction, then he also has a discourse commitment to p. That means that both discourse participants have the same commitment, cf. (22a). Subsequently, p becomes part of the cg as a consciously shared public commitment, cf. (22b), and the issue is removed from the table. (22) K3: B confirmed A’s contribution a. part 1 DCA

Table

DCB

p

〈The linguist trusts her dentist [D]; {p}〉

p

Common Ground s3 = s1

Projected Set ps3 = ps2

9 In order to focus on this aspect, I will actually place p ∨ ¬p on the table later when an assertion is uttered. One reviewer remarks that he considers this to be the effect of polar interrogatives. According to Farkas and Bruce (2010: 82), polar interrogatives and assertions behave analogously concerning opening up an issue on the table. They differ in the way that only assertions involve a commitment on part of the speaker. And, of course, polar interrogatives denote a set of propositions, whereas assertions denote a single proposition. The different appearance of the respective projected set mirrors this difference.

Combining ja and doch: A case of discourse structural iconicity 





 213

b. part 2 DCA Common Ground s4 = {s1 ∪ {p}}

Table

DCB

Projected Set ps4 = {s4}

The context changes described are the moves that are principally assumed to capture the effects that assertions have on discourse.

3.2 Canonical moves Furthermore, the authors assume that conversations are generally driven by the two aspects in (23) (cf. Farkas and Bruce 2010: 87): (23) a. Increasing the cg b. Reaching a stable state The first one is that participants follow the need to increase the cg. As they strive for that, they place elements on the table. The second drive is that participants strive for reaching a stable state, that means a state in which nothing is under debate. Nothing is placed on the table and because of this intention, discourse partners remove elements from the table in that way that the cg increases. More concretely, they also assume that every single move in discourse which places an element on the table is associated with a canonical move to remove this element from the table. The canonical way to remove the issue from the table is to reach a context state in which p or ¬p is part of the cg. The canonical reaction to assertions is, therefore, confirming the assertion. As mentioned above, assertions make proposals by someone making a public commitment. Accepting the proposal is then the most direct way to make the proposition introduced into the context cg-contents. Farkas and Bruce assume a kind of “conversational pressure” to increase the cg by making public commitments to shared public commitments. In order to reach this state, the discourse partner has to accept the contents. This idea of the canonical move which removes an element from the table is captured by its own component, the projected set (ps). A move in discourse which places an element on the table simultaneously projects a set of future cgs in which the topic on the table is decided. An assertion, for example, projects confirmation. This utterance type projects a projective cg in which the asserted proposition is contained (cf. K2 in [20]). In this sense, an assertion has a bias towards the proposition it expresses.

214 

 Sonja Müller

It is also important for the subsequent argumentation that Farkas and Bruce make specific assumptions concerning the effects which assertions standardly have; cf. (24) and the paraphrases in (25). (24) A(S[D], a, Ki) = Ko such that a. DCa,o = DCa,i ∪ {p} b. To = push(〈S[D]; {p}〉, Ti) – {p}10 c. pso = psi  ∪ (Farkas and Bruce 2010: 92) (25) After a speaker has uttered an assertion with proposition p ... a. the speaker’s discourse commitment set contains p. b. the proposition p (vs. ¬p) is put on the table. c. the cg is biased towards p. (25a) is common to all assertions, i.e. all utterances which are classified as assertive (regardless of standard or non-standard) have the effect to update the discourse commitment set of the author of the assertion. This also applies to the subordinate clauses which I introduced in Section 2. A prototypical assertion fulfils all three criteria in (25a) through (25c): It commits the author to its propositional contents, opens up a topic by placing an element on the table and directs the conversation towards solving the topic, i.e. towards a confirmation of the assertion. That means a canonical assertion leads to a context which is categorically biased towards a confirmation of the asserted proposition. After introducing those aspects of the discourse model and of the effect of (non)canonical assertions in context which are decisive for my analysis of the orders ja doch and doch ja in Section 5, the following section will present the relevant issues concerning my notion of MPs.

4 Modelling the meaning of MPs My analysis of the contribution of MPs which will be part of my argumentation on the MP-combinations of ja and doch in Section 5 is based on assumptions about

10 i = input context, o = output context, a = speaker, A = assertion, K = context, T = table, ps = projected set.



Combining ja and doch: A case of discourse structural iconicity 

 215

MPs held by Diewald in a series of pieces of work (cf. Diewald 1997, 1999, 2006, 2007; Diewald and Fischer 1998; Diewald and Kresić 2010).11 In my opinion, her modelling translates easily into Farkas and Bruce’ (2010) model. Diewald describes the MPs’ contribution in three steps: Firstly, the context state immediately preceding the context in which the MP-utterance is made gets characterised. A proposition (or in some cases a speech act) is at issue. Secondly, the speaker takes a certain stance towards the proposition. Thirdly, he actually makes the MP-utterance. She illustrates the specific contribution by doch in assertions like (26) as illustrated in the scheme in (27). (26) Das war doch richtig. That was MP right ‘But that was right.’ (Diewald and Fischer 1998: 92) (27) Meaning scheme for the MP doch (Diewald and Fischer 1998: 92) pragmatically given unit

it is at issue: whether that was right

relevant situation

I think: that was right

→ utterance

Das war doch richtig.

The doch-utterance is preceded by a situation in which the question is open whether it was right, that means the possibilities are: ‘it was right’ or ‘it was not right’. Against this background, the speaker represents one of the two possibilities. Consequently, doch indicates a concessive relation under this view: The speaker decides on the proposition expressed with his assertion although the opposite assumption is also contextually activated. In my opinion, Diewald’s basic idea is to describe the context state in the situation immediately preceding the MP-utterance and after the MP-utterance has been made. I translate these ideas by formulating two context states: one is before and the other after the MP-utterance.12

11 Other approaches to ja and/or doch have been suggested by Doherty (1987), Thurmair (1989), Jacobs (1991), Meibauer (1994), Ormelius-Sandblom (1997), Waltereit (2001), Karagjosova (2003), Rinas (2007), Egg (2013), among others. 12 For reasons of simplification, I will ignore the content of the projected set in the following.

216 

 Sonja Müller

For doch, I assume that in the context immediately preceding the dochutterance, p is under debate. In the discourse model, this means that it is an open issue whether p or non-p is true. Decisively, the cg is unchanged. (28) Context preceding the doch-utterance DCA

Table

DCB

p ∨ ¬p (it was right vs. it was not right) Common Ground s1 Subsequently, the doch-utterance is made and the effect is that the speaker makes a public commitment to p against the background that the two possibilities are open. Decisively, the cg is still unchanged. (29) Context following the doch-utterance DCA

Table

p it was right

p ∨ ¬p (it was right vs. it was not right)

DCB

Common Ground s2 = s1 There is a certain variability when it comes to mentioning what can be at issue and for what reason. Contents can be at issue because it is implied, implicated or presents a speech act condition. In the following, I will present a doch-utterance which relates to an implicature. The existence of such examples ties in nicely with Farkas and Bruce’s original account as they assume that implicated contents can also be placed on the table (cf. Farkas and Bruce 2010: 94). Examples of the type in (30) are typical for uses of doch as an MP. (30) A: Peter wird auch mitkommen. Peter will also come along ‘Peter will also come along.’



Combining ja and doch: A case of discourse structural iconicity 

 217

B: Er ist doch krank. He is MP ill ‘But he is ill.’ (Egg 2013: 126) In some sense, the doch-utterance refers to the consequences of the first utterance. With Ormelius-Sandblom (1997: 83–84), I assume that a conventional implicature is involved in examples of that sort. (31) Peter will also come along. > Peter is not ill. The relation in (30) certainly cannot be considered a logical conclusion, but a pragmatic inference is at hand which is based on background/world knowledge and which represents a plausible connection (for instance: If Peter will come along, he is normally not ill.). As Farkas and Bruce (2010) assume that implicated contents can be at issue on the table, the contexts involved in (30) can be modelled as in (32) and (33). (32) Context preceding the doch-utterance

A: Peter wird auch mitkommen. (= p) > Er ist nicht krank. (= ¬q) DCA

Table

p

p ∨ ¬p q ∨ ¬q

DCB

Common Ground s1 A has a public commitment to p (via his assertion), thus putting p ∨ ¬p up for discussion (i.e. the bipartition lies on the table). The proposition p has the implicature ¬q, so that the further open question on the table is q ∨ ¬q. B reacts to this open question with his public commitment to q.13

13 Cf. Müller (2014a: 188, fn. 20) on a discussion of whether one has to attribute a commitment towards ¬q to A and on whether the relation p > ¬q has to be considered cg-content.

218 

 Sonja Müller

(33) Context following the doch-utterance B: Er ist doch krank. DCA

Table

DCB

p

p ∨ ¬p q ∨ ¬q

q

Common Ground s2 = s1 As mentioned above, doch-utterances can also relate to implied propositions and speech act conditions. Due to lack of space, I cannot discuss such examples here. The same applies to discourse-initial uses of doch which seem to go against the modelling presented above at first glance (cf. Müller [2014a: 185–187] for the parallel treatment of such doch-utterances). Modelling the contribution by ja, according to Diewald and Fischer (1998), in the context preceding a ja-utterance such as (34), there is someone else (usually the hearer) who presumes exactly the proposition which is contained in the speaker’s assertion. (34) Es soll ja fliegen. It shall MP fly ‘As you know, it is supposed to fly.’ (after Diewald and Fischer 1998: 93–94) That means, the speaker affirms an already existing assumption, cf. (35). (35) Meaning scheme for the MP ja (after Diewald and Fischer 1998: 93–94) pragmatically given unit

it is at issue: someone thinks that it is supposed to fly

relevant situation

I think: it is supposed to fly

→ utterance

Es soll ja fliegen.

Translating this description into Farkas and Bruce’ (2010) model, I assume that the hearer already has a discourse commitment to the contents of the ja-assertion which will be introduced to the context in the next step, cf. (36).

Combining ja and doch: A case of discourse structural iconicity 



 219

(36) Context preceding the ja-utterance DCA

Table

DCB p (it is supposed to fly)

Common Ground s1 After the ja-utterance has been made, cf. (37a), the speaker A has a public commitment to the proposition contained in the assertion. By making this assertion, A places p on the table and the topic is opened up whether p or non-p is the case. However, different from assertions without ja, p gets directly removed from the table because B already has his public commitment to p which is needed for p to get into the cg. As both discourse partners have a public commitment to p after A’s ja-utterance, p can enter the cg as in (37b). (37) Context following the ja-utterance a. part 1 DCA

Table

DCB

p (it is supposed to fly)

p ∨ ¬p (it is supposed to fly vs. it is not supposed to fly)

p (it is supposed to fly)

Table

DCB

Common Ground s2 = s1

b. part 2 DCA Common Ground s3 = {s2 ∪ {p}}

(it is supposed to fly)

As mentioned above, there are different uses of doch-assertions one can make out. Similarly, one can differentiate between diverse contexts of use of ja-utterances. I will restrict the presentation again to one case. In the situation in which (38) is made, the speaker notices that a certain element does not/will not fit.

220 

 Sonja Müller

(38) Oh dann mußt du es ja nochmal abmachen. Oh then must you it MP again remove ‘Oh, you’ll have to remove it again then.’ (Diewald and Fischer 1998: 93) As long as the speaker and the hearer share the same perceptual situation, the speaker can transparently ascribe the public commitment to the addressee that the piece is not fitting and that he (the hearer), therefore, has to remove it. Consequently, in this context of use, there is honestly and transparently an actual reason for assuming the presupposed hearer commitment. Among other uses of ja, there are also contexts in which it is obvious that the speaker only insinuates the hearer’s commitment. This can happen in harmless scenarios in which the hearer simply cannot know about the proposition whereas the speaker is absolutely certain or in perfidious contexts in which the hearer would never agree with the speaker on his own. The above modelling of ja’s contribution also covers the use in which p is already part of the cg and (for some reason) gets referred to again in the conversation (cf. Müller [2014a: 189–191] for concrete examples and more detailed descriptions).14

14 One reviewer is right in pointing out that doch is accompanied by some sort of controversy in most cases. I derive from that comment that he considers the openness of p too weak a requirement. I agree that the openness of p often arises because non-p can be inferred from another utterance. However, this does not apply to all uses (e.g. discourse initial uses and the occurrence of doch in verb-first-/wo-verb-final assertions). I see the advantage in Diewald’s modelling I am building my analysis on in capturing the controversy-contexts as well as the noncontroversy-contexts. Openness can be due to controversy, but it does not have to. The reviewer also points out that doch is not used to confirm p, even if it is open. However, this still does not convince me that non-p always has to be available in the context preceding a doch-utterance and should, therefore, be part of the meaning of doch. In directives, the use of doch seems also possible in confirming reactions: (i) A: Ich gehe dann jetzt. I go then now ‘I’ll go right now.’ B: Ja, geh doch. Yes go MP ‘Yes, do so.’ The question then is whether one still wants to ascribe controversy to such examples or whether it is not rather the openness of the action being realised or not which motivates the use of doch here. It is a well-known problem in the literature on MPs that (alleged) aspects of meaning have been ascribed to single particles which turned out to be attributable to other factors. Therefore, I think, one should be careful with formulating very rich MP-contributions. I agree that ja cannot refer to a proposition which has been mentioned in the previous turn. This is a phenomenon known and presents a further requirement on the context. If explicitly uttered, the proposition which the hearer is assumed to hold already has to be an older contribution (cf. Repp 2013: 246–248).



Combining ja and doch: A case of discourse structural iconicity 

 221

After establishing the view to MPs in isolation to be relied upon in my account, the next section will first look at the interpretation of the combined use of MPs and will afterwards offer an approach that accounts for the marked and unmarked order of ja and doch in sequences.

5 The combination of ja and doch When it comes to interpreting MP-combinations, there is quite some controversy in the literature regarding how the MPs involved take scope. When two MPs occur and each of them takes scope, there are in principle four possibilities how they could take scope: (39) a. Different scope i. ja(doch(p)) (Ormelius-Sandblom 1997; Rinas 2007) ii. doch(ja(p)) (Lindner 1991) b. Same scope iii. 1. ja(p) & 2. doch(p) (Thurmair 1989) iv. 1. doch(p) & 2. ja(p) (Doherty 1985) That means, there is an open question whether a hierarchical relation obtains between two particles in a combination or whether they display the same scope. A surprising fact is that all interpretations in (39) have been suggested for the sequence ja doch.

5.1 Against a scope relation between the two MPs In my opinion, what happens in context in case the combination ja doch occurs is that ja and doch take the same scope. That means they both scope over p and the result is an interpretation in which the two MP meanings add up. This assumption has also been held by Thurmair (1989, 1991). However, a dominant assumption in the literature is that (39a) is the adequate interpretation and authors who argue for that assume that the (putatively) fixed order is a reflex of the scope relation (cf. Ormelius-Sandblom 1997; Rinas 2007; among others): Because ja takes scope over doch, ja precedes doch. Due to lack of space, I cannot back up my decision for (39b) by comparing the four interpretations when applied to data (for a more detailed discussion, cf. Müller [2014a: 192–194]). However, I believe that the fact that the reversed order is attested is an argument against a scope relation and against pieces of work which want to derive the linear order from the scope relations in this case – unless one wanted to assume that the scope relation gets reversed by reversing

222 

 Sonja Müller

the order of the MPs. If one wanted to assume that the two orders have different meanings in terms of diverse scope relations, one would have to explain the different possibilities for substitutions. At this point, I can only show what the decision for “same scope” means for the interpretation of an example such as (40) (for an example in context, cf. Müller [2014a: 195–197]).15, 16 (40) Konrad ist ja doch verreist. Konrad is MP1 MP2 out of town (41) Context preceding the ja doch-utterance 1. ja(p), 2. doch(p) DCA

Table

DCB

p ∨ ¬p (doch)

p (ja)

Common Ground s1 The context state preceding the ja doch-utterance is such that B has a public discourse commitment towards p (this is the contribution by ja). Additionally, it is open whether p, that means p vs. non-p is placed on the table (this is the contribution by doch). By uttering the ja doch-sentence, A makes a public commitment

15 This assumption on the interpretation of this MP-combination is definitely not compatible with all existing syntactic accounts concerned with representing MPs within a tree geometry. There is only few work on the syntactic modelling of MPs in general (cf. Coniglio 2007; Struckmeier 2014, among others). As most representations focus on hierarchical constellations, they usually predict a scope relation between particles in a sequence (even though I only know of Vismans [1994], Eroms [2006: 1021] and Coniglio [2011: 98, 118] who are explicitly concerned with MP-combinations in this context). In my opinion, this applies to Abraham’s (1995) generative account in which MPs occupy the head-positions of three different topic phrases on the level of logical form as well as to Visman’s (1994) functionalist approach in which the particles operate on different sentential levels which have dominance relations to each other. Coniglio (2011: 98, 119), however, suggests that MPs form particle-clusters via adjunctions. Capturing the additive interpretation of MP-sequences becomes possible more easily under this view. 16 For an outline of the few existing approaches on the combination of ja and doch (Doherty 1985, 1987; Ickler 1994; Lindner 1991; Ormelius-Sandblom 1997; Rinas 2007), see Müller (2014a: 170–172). My main objection towards these accounts is that ‒ even if one wants to reject my thesis that the reversed order doch ja exists ‒ (apart from Ickler 1994: 404) no author provides an explanation for the reason that reversing the order should not be possible. They offer explanations for deriving the “acceptable” sequence, but do not exclude the “unacceptable” order.

Combining ja and doch: A case of discourse structural iconicity 



 223

to p against the background of p vs. non-p being an open issue and as B already has a discourse commitment, p becomes part of the cg. (42) Context following the ja doch-utterance 1. ja(p), 2. doch(p) a. part 1 DCA

Table

DCB

p

p ∨ ¬p

p

Table

DCB

Common Ground s2 = s1

b. part 2 DCA

Common Ground s3 = {s1 ∪ {p}}

5.2 Explaining the unmarked order ja doch Assuming that this is the correct interpretation in terms of scope, in the following, I will derive the order of ja and doch by referring to some form of iconicity. Iconicity has been assumed to hold for a number of phenomena in functional approaches to word order. Authors who favour such approaches assume that there is a relation between the syntactic order and the order in which things get interpreted. Within the framework of Functional Grammar, Dik (1997: 399) formulates the principle in (43). (43) General Principle 1 The Principle of Iconic Ordering: Constituents conform to (GP1) when their ordering in one way or another iconically reflects the semantic content of the expression in which they occur. To give one example, Dik (1997: 399) assumes that in the unmarked case, the order of sentences in a text mirrors the order of events they describe. Subordinate clauses with the meaning ‘after X’, therefore, precede the main clause, cf. (44).

224 

 Sonja Müller

(44) a. After John had arrived, the meeting started. (unmarked) b. The meeting started after John had arrived. (marked) Consequently, I would like to argue that the order of MPs is motivated in a sense in which it is motivated why temporal sequences, for instance, have an impact on marked and unmarked orders of main and subordinate clause. My case is not about temporal relations, but about the most direct mapping of what happens in discourse. In Section 3.2, I mentioned that Farkas and Bruce (2010) formulate two aspects that drive conversations in general: The first one is that participants follow the need to increase the cg and the second one is that participants strive for reaching a stable state, that means a state in which nothing is under debate. (45) a. Increasing the cg b. Reaching a stable state According to what I assume to be the contribution of ja and doch in discourse, doch refers to an unstable context state. The bipartition p vs. non-p is placed on the table before the MP-utterance and the speaker makes a commitment to p when uttering the MP-assertion. However, agreement can never be reached with a doch-utterance as the issue is open before and after the MP-statement. The use of ja, however, always results in a stable context state. In the context preceding the MP-utterance, the discourse partner has the same assumption as the assertion will introduce. Therefore, speaker and hearer have the same public commitment and the proposition becomes cg. Thus, when ja and doch occur together, there is always one element which creates a stable context state (ja) and one element which refers to an unstable state (doch). If it is the topmost intention of a conversation to increase the cg and to reach a stable state, an assertion complies with this aim in the most direct way if it offers that element which makes the context stable and which renders the proposition part of the cg immediately. If ja is offered first in the combination of ja and doch, what speaker and hearer want (namely stability) is immediately established because they agree on the proposition which is under debate. However, if doch is offered first, the speaker at first only refers to the concessivity (i.e. although two options are possible, he chooses one of them) and in the next step, he says that he is committed to the same proposition as the hearer is and, therefore, they agree on the issue and p is part of the cg. In this sense, I believe it is unmarked – because iconic in discourse structural terms – to offer the ja, which resolves the unstable state presupposed by doch, first. What is formulated in (45) is a superordinate principle operating in discourse and, therefore, it is no surprise that it applies across constructions whenever ja and doch occur. As a consequence, ja doch is an acceptable order in all contexts in which ja and doch combine.



Combining ja and doch: A case of discourse structural iconicity 

 225

Section 3.2 introduced what Farkas and Bruce (2010) consider to be prototypical assertions, cf. (25), repeated here in (46). (46) Prototypical assertions a. The speaker’s discourse commitment set contains p. b. The proposition p (vs. ¬p) is put on the table. c. The cg is biased towards p. In my opinion, this type of assertion easily adheres to the superordinate principle in (45) (from which I derive the order of ja and doch) because it wants p to become cg anyway – even if ja and doch do not occur. The way I interpret the contextual effect of the combination of ja and doch, this type of assertion does what the speech act independently wants. If such a speech act gets an element which can create the situation the speech act wants to create anyway, this element will be put to use before other elements which cannot do that. Therefore, in my opinion, it is natural that such types of assertions favour the order ja doch. The notion of iconicity I rely on here and which is typically illustrated by examples of the sort in (44) is called diagrammatic iconic motivation in Haiman (1980) and is defined as follows: “a grammatical structure, like an onomatopoeic word, reflects its meaning directly” (Haiman 1980: 516). This form-function parallelism or isomorphic correspondence is not explicitly tied to Dik’s (1997) work, but is considered a principled form of iconicity (cf. the overview in Pusch [2001]). Obviously, iconicity is a concept integrated in functional models of grammar (however, cf. Newmeyer [1992] on built-in iconicity in generative approaches). The more general underlying assumption is formulated by Croft (1995: 129): “the principle that the structure of language should, as closely as possible, reflect the structure of experience, that is, the structure of what is being expressed by language”. I do not believe that the violation perceived with doch ja is too strong to be captured by such a principle. On the one hand, I assume that the reversed order should not be excluded completely. On the other hand, although (44) is a typical textbook example to illustrate iconic ordering, in the literature on the topic, it is no must that the abnormal case is acceptable and simply less frequent. Other constructions which have been claimed to be instances of this principle, involve clearer deviations regarding their acceptability, cf. the coordinations in (47) and (48) and the binomials in (49) and (50). (47) a. He opened the door, came in, sat and ate. b. *He sat, came in, ate and opened the door. (48) a. He shot and killed her. b. *He killed and shot her. (Givón 1991: 92)

226 

 Sonja Müller

(49) a. auf und ab up and down b. *ab und auf (50) a. Herz und Nieren heart and kidneys b. *Nieren und Herz (Plank 1979: 140) In fact, I doubt that the difference in markedness between the two particle orders is felt strongly by speakers.17

5.3 Explaining the marked order doch ja Although prototypical assertions easily adhere to (45), the prototypical assertion is only one type of an assertion. Other constructions which are classified as assertive (because the speaker has a commitment to the proposition expressed) can have their own intentions which deviate from prototypical assertions. I argue that those are cases in which it is not the topmost intention of the concrete construction type to make the expressed proposition become cg. Therefore, for those structures, it is also natural not to offer the cg-marker ja early in the MP-combination. The result is the order doch ja, even if ja doch is possible, of course, because the speaker can always stick to (45) when uttering an assertion. Consequently, I do not expect the reversed order doch ja to occur in prototypical assertions whose intention it is to make its contents part of the cg. Then the last question is: In which way do the three linguistic contexts in which the order doch ja is attested (evaluations, epistemically modalised sentences, modal causal (subordinate) clauses, cf. Section 2) confirm this idea? That means to what extent it is not their topmost aim to make their contents cg-contents. In my opinion, the three contexts confirm this idea in so far as they share the property to display a form of subjectivity: The proposition expressed is not primarily “told” by the speaker, but it gets “evaluated” by him. In this sense, assertions in the three contexts are not predominantly about convincing the discourse partner of this rating.

17 One reviewer is sceptical about whether iconicity could be lexicalised/grammaticalised in the MP-sequence. In my opinion one should not rule out this possibility completely. There are pieces of work which argue for the relevance of iconicity in lexicalisation/grammaticalisation processes (cf. Giacalone Ramat 1995; Fischer 1999).



Combining ja and doch: A case of discourse structural iconicity 

 227

In the first context (evaluations), such a rating arises with regards to content. typisch. (51) Das ist doch ja wieder That is MP2 MP1 again typical ‘That’s typical again.’ (A “nerd” runs amok because of frustration on wife and teacher.) (52) [‘I think most women will notice in time that they are female and when they are addressed as a man now and then, they will not be insulted.’] „Sehr geehrte Frau Minister!“ Ist doch ja auch ganz very honoured Mrs minister is MP2 MP1 also quite hübsch. nice ‘“Dear Minister!” This is also quite nice.’ (DECOW2012-06: 697189512 – taken from Müller 2014a: 28) (53) dass bronski jetzt auch noch so genial ist das mit that Bronski now also even so brilliant is this with dem thema kultur überhaupt zu verflechten the topic culture at-all to interweave ist doch ja schon die speerspitze der europäischen is MP2 MP1 already the spearhead of-the European bewegung (*grins) movement (*grin) ‘That Bronski now also is so brilliant to interweave this with the topic “culture” at all is already the spearhead of the European movement (*grin).’ (DECOW2012-00:B00: 331498526) (54) Das ist doch ja auch einer der interessanten Aspekte This is MP2 MP1 also one-of the interesting aspects an Star Trek, dass es so eine klare Trennung zwischen of Star Trek that it such a clear difference between Gut und Böse nicht gibt. good and evil not gives ‘One of the interesting aspects of Star Trek is that there isn’t such a clear difference between good and evil.’ (DECOW2012-03:B03: 254862193) In my opinion, with all these utterances the speaker gives an evaluation of an issue. In almost all of the cases of this type a copula-construction is involved of the form this is + X. Typically, the predicate is realised by an evaluative adjective,

228 

 Sonja Müller

as typisch ‘typical’ in (51) (= [11]) and hübsch ‘nice’ in (52). Apart from adjectives, X can also be realised by nominal phrases which either display evaluative parts of meaning by themselves, e.g. Speerspitze ‘spearhead’ in (53), or which cooccur with an evaluative adjective in case they are themselves semantically relatively empty, e.g. Aspekte ‘aspects’ in (54). A speaker who makes such an evaluation does not pursuit the topmost aim to make the hearer agree on this evaluation. Since increasing the cg is not the overall aim of this type of utterance, the MP which can create this state directly does not have to be offered as soon as possible. Of course, the expressed content becomes shared information (otherwise ja would not have to be used at all), but it is not the speech act’s main aim. Something similar can be said about epistemically modalised sentences which present a further context in which the order doch ja can be found, cf. (55) through (57) (see also Section 2). (55) Wir haben doch ja nun wirklich bei Gott genügend We have MP2 MP1 now really by God enough weiße Flecken auf der Karte […]. white spots on the map ‘By God, we now really have got enough white spots on the map […].’ (DECOW2012-00:B00: 618521882) (56) [‘Against who should the blu-ray establish itself?’] Sie hat doch ja schon gegen HDDVD gewonnen oder? She has MP2 MP1 already against HDDVD won or ‘It has already won against HDDVD, hasn’t it?’ (DECOW2012-02:B02: 568050573) (57) Die Einzige, die mir da Mut gemacht hat, war The only-one who myself there courage made has was meine  Logopädin, die mir sagte, ich hätte auch my speech-therapist who me said I had also ohne CI mein Sprechen gut kontrollierte, es konnte without CI my speech well controlled it could doch ja also so schlimm nicht gewesen sein. MP2 MP1 thus so bad not have been ‘The only person who encouraged me was my speech therapist who told me, I had controlled my speech well without CI. Consequently, it couldn’t have been that bad.’ (http.//www.kestner.de/n/elternhilfe/berichte/nf2.htm, accessed 24.10.2014)



Combining ja and doch: A case of discourse structural iconicity 

 229

The concrete linguistic means which yield the epistemic modalisation can vary (adverbs, e.g., nun, wirklich, bei Gott in [55], tag-questions, e.g. oder? in [56], or epistemic modal verbs, e.g. konnte in [57]). As seen in Section 2, these means also like to combine. The effect of modal expressions as occur in (55) through (57) is to express that the speaker ascribes a smaller or bigger likelihood to the issue described. Therefore, it is not central to render p shared knowledge, but to express one’s evaluation regarding the likelihood of p’s realisation. And this again is the reason to say that the structure’s need does not independently prefer to have ja occur before all other MPs because the speaker wants to comply with discourse’s superordinate aim and establish cg. The last context in which the order doch ja can be found are causal clauses which are interpreted epistemically or illocutionarily. That means, it is never the case that a cause-effect-relation between two issues is established. It is either motivated why the speaker makes a certain assumption (epistemic interpretation, cf. [15], repeated in [58]) or why he performs a certain speech act (illocutionary interpretation, cf. [16], repeated in [59], and [60]) (cf. Sweetser 1990; Küper 1992; Blühdorn 2006; Volodina 2010, among others). Modal causal clauses are assumed to be more subjective than propositional causal clauses as they refer to attitudes. (58) Am 28.02.09 dürfte er dann noch einige junge Zuschauer On 28/02/09 may he then else some young viewers weniger haben, denn diese möchten doch ja less have because these want MP2 MP1 garantiert DSDS sehen, da geht es ja um „Alles surely DSDS watch, there goes it MP about all oder Nichts“ bei den 15 DSDS Sternchen.    or nothing for the 15 DSDS starlets ‘On 28/02/09, he will probably have even less young viewers because they certainly want to watch DSDS. All or nothing is at stake for the 15 DSDS starlets.’ (http://www.deutschlands-superstar.de/2009/02/27/wetten-dass-gegendsds-die-entscheidung/ accessed 09/06/12 – taken from Müller 2014a: 176) (59) [‘The Capa-Haus in Jahnallee 61 is still in danger of acute collapse. The historical building is now secured in order to prevent it from collapsing. [...] To ultimately save the Capa-Haus, it would have to be sanified completely by tightening the ceiling and repairing those parts that were damaged by the fire. The agency further reports that the instigation of those works falls to the owner. Posted: 01.02.2012 12:31 anonymous’]

230 

 Sonja Müller

Einfach abreißen! Just down.tear ‘Just tear it down!’ denn es fehlen doch ja schon sämtliche zwischen because it lacks MP2 MP1 already all false decken ceilings ‘Because all false ceilings are already missing.’ (http://www.leipzig-fernsehen.de/default.aspx?ID=5846&showNews=1107962 – taken from Müller 2014a: 9) (60) Muss da echt das Heckrohr gewechselt werden? Wenn Must there really the tail-pipe replaced be if der Riemen noch schön läuft? the belt still fine runs Die Heckabspannung ist doch ja auch nur für die The tail-anchoring is MP2 MP1 also only for the Optik. optics ‘Does the tail pipe really have to be replaced? If the belt is still running fine? As the tail anchoring is only for optical effects anyway.’ (DECOW2012-02: 318143690) The prominent function of such causal clauses is to give reasons for and motivate the main clause which they relate to. However, it is not their topmost aim to bring about that their own contents is accepted by the hearer. Epistemically and illocutionarily interpreted causal clauses can be claimed to carry subsidiary illocutionary force which means that they serve the function to assure the success of another illocutionary type (cf. Motsch 1987: 58; Brandt and Rosengren 1992: 21; Pittner 2007: 54, 2011: 178, on this concept). This captures exactly my assumption that the focus is on the illocutionary types they give reasons for and whose success they want to guarantee and not on themselves. Thus, it is again not so much the utterances’ own need to make the proposition expressed become cg and, therefore, ja which can directly fulfil this aim does not have to be provided immediately in the MP-combination. And again, the order of the two MPs can be reversed. As preliminarily stated, a more abstract look at the three contexts reveals that one is dealing with subjectivised contexts here. There is an evaluation marked as such by the speaker. This can arise due to the sentences’ content (context 1), the sentences can be epistemically modalised by adverbs, modal verbs, tags (context 2) or modally interpreted, i.e. epistemic/illocutionary causal clauses can occur. The



Combining ja and doch: A case of discourse structural iconicity 

 231

sentences thus have the property to foreground the speaker’s attitude and his assessment of the proposition. And in this sense, none of the three contexts is predominantly about convincing the discourse partner of this evaluation. If the speaker does not intend mutual consent with the utterance in the first instance (what ja can instantly bring about), it is also motivated why it does not have to make up the left margin of the combination. Put differently, in my opinion, there is reason why the order gets reversed (if it is reversed) in those three subjectivised contexts for which making p cg is not so very central. However, in all three contexts, it is of course also conceivable to follow the superordinate discourse principle and to create a stable context state as soon as it is possible. In this case, the order chosen is ja doch.18

6 The status of the sequence doch ja As my claim on the existence of the reversed order doch ja challenges a longbelieved assumption, I would not be surprised if some aspects might evoke scepticism in the reader. As mentioned in footnote 7, many of the examples are from web sources. One might reject using such data for linguistic theorising at all. The argument could go that one can find anything on the web. Obviously, working with data of any sort does not work that way to build theories on single examples and, therefore, just finding something in web data does not prove its existence either. However, I believe that the hits are not only sporadic occurrences. Collecting approximately 150 examples across diverse corpora/data sets for doch ja (for which I see no reason to assume that the two elements clearly belong to different parts of speech, see below), leads me to the conclusion that the reversed order occurs systematically in the three contexts described. The three contexts are highly re-occurring patterns, especially the modalisations.

18 As the reviewers correctly point out, my explanation presupposes that the structure gets interpreted from left to right or top-to-bottom. The alternative (which has also been suggested for MP-sequences in Doherty [1985]) is a bottom-to-top-interpretation. Obviously, the explanation then cannot rely on the fact that information tied to the first particle in the sequence is earlier part of the interpretative mechanism than that related to the second one. I believe the first view is commonly presupposed in information structure-affine pieces of work (by that I mean tendencies such as early introduction of the topic in a clause, late presentation of rheme/ information focus).

232 

 Sonja Müller

I also do not believe that the reason for finding it predominantly in this type of data (even if not exclusively) is due to the sloppy nature of the language used there. One might criticise the incorrect/missing punctuation and orthography in some of the examples and the colloquial style. Of course, these are issues one normally does not face when working with (seemingly) “serious” data such as data from newspapers or minutes from political debates. However, I do not think that such data does justice to a phenomenon of (colloquial) oral speech. One also does not know in how far such things get corrected/standardised before they find their way into such data sets. In my opinion, oral data is the only sensible alternative. To me, web data is as valid a data source as spoken language. They are more appropriate for my purpose because corpora of oral language are too small to investigate rare phenomena. As the reversed order is definitely clearly less frequent in comparison to the unmarked sequence, the data set has to produce a respectively large amount of ja doch before I would even expect to find a doch ja. In my view, one should in fact highlight the advantage of such data: It is exactly that type that gives a realistic picture of conceptually oral speech in a written medium. These thoughts are meant to show that the sloppiness or colloquial speech sometimes found on the web is not the reason why one finds the reversed order there. However, for reasons to be mentioned in the next paragraph, this particular phenomenon is more prone to errors than others. But this aspect is true, relatively independent of the type of data involved. Concretely, one might object that we cannot be sure that we are really dealing with the sequence doch ja here. The examples might be cases of selfcorrection: The speaker uttered doch and replaced it by ja. As no prosodic clues are available in written data, we indeed cannot know about that. However, the same problem also arises with all unmarked orders. The writer could have started off with ja and decided to replace it by doch. The resulting order ja doch would not exist then. If that was the case (which one can neither prove nor refute), I would be surprised that the reversal happens in exactly those three contexts. It should equally happen in standard assertion contexts. I would also be surprised about the fact that this occurs in written data as the speaker has the chance to erase his writing. One would expect to find other remains of corrections as well then. Apart from the fact that one definitely wants to exclude corrections in the data, I do not believe that a possibly occurring prosodic break between two particles is per se an argument against the combination. It would be a problem if one wanted to treat an MP-sequence as a complex element of the form



Combining ja and doch: A case of discourse structural iconicity 

 233

[MP1 + MP2](p). As my assumption is that the MPs operate on the same proposition ­separately, an intonational break is no counter argument. It is well known that material can occur between two particles. This also applies to doch ja, cf. (61) and (62). (61) Aber das wird doch beim Internisten ja gar nicht But that will MP2 at.the internist MP1 at-all not lange dauern oder? long take or ‘But it won’t take long at the internist, will it?’ (DECOW2012-06X: 685575138) (62) wenn das klappt müsste sich das doch im prinzip ja if that works had self that MP2 in principle MP1 auf jedes model von u1 bis mw ausweiten lassen, oder? to every model from u1 to mw expand let or ‘If that works out, it should be possible to expand it to every model from u1 to mw, shouldn’t it?’ (DECOW2012-06X: 999332683) Nobody would refute the existence of an established sequence on that ground. Another objection one might raise is that one is not dealing with the MPsequence doch ja here, but with parts of frozen constructions involving the following material, such as das-ist-doch + ja-wieder or es-Dir-da-doch + ja-so. I doubt that this applies to all examples in the paper ‒ aside from the fact that I do not know how to decide when a construction is at hand and, in particular, when it is not. If one wants to assume that ja-am-Monitor-dran, ja-wohl, ­ja-garantiert, ­ ja-schon, ja-auch, ja-nun-wirklich-bei-Gott, ja-also, ja-gar and ja-auf are all formulaic, I have no counter argument. In that case, one would also have to assume that ja only scopes over its construction. I doubt that (at least for the cases without material between the two particles). If one wanted to argue that the examples I give are not true examples, one had to independently define the constructions involved. For me, it is no counter argument to the data when arguing for a semi-frozen structure whenever something appears one does not want to believe in. One might also think that the doch preceding the ja is not the MP, but the stressed adverbial variant. It is obviously the case that the duplicates in other parts of speech are a problem accompanying this topic. Therefore, it can become difficult to decide which reading is intended, especially in the combinations. Thurmair (1989: 209) writes this about ja doch and I believe that it is rarely ­possible at all to exclude the occurrence of the duplicate when a “MP” occurs

234 

 Sonja Müller

in authentic data presented in written form. In this sense, I believe it is impossible to argue that non-MP-readings are altogether excluded in the doch ja-cases. All that is required (as in the case ja doch) is that the MP-reading is available. The ­argumentation cannot work in the following way: Forms might be ambiguous in the order one knows and believes in. But when one finds the reversed order (which one does not want to believe in), one knows for sure that the respective form is not the MP. When looking at the examples I am referring to in this article, I believe that as far as imagination and contextual information go, the adverbial reading of stressed doch is highly implausible in (11), (14) through (16), (52) through (57) and (61). I do not believe that doch ja suffers more from this risk than ja doch. Apart from that, I see two more general aspects which speak against this possible objection: Thurmair (1989: 216) refers to the combinations doch+halt/ eben and halt/eben+doch, writing that the precedence of doch rather goes together with its unstressed occurrence whereas it is stressed in final position. I, therefore, consider it rather implausible that a doch preceding another particle behaves completely different in this respect. Thurmair (1989: 215–216) also assumes that doch eben as well as eben doch can contain two particles (the same applies to doch halt and halt doch). In the first case (with the preceding doch), the particle reading is clearer. I believe, the further a “MP” occurs to the right, the more difficult it becomes to differentiate it from its adverbial counterpart. Of course, there are authors who assume that adverbs can scramble. As the accent on adverbial doch has been related to contrast (cf. Meibauer 1994, among others) and as contrastively focussed material can scramble, the possibility seems sound from that perspective. However, to my mind one cannot claim that enough is known about the interaction of adverbs and MPs (in my opinion existing accounts rather show that the data is unclear, cf. Coniglio [2007]) in order to rely on that as a counter argument. As there are also accounts which argue that MPs can move whereas the adverb’s position is fixed (Coniglio 2007), we would have to know about DOCH’s base position first in order to see how ja can be placed in relation to it. To conclude this discussion on the data’s status, I believe that a) there is no reason not to work with web data (on the contrary it presents a realistic picture of language use and allows studying rare constructions) and b) the reversed orders are no illusions in the sense that one is clearly not dealing with the two particles in all the examples. MPs are considered the result of grammaticalisation. To my knowledge, nothing is known about the diachrony of MP-combinations. There could be



Combining ja and doch: A case of discourse structural iconicity 

 235

changes as well (even though I actually found older hits dating back to Luther). If it had not been for potential ambiguity, no MP would exist at all in grammatical description.19, 20

7 Summary and conclusion By looking at the sequences ja doch and doch ja, I examine one particular combination of MPs. By approving of the literature on MPs that one has to act on the assumption of a principle restriction on the combinability to assertive contexts (cf. Section 2), I argue on the basis of corpus data that one cannot assume that ja doch is the only acceptable order. I suppose that one is dealing with a difference in markedness here, with ja doch being the unmarked and doch ja the (highly) marked order. After introducing the relevant background concepts concerning the effect of (prototypical) assertions in context within the discourse model by Farkas and Bruce (2010) in Section 3 as well as the two particles’ contributions to discourse within this modelling (Section 4), I derive the unmarked order of ja and doch by presuming that this sequencing is the matter of an iconic reflex of a general strive in discourse: increasing the cg and reaching a stable context state. This assumption presupposes that both MPs “equally” relate to the proposition expressed. Therefore, I argue against a scope relation between the two MPs in Section 5.1. In Sections 5.2 and 5.3, I go back to the characterisation of a prototypical assertion from Section 3.2 and ascribe the possible occurrence of doch ja in certain linguistic contexts (evaluations, epistemically modalised sentences, modal causal clauses) to the presence of more or less prototypical assertions. A standard assertion, for which it is true in particular that it intends to make p part of the cg, easily adheres to the two general trends in discourse (see above)

19 I am grateful to the reviewers for bringing into discussion these aspects discussed up to this point in this section and, thereby, giving me the occasion to comment on them. 20 I believe another aspect related to my claim which one should put an effort in investigating, is how the contexts in which I find the reversed order, are genuine contexts for this order. One cannot exclude that its prominence is due to the type of data looked at. Of course, ja doch can occur in more contexts than subjectivised ones, but, nonetheless, these are probably a domain which enhance the use of MPs in general. I think, this is an aspect which is often missed out when making out domains in which particular phenomena are claimed to be found. However, it is difficult to provide evidence for this question, especially when the data is from various sources.

236 

 Sonja Müller

based on which I motivate the ja doch-order. Assertions whose topmost aim it is not to increase the cg with the proposition they express also allow the order doch ja. This order is marked in the sense that it is restricted to utterance types this communicative intention can be attributed to. This applies to the three subjectivised contexts that I assume as the speaker’s evaluation of the issue takes centre stage in these utterances. However, turning back to the superordinate principle (i.e. the two general discoursive requirements) is always possible which is why the ja doch-order – which mirrors this principle in my view – can occur. This captures the observation that doch ja can always be substituted by ja doch, but that the reversed replacement is not always possible. This paper has investigated a particular MP-combination and several points for discussion show up. A question that arises is what the analysis can contribute to the larger picture. According to Thurmair (1989: 280), there are 171 possible combinations of two MPs. I do not believe that every combination has to be explained in the same way. MPs have come into existence at different points in time and have undergone different developments. Against this background, it is highly plausible that the combinations can be motivated very differently, too. Looking at existing accounts (cf. footnote 4), they come from different areas of grammatical description such as phonology, semantics, information structure as well as from a diachronic perspective. I absolutely do not consider this a disadvantage. However, I believe that this type of explanation I am proposing is transferable to other combinations for which other independently formulated semanticpragmatic principles are relevant then (cf. Müller 2016, to appear, on halt eben and eben halt). Indeed, I believe that a fairly general assumption follows from my considerations: It is known that the MP ja in the unmarked case marks the left margin of a combination, it usually precedes all other particles in sequences. As it is restricted to assertions and codes the information as cg-contents, it is more than plausible that this superior goal of discourse gets coded early. The MP doch addresses the topic of the conversation and should also be introduced early (unless even superordinate goals are expressed such as being part of the cg). This seems to be borne out as doch in the unmarked case precedes other particles in combinations of two in assertions (doch auch, doch eben, doch halt, doch schon, doch wohl, doch eh, doch sowieso, doch einfach) as well as in other illocutionary types (doch bloß, doch nur, doch mal, doch ruhig, doch nicht etwa). An exception seems to be denn doch in assertions. I am referring to two concepts, cf. (45), which are undoubtedly involved in discourse organisation, in order to derive the order of ja and doch, which serve discourse purposes. If my general idea can be applied to other combinations (what I am optimistic about), my account makes predictions of the sort above.



Combining ja and doch: A case of discourse structural iconicity 

 237

I also believe that one should make further investigations on the reversed orders. For many MP-combinations, it is not difficult to find both orders, cf. (63) and (64). One should use the sources that new corpora (such as the DECOWcorpus) provide and which were not available when other pieces of work on MP-combinations were written. (63) a. Auwei was ist bloß nur aus den Ruhrbaronen Oh-oh what is MP1 MP2 of the Ruhr-barons geworden? become ‘Oh, oh, what on earth has happened to the Ruhr barons?’ (DECOW2012-06X: 709617545) b. was könnte das nur bloß gewesen sein? *grübel, grübel* what could it MP2 MP1 been has *think, think* ‘What on earth could it have been?’ (DECOW2012-03X: 622063131) (Müller 2014b: 13f.) (64) a. Und sorry, hört langsam mal auf alle Namen mit And sorry stop MP1 MP2 up all names with „glänzend und hell“ zu übersetzen.   shiny and bright to translate ‘And sorry, it’s high time to stop translating all names with “shiny and bright”.’ (wikipedia-discussion-Arash; mythology) b. Kommen Sie mal langsam zum Haushalt! Come you MP2 MP1 to.the budget ‘It’s high time to debate on the budget!’ (Protocol of the session of the state parliament Niedersachsen, 22.12.2010) In my opinion, both elements can perfectly be understood as particles in (63) and (64). And one should put an effort on investigating the relation between the two sequences.

References Abraham, Werner. 1995. Wieso stehen nicht alle Modalpartikel in allen Satzformen? Die Nullhypothese. Deutsche Sprache 23(2). 124–146. Blühdorn, Hardarik. 2006. Kausale Satzverknüpfungen im Deutschen. Pandaemonium Germanicum. Revista de Estudos Germanísticos 10. 253–282. Brandt, Margareta & Inger Rosengren. 1992. Zur Illokutionsstruktur von Texten. Zeitschrift für Literaturwissenschaft und Linguistik 86. 9–51.

238 

 Sonja Müller

Coniglio, Marco. 2007. Deutsche Modalpartikeln: ein Vorschlag zu ihrer syntaktischen Analyse. In Eva-Maria Thüne & Franca Ortu (eds.), Gesprochene Sprache und Partikeln, 103–113. Frankfurt/M.: Peter Lang. Coniglio, Marco. 2011. Die Syntax der deutschen Modalpartikeln: Ihre Distribution und Lizenzierung in Haupt- und Nebensätzen. Berlin: Akademie Verlag. Croft, William. 1995. Modern syntactic typology. In Masayoshi Shibatani & Theodora Bynon (eds.), Approaches to language typology, 85–144. Oxford: Clarendon Press. Burkhardt, Armin. 1995. Zur Übersetzbarkeit von Abtönungspartikeln. Am Beispiel von Hofmannsthals ‘Der Schwierige’. Zeitschrift für germanistische Linguistik 23(2). 172–201. Dahl, Johannes. 1988. Die Abtönungspartikeln im Deutschen: Ausdrucksmittel für Sprechereinstellungen mit einem kontrastiven Teil deutsch-serbokroatisch. Heidelberg: Groos. De Vriendt, Sera, Willy Vandeweghe & Piet van de Craen. 1991. Combinatorial aspects of modal particles in Dutch. Multilingua 10(1–2). 43–59. Diewald, Gabriele. 1997. Grammatikalisierung. Eine Einführung in Sein und Werden grammatischer Formen. Tübingen: Niemeyer. Diewald, Gabriele. 1999. Die dialogische Bedeutungskomponente von Modalpartikeln. In Bernd Naumann (ed.), Dialogue analysis and the mass media. Proceedings of the international conference, Erlangen, April 2–3, 1998, 187–199. Tübingen: Niemeyer. Diewald, Gabriele. 2006. Discourse particles and modal particles as grammatical elements. In Kerstin Fischer (ed.), Approaches to discourse particles, 403–426. Leiden: Brill. Diewald, Gabriele. 2007. Abtönungspartikel. In Ludger Hoffmann (ed.), Handbuch der deutschen Wortarten, 117–141. Berlin: de Gruyter. Diewald, Gabriele & Kerstin Fischer. 1998. Zur diskursiven und modalen Funktion der Partikeln aber, auch, doch und ja in Instruktionsdialogen. Linguistica 38. 75–99. Diewald, Gabriele & Marijana Kresić. 2010. Ein übereinzelsprachliches kontrastives Beschreibungsmodell für Partikelbedeutungen. Linguistik Online 44(4). 5–18. https://bop.unibe. ch/linguistik-online/article/view/400/631 (accessed 7 June 2016). Dik, Simon C. 1997. The theory of functional grammar. Part 2: Complex and derived constructions. Ed. by Kees Hengeveld. Berlin: Mouton de Gruyter. Doherty, Monika. 1985. Epistemische Bedeutung. Berlin: Akademie Verlag. Doherty, Monika. 1987. Epistemic Meaning. Berlin: Springer. Eroms, Hans-Werner. 2006. Satzadverbien und Diskurspartikeln. In Vilmos Ágel, Ludwig Eichinger, Hans-Werner Eroms, Peter Hellwig, Hans Jürgen Heringer & Henning Lobin (eds.), Dependency and valency – Dependenz und Valenz. Ein internationales Handbuch der zeitgenössischen Forschung – An international handbook of contemporary research, vol. 2, 1017–1036. Berlin: Walter de Gruyter. Egg, Markus. 2013. Discourse particles, common ground, and felicity conditions. In Daniel Gutzmann & Hans-Martin Gärtner (eds.), Beyond expressives. Explorations in use-conditional meaning, 125–149. Leiden: Brill. Farkas, Donka & Kim Bruce. 2010. On reacting to assertions and polar questions. Journal of Semantics 27(1). 81–118. Fischer, Olga. 1999. On the role played by iconicity in grammaticalisation processes. In Max Nänny & Olga Fischer (eds.), Form miming meaning. Iconicity in language and literature, 345–374. Amsterdam: Benjamins. Giacalone Ramat, Anna. 1995. Iconicity in grammaticalization processes. In Raffaele Simone (ed.), Iconicity in language, 119–139. Amsterdam: Benjamins. Givón, Talmy. 1991. Isomorphism in the grammatical code: Cognitive and biological considerations. Studies in Language 15(1). 85–114.



Combining ja and doch: A case of discourse structural iconicity 

 239

Haiman, John. 1980. The iconicity of grammar: Isomorphism and motivation. Language 56(3). 515–540. Helbig, Gerhard & Werner Kötz. 1981. Die Partikeln. Leipzig: Langenscheidt Verlag Enzyklopädie. Ickler, Theodor. 1994. Zur Bedeutung der sogenannten Modalpartikeln. Sprachwissenschaft 19. 374–404. Jacobs, Joachim. 1991. On the semantics of modal particles. In Werner Abraham (ed.), Discourse particles, 141–162. Amsterdam: Benjamins. Karagjosova, Elena. 2003. Modal particles and the common ground: Meaning and functions of German ja, doch, eben/halt und auch. In Peter Kühnlein, Hannes Rieser & Henk Zeevat (eds.), Perspectives on dialogue in the new millenium, 335–349. Amsterdam: Benjamins. Küper, Christoph. 1992. Zum sprechaktbezogenen Gebrauch der Kausalverknüpfer denn und weil: Grammatisch-pragmatische Interrelationen. Linguistische Berichte 92. 15–30. Lindner. Katrin. 1991. ‘Wir sind ja doch alte Bekannte’. The use of German ja and doch as modal particles. In Werner Abraham (ed.), Discourse particles, 168–201. Amsterdam: Benjamins. Masi, Stefania. 1996. Deutsche Modalpartikeln und ihre Entsprechungen im Italienischen. Äquivalente für ‘doch’, ‘ja’, ‘denn’, ‘schon’ und ‘wohl’. Franfurt a.M.: Lang. Meibauer, Jörg. 1994. Modaler Kontrast und konzeptuelle Verschiebung. Studien zur Syntax und Semantik deutscher Modalpartikeln. Tübingen: Niemeyer. Motsch, Wolfgang. 1987. Zur Illokutionsstruktur von Feststellungstexten. Zeitschrift für Phonetik, Sprachwissenschaft und Kommunikationsforschung 40. 45–67. Müller, Sonja. 2014a. Zur Anordnung der Modalpartikeln ja und doch: (In)stabile Kontexte und (non)kanonische Assertionen. Linguistische Berichte 238. 165–208. Müller, Sonja. 2014b. Modalpartikeln. Heidelberg: Winter. Müller, Sonja. 2016. Halt eben vs. eben halt – Dialekt, Satzmodus, Rhythmus oder Interpretation? Sprachwissenschaft 41(2). 139–184. Müller, Sonja. To appear. Normalität im Diskurs – Implikationsverstärkung in (halt eben-/eben halt) Assertionen. In Linguistische Berichte (Special issue Normalität in der Sprache, ed. by Franz d’Avis & Horst Lohnstein). Newmeyer, Frederick J. 1992. Iconicity and Generative Grammar. Language 68(4). 756–796. Ormelius-Sandblom, Elisabet. 1997. Die Modalpartikeln ‘ja’, ‘doch’ und ‘schon’. Zur ihrer Syntax, Semantik und Pragmatik. Stockholm: Almqvist & Wiksell. Pittner, Karin. 2007. Dialog in der Grammatik: Doch in Kausalsätzen mit Verberststellung. In Sandra Döring & Jochen Geilfuß-Wolfgang (eds.), Von der Pragmatik zur Grammatik, 39–56. Leipzig: Universitätsverlag. Pittner, Karin. 2011. Subsidiäre Begründungen. In Gisella Ferraresi (ed.), Konnektoren im Deutschen und im Sprachvergleich: Beschreibung und grammatische Analyse, 157–182. Tübingen: Narr. Plank, Frans. 1979. Ikonisierung und De-Ikonisierung als Prinzipien des Sprachwandels. Sprachwissenschaft 4. 121–158. Pusch, Claus D. 2001. Ikonizität. In Martin Haspelmath, Ekkehard König, Wulf Oesterreicher & Wolfgang Raible (eds.), Language typology and language universals – Sprachtypologie und sprachliche Universalien, vol. 1, 369–384. Berlin: Walter de Gruyter. Repp, Sophie. 2013. Common ground management: Modal particles, illocutionary negation and VERUM. In Daniel Gutzmann & Hans-Martin Gärtner (eds.), Beyond expressives. Explorations in use-conditional meaning, 231–274. Leiden: Brill. Rinas, Karsten. 2006. Die Abtönungspartikeln ‘doch’ und ‘ja’. Semantik, Idiomatisierung, Kombinationen, tschechische Äquivalente. Frankfurt/M.: Peter Lang.

240 

 Sonja Müller

Rinas, Karsten. 2007. Abtönungspartikel-Kombinationen und Skopus. Sprachwissenschaft 32(4). 407–452. Schäfer, Roland & Felix Bildhauer. 2012. Building large corpora from the web using a new efficient tool chain. In Nicoletta Calzolari et al. (eds.), Proceedings of the Eighth International Conference on Language Resources and Evaluation (LREC’12), 486–493. Istanbul, Turkey: European Language Resources Association. http://www.lrec-conf.org/ proceedings/lrec2012/pdf/834_Paper.pdf (accessed 7 June 2016). Schubiger, Maria. 1965. English intonation and German modal particles: A comparative study. Phonetica 12(2). 65–84. Struckmeier, Volker. 2014. Ja doch wohl C? Modal particles in German as C-related elements. Studia Linguistica 68(1). 16–48. Sweetser, Eve. 1990. From etymology to pragmatics: Metaphorical and cultural aspects of semantic structure. Cambridge: Cambridge University Press. Thurmair, Maria. 1989. Modalpartikeln und ihre Kombinationen. Tübingen: Niemeyer. Thurmair, Maria. 1991. ‘Kombinieren Sie doch nur ruhig auch mal Modalpartikeln!’. Combinatorial regulari­ties for modal particles and their use as an instrument of analysis. Multilingua 10(1–2). 19–42. Thurmair, Maria. 2013. Satztyp und Modalpartikeln. In Jörg Meibauer, Markus Steinbach & Hans Altmann (eds.), Satztypen des Deutschen, 627–651. Berlin: de Gruyter. Vismans, Roel. 1994. Modal particles in Dutch directives: A study in functional grammar. Amsterdam: IFOTT. Volodina, Anna. 2010. Sprechaktbezogene Kausalität. In Olga Souleimanova (ed.), Sprache und Kognition: Traditionelle und neue Ansätze. Akten des 40. Linguistischen Kolloquiums in Moskau 2005, 309–321. Frankfurt/M.: Peter Lang. Waltereit, Richard. 2001. Modal particles and their functional equivalents: A speech-acttheoretic approach. Journal of Pragmatics 33(9). 1391–1417.

Corpora/Tools COW Corpora from the Web, http://corporafromtheweb.org, a project by German Grammar Group at Freie Universität Berlin. DeReKo Das Deutsche Referenzkorpus DeReKo, http://www.ids-mannheim.de/kl/ projekte/korpora/, am Institut für Deutsche Sprache, Mannheim.

Werner Abraham

Discourse marker = discourse particle = thetical = modal particle? A futile comparison Data regarding three claims concerning the syntax of modal particles will be presented. The first claim is that modal particles (MPs) are a category of particles that is different from other discourse markers, and that they are found only in German, a few Germanic languages and possibly in Russian, but not in Romance languages and in English. The second claim states that MPs as illocutive operators are free grammatical morphemes. If this claim is correct, it debunks the attempts of entire generations of researchers to identify the word class of MPs and put them on a par with lexical discourse markers. The third claim further develops Höhle’s (1982), Jacobs’ (1992), and Klein’s (1998) arguments for a decomposition of Fin(iteness): i.e. that Fin should be understood to involve two independent components – agreement and truth valuation. The mediating empirical components are emphasized assertion and polarity accent (verum focus [VF]). An important conclusion will be that MPs and VF are intimately related before a semantic and pragmatic background.1

1 Introducing the topic The following ideas and lines of argumentation were inspired by Degand, Cornillie, and Pietrandea (2013) and Kaltenböck and Heine (2014) on discourse markers, “theticals”, and modal particles and, in particular, on problems of their identification in terms of categorial status. This is due to at least two facts. First,

1 The following abbreviations will be employed: Addr = addressee (hearer); CG = common ground; DM(s) = discourse marker(s); EMV = epistemic modal verb (reading); MP(s) = modal particle(s); Sp = speaker; VF = verum focus. Acknowledgements: The present version of this paper is the result of three insightful reviews and suggestions for format and style. Special thanks go to Jason Whitt (Nottingham) and to Josef Bayer (Konstanz), one of the volume editors, for his scrutinizing input to many essential details. Critical comments during the presentation in the Marburg Conference proved to be highly useful. Werner Abraham: Institut für Sprachwissenschaft, Universität Wien, Wien, e-mail: werner_ [email protected] DOI 10.1515/9783110497151-010

242 

 Werner Abraham

counter to the claim that the presentations in these works mirror the present state of the art on discourse particles and modal particles,2 only a section of the empirical and theoretical horizon is covered. Second, the present reaction is conceived within the context that Degand, Cornille, and Pietrandea (2013) and Kaltenböck and Heine (2014). The position represented by Degand, Cornille, and Pietrandea (2013) concerns primarily the question as to what grammatical category MPs could be ascribed to. Both Coniglio (2011) and Cardinaletti (2011) (unmentioned in Degand, Cornille, and Pietrandea [2013] and Kaltenböck and Heine [2014]) have characterized the class of MPs as “light adverbials”. Cardinaletti (2011) offers strong arguments against head status of MPs while identifying them as grammatically free (unbound) morphemes. The present discussion will claim, in line with Bayer and Obenauer (2011), that MPs are free grammatical morphemes with head status. It is this general claim that will lead us through several descriptive and theorizing sections with the aim of clarifying the following more specific questions and making the claims below the background of our considerations: –– MPs are categorially, and in terms of serialization, fundamentally different from all other discourse markers (DMs). –– Clarification is offered as to what extent are they really “modal”. –– It will be shown that MPs are intimately related to intonational (verum) focus. The function of accent focus on MPs will be discussed. Does emphasis have an effect on the categorial status of sentential MP or its identity? What is the relation between MP and its focused variant? –– What is behind the serialization of MPs? –– How MPs need to be seen as truth gradients. –– We shall explore to what extent the status of MP – counter to other lexical discourse markers – is contingent upon: topicality, categoriality (as opposed to theticity), the utterance to the effect that clausal categories other than VP, Aspect, and Inflection have to be involved for a successful MP-merge.

2 As much as they do not speak about tags, Kaltenböck and Heine (2014) do not address the class of MPs at all as the latter do not figure in their inventory of text categories in the first place (Kaltenböck and Heine 2014: 360). Nevertheless, their topic is discourse grammar and the categories that figure in it as a separate “domain” of speech are claimed to be cognitively privileged over clause grammar (Kaltenböck and Heine 2014: 361–362). However, what remains totally undiscussed is to what extent the inventory of thetic lexicals and clusters are amenable to some sort of text syntax. No reference is made to theticity in the language philosophical sense of the inventors of concept and terminology, Marty (1910) and Brentano (1956). Without a formal syntax, a general text grammar is hard to conceive. Before this background, Kaltenböck and Heine’s (2014) claim that clausal grammar and the lexical inventory of “theticals” are competing grammatical domains seems to be a misconception.



Discourse marker = discourse particle = thetical = modal particle?  

 243

Just as Aikhenvald (2012, 2015) and de Haan (2001, 2011) have made a strong case for evidentiality as a purely grammatical category, stating that morphological evidentiality or evidentiality proper must be kept distinct from lexical evidential expressions, we posit a similar range of necessary conditions for MPs as they appear in German and Dutch: –– MPs are grammatical (as opposed to lexical) elements. –– MPs have speaker (Sp) deixis, and possibly also addressee (Addr) deixis, as their primary meanings. –– MPs are part of an obligatory system and paradigm to mark their independent syntactic (and semantic) status (illocutionary autonomy). –– The Sp’s MP-producing strategy affords a considerably higher degree of complexity and social awareness than any lexical alliteration. By the same token, the Addr’s competence at deciphering the exact attitudinal import of MPs presupposes a similar degree of grammatical awareness and socialpsychological flexibility (perspectivization with respect to world access) and adaptability. Every language has some way of referring to the source of information, albeit not every language has grammatical evidentiality. It is equally evident that not every language has MPs – in terms of the expressive shortness, derivation from lexicals with different meanings (giving credit to their polyfunctionality), clear constraints on the type of attitudinal function (Sp deixis), syntactic position and structural status, etc. In other words, just as much as the lexeme allege in the alleged killer of X is tangential to the discussion about grammatical evidentials, great care is taken in the recent discussion about MPs to exclude lexical adverbial epistemics (such as, e.g., vermutlich in Das ist vermutlich so ‘This may presumably be so’ in contrast to Das ist wohl so ‘This may well be so’; however, see Abraham [2012a] on the categorial status of wohl). The respective syntactic slots for MPs are identified between the so-called Wackernagel position and to the left of vP. Strong intonation is argued to target either polarity emphasis (VF) or the most deeply embedded category in VP, by default the latter. With regard to modality as projected by MPs, it is shown that these features are projected rather low in certain other languages, i.e. immediately to the left of the predicate phrase. This structural aspect will be taken over for German MPs despite the fact that German is OV, while the languages compared are of the VO type. Based on evidence from imperative, interrogative, and modal constructions, this low structural position will be shown to host many other illocutionary features. Parallels are drawn from Spanish, Italian, and Vietnamese, the latter a topic prominent language, for the very reason that German MPs relate effectively to topicality. Considering the strong Universalist tenet that the

244 

 Werner Abraham

appearance of a phenomenon investigated and explained rigidly and profoundly in analytic terms qualifies this phenomenon and its deriving processes as universal, we suggest that the evidence drawn from languages other than OV-German (i.e. VO-French, VO-Italian, VO-Spanish, and VO-Vietnamese) support our initial claim that indeed German MPs are free grammatical morphemes with their lower positioned illocutive potential to eventually raise to the highest position taking scope over the entire sentence. As such they are a far cry from discourse oriented lexicals (termed as thetics, markers, adverbials). It will be the primary goal of the following discussion to argue to keep MPs separate not only from synonymous lexemes, but also from other syntactic categories (among which (light) adverbial status; see Coniglio [2011], Cardinaletti [2011]). These topics will be dealt with in the present discussion under the following headings: Modality at issue (Section 2); MP-source categories and their underspecified MP-results (Section 3); Contrastive accentuation, verum focus, tense, and truth valuation (Section 4); Contrast focus on German MP (Section 5); Free grammatical MPs in German and in other languages (Section 6); Verum focus and MP-readings (Section 7); MPs and epistemic vs. root modal scoping (Section 8); Indismissable topic reference on MPs (Section 9); Word order options for MPs under finiteness and non-finiteness (Section 10); Serialization options and constraints between MPs (Section 11).

2 Modality at issue We will discuss modality within a narrow conceptualization. Unlike mood (Latin-German Modus), which takes us into formal verbal paradigms of realis and irrealis – and thereby into scalar truth appreciations of Sp or viewer, in Bühler’s ([1934] 1999) sense of the origo perspective (see Abraham 2012a, 2012b) –, modality makes incumbent the Sp and observer so that a specific possible world becomes accessible. In addition, modality introduces a common ground (CG) level to be decided upon both by Sp and Addr: Sp puts the utterance of p up for negotiation in a variety of speech acts, for example, a challenge of somebody else’s assertion. Epistemic readings of modal verbs (EMV) achieve the former (i.e. modality only at the Sp’s disposal), but not the latter. Only German and Dutch modal MP make available a more complex type of modality to the latter. An important criterion of the expressive trigger of modality adds an extra qualification to both Sp’s (intransitive) and Sp-Addr’s (transitive) modality: They are not triggered lexically (by adverbials), but as free grammatical morphemes – i.e. in purely grammatical terms, which, in line with Coniglio (2011) and Bayer and Obenauer (2011),



Discourse marker = discourse particle = thetical = modal particle?  

 245

have the categorical status of adverbials albeit as functional head. Notice that it follows from this functional status that MPs are subject to syntactic distributional constraints that adverbials are not (e.g. as regards the appearance in the prefield/ SpecCP position in German and Dutch, where MPs are disallowed for the very fact that free grammatical items cannot be accentuated, and, consequently cannot be fronted to SpecCP, as any adverbial can). Let us jump right into illustrations. Take the German MP ja (‘yes’ in its non-MP, pregrammaticalized use) in the following sentences: (1) Er hat sie ja belauscht he has her MP overheard ‘He overheard her.’

utterance status: argument in support of a previous statement of Sp, e.g. “He did not behave in style.” (e.g. chancelor Merkel in the NSA spy case)

(2) Er ist ja immer beschäftigt. he is MP always busy ‘He is always busy.’

utterance status: explanation of Sp’s own experience with person Er, who could not be approached on a question to be solved.

Following the thesis of Foreign Consciousness Alignment (Abraham 2012a, 2012b), according to which the Addr of (1) is invited by the Sp to further negotiate the truth or appropriateness of the Sp-supposed CG, we may assume that something like the following exchange of thought takes place between the communicating parties. Note that the ja-sentence in (2) yields an analytic result different from that of (1). (3)–(6) appear to function as (at least part of) sincerity conditions for the use of ja in the respective illustrations: (3) With reference to (1), Sp assumes that Addr is prepared to see some appropriateness in Sp’s claim of p. Therefore he invites Addr by using ja to seek further CG with Sp’s claim of p to the effect that the utterance of p is warranted (or is not too much unwarranted). (4) (2) is dialogical as well, although maybe given the different utterance status, not as explicitly as (1). Let us assume that in both (1) and (2) we are dealing with portions of a dialogue between Sp and Addr, where in (2) the Addr is also the Sp (in terms of an inner dialogue). On the latter assumption, the analysis (5) with respect to (2) is close to (3). (5) Sp gives expression to his understanding why er ‘he’ (subject in [2]) has not reacted – Sp assumes that (Sp’s inner) Addr is prepared to see appropriate Sp’s claim of p. Therefore, by virtue of the MP ja, Sp invites Addr to seek so much CG with Sp’s claim of p.

246 

 Werner Abraham

Notice that, except for the inceptive line in (4), the analyses (3) and (4) are equivalent. We thus have a somewhat simplified analysis in (6) for both (1) and (2) and instead of (3)–(5): (6) Sp believes that Addr is prepared to accept p, or at least consider part of p as acceptable – which is near-equivalent to saying or even further simplified , which, in turn, may even turn out as – i.e. the belief-act recurring cyclically (and without natural termination) between Sp and Addr. Let us consider other MPs in German to see which other CG or Foreign Consciousness Alignment analyses they require. See (7): (7)  Deutsch ist eben // aber // ja // auch // schon // vielleicht // eigentlich // wohl German is MP  MP … schwer. difficult Diewald (2007: 34) has the following paraphrases for the MP-clauses in (7) – the English lexicals in parentheses are lexical translations usually of the categorial status of an interjection, coordinator, adverbial, or temporal focus particle. English, of course, does without the category of MP): (8) ‘German is difficult – I iterate this (eben) // in contrast to the opposite assumption (aber ‘but, however’) // we all know (ja ‘yes’) // this and other things hold (auch ‘also’) // admittedly (schon ‘already’).’ For vielleicht/eigentlich/wohl3 we might add the following paraphrases as modalillocutive modifiers of p in (9): (9) ‘German is difficult – I emphasize this statement p (vielleicht ‘perhaps’) // counter arguing against doubts about p (eigentlich ‘in fact’) // presumably (wohl ‘well’).’ We shall return to such paraphrases and will claim that they fall short of meaningful substitutes in terms of several criteria. A few more clarifying words appear to be in order at this point.

3 Not contained in Diewald’s (2007) list; for eigentlich see Oppenrieder and Thurmair (1989); for a possible disqualification of wohl as an MP see Abraham (2012a, 2012b).



Discourse marker = discourse particle = thetical = modal particle?  

 247

3 MP-source categories and their underspecified MP-results Let us address the question how to relate the illocutive MP-function to the lexical source meaning and category – a question that has never been solved in a convincing way. This question may be identical to the grammaticalizing path of the MP since both categories still exist side by side. Let us look at aber ‘but’ and doch (etymologically related to though). The MP-function of aber as in (6) above derives from the lexical’s categorial status as the coordinator aber ‘but, however’. Note that coordinators are indisputably to the left of the German prefield/SpecCP – thus, not CP-integrated syntactically, whereas MPs are located in the middle field of the German clause, definitely above VP, but below IP and the Wackernagel position, and consequently, CP-integrated. The semantics of the coordinator is sketched in (10a), (10b): (10) a. Sie ist klein, aber/doch (sie ist) dumm. ‘She is small, but (she is) dumb.’ b. Es ist Sommer, aber/doch es schneit. ‘It is summer, but it is snowing.’ a.’ p, {aber, doch} q =def (p → ¬q) or, depending on how plausible the relation between klein and dumm is considered by others/Addr: b.’ p, {aber, doch} q =def (∃r) (p → r) ∧ (q → ¬r) (Abraham 1975) Uttering (10a)/(10b), Sp simply lets us know that in Sp’s perspective from the property of being klein/small follows the opposite of being dumm/dumb. The same holds as a relation between being in summer and be snowing: one excludes the other. (10b) involves an intermediate step: from being summer/p follows not being winter/r, while winter/¬r is entailed by be snowing/q. Such intermediate positions seem to be required if the but-statement is too audacious as in (11) since there is no obvious way of relating the protasis/p to the apodosis/q semantically. Notice again that there is nothing ontologically solid on such uses of aber/doch. Nothing with respect to reality or truth validity is presupposed from such usage. (11) Er ist mutig, schaut aber gerne zu viel ins Glas. ‘He is courageous, but drinks too much.’ Similarly, even if (11) should sound too audacious or uninterpretable for certain Addrs, the conclusion to be drawn from (11) is that Sp’s utterance runs under the analytic premises of (10a’) or (10b’). In other words, interpreting (11) presupposes

248 

 Werner Abraham

an analytic understanding of (10a’)/(10b’) or change of the accessible world on the side of Addr such that (10a)/(10b) are accommodated. Let us now pursue the question concerning what aber preserves from (11) when used as an MP, and what it loses or suppresses, and which parameters cause such a change. Note, first, that for the MP aber single sentences might not be taken to be the norm: (12) a. Das ist aber überraschend! this is MP surprising b. Der hat das doch getan ohne mit he has that MP done without with der Wimper zu zucken. the eyelid to bat ‘He did that without batting an eye.’

emphatic surprise

assertive surprise

Although MPs in (12a)/(12b) are single-utterance operators, they react to preceding utterances or events that have the quality of predications expressing something unexpected: if q is what Sp expected, but p happened, then Sp sees reason to make statements like (12a)/(12b), with p relating to q in accordance with (10a’)/ (10b’). We may thus say that while coordinators operate on two successive sentences, MPs operate on just one while presupposing an event corresponding to the protasis of a coordination of two sentences. The syntactic position plays a crucial role. Coordinators are located and execute their operating function outside (but in front of) the single CP-apodosis. In contrast, MPs, like other adverbials, are located above vP, more accurately between the thematic and the rhematic middle fields. See (13) for this (T = tense-agreement, MF = middle field, THMF = thematic middle field, RHMF = rhematic middle field):4 (13) Der [T’ hat [MF [THMF < aber/doch> seinem Bruder] < aber/doch> [RHMF einen he.dem has  MP his.dat brother MP a Schubs]] * [vP [V’ gegeben]]], dass … shove MP given that

4 The following abbreviations are used in glossing examples: acc = accusative; ass = assertion (marker); cl = clitic; dat = dative; dem = demonstrative; excl = exclamative (marker); imp = imperative mood; neg = negation; prn = pronoun; refl = reflexive (pronoun); sg = singular; q-part = question particle; 1/2 = first/second person.



Discourse marker = discourse particle = thetical = modal particle?  

 249

In what follows we build on Klein’s (1998) argument for the decomposition of Finiteness into Tense and Assertion (i.e. truth valuation). Furthermore, it is worth pointing out that in other languages Assertion (as much as truth valuation) is projected rather low (immediately to the left of vP, the predicate phrase, below all other functional categories (except for Mood). It has been argued that sentential negation, and other polarity elements, occupy a separate, and higher, structural position than those of the assertion phrase (Rizzi 1990, 1997). The evidence for this claim is derived from imperative, interrogative and modal constructions. Thus, this low structural position hosts all other illocutionary (force) features (as they indeed do in languages like Vietnamese, according to Duffield [2013] ‒ notwithstanding the evidence from Romance and Germanic languages, which seem to support a much higher position for such features ‒ namely, ForceP ‒ on the left periphery). The results of these investigations thus resurrect the idea originally proposed in Chomsky (1957) that features of (emphatic) assertion as well as other features of illocutionary force are initially associated with a low syntactic position: see also Cormack and Smith (2002). As to the theoretical implication of this, the appearance of illocutionary features in the (extended) C-domain must be due to overt movement of the lexical elements hosting these features (i.e., lexical elements that move to CP are attracted by illocution-related features of the functional C head), rather than to basegeneration (see also Duffield 2013 on English; see Abraham 2015 for German MPs). See (14)–(15) for what clausal accentuation in imperative constructions yield in terms of IP-positions in English (Duffield 2013: 266) and in German dialects: (14) a. Do sit you down there! (Belfast) English Sitz ( du di) nieda duat sit.imp  you 2sg.refl down there Tua ( di) niedasitzn!Austrian-Bavarian do.imp 2sg.refl down.sit Tua ( du) derta dori hocka! High Alemannic do.imp  you there down sit ‘Do sit down there!’ b. *You sit down there! Belfast English Du sitz di nieda duat!Austrian-Bavarian you sit.imp 2sg.refl down there c. Du – sitz di nieda duat!Austrian-Bavarian ‘You – sit yourself down there!’ *Du tua derta dori hocka! High Alemannic You do.imp there down sit

250 

 Werner Abraham

(15) a. Ask not for whom the bell tolls. John Donne Frag nicht, wem die Stunde schlägt. Modern Standard German Tu (du) net fragen, wea (dass) des isAustrian-Bavarian do.imp  you not ask who that this is Tua Du netta fraga, wer (dass) des isch!  High Alemannic do.imp you not ask who that this is b. Fear not the future, for it doesn’t exist and never shall. Fürchte nicht die Zukunft, denn sie existiert nicht.  Modern Standard German Tua (du) di net fiachtn!  Austrian-Bavarian do.imp you 2sg.refl not fear Note what the illustrations are to show: the optimal position for the imperative auxiliary, tun/do, is Comp, with SpecCP bracketed out by preference. The illustrations for English and (dialectal) German are complementary to the extent that where English standardly inserts the auxiliary do to avoid the lexical predicate from raising to IP. English lexical forms are possible as well in the non-standard (which, as Duffield [2013] claims, is supported by L1 evidence). By contrast, Standard German never uses the auxiliary tun in imperatives (except in rare cases of emphatic accentuation such as Tu mir das nicht anrühren! ‘do (to) me that not touch’), while it is normal and idiomatic in the Southern dialects, but not confined to them in German. This shows that (emphatic) Assertion is best served high in CP, whereas mood and modality “lexicals” are located lower down in the middle field just above vP. MPs are among the latter, although their scoping force extends from the highest possible position in CP (or, better, in its Rizzi-extension; cf. Abraham [2015]). Ass(ert)P is relabeled as Mood Phrase in (16) below, reflecting its more comprehensive feature set {interrogative, declarative, imperative} as sentence mode. Asp1P, the higher aspect phrase, refers to Perfective aspect (derived by raising to the definiteness in the German middle field, whereas, in the cases of Germanic Aktionsart, from within VP), while Asp2P is the unmarked Imperfective aspect (cf. Abraham 2013: 587–635, for German and Slavic):5

5 Asp1P (in our case, perfectivity) and definite-specific DP in the left middle field of German clause structure have the mereological features of [–additivity, +boundedness] in common, while Asp2P (imperfectivity) and indefinite DPs share the characteristics [+additivity, –boundedness]. These features are responsible for the probing processes restricting overt movement. If topic is a structural hypercharacteristic for definiteness and topic-about, it follows that MPs are positioned (low) in the left middle field (above vP) of German, as this clause structural field is responsible for categories reserved for dialogical usage as MPs typically are. Note with respect to (16) that IP is different from TP in German with IP scoping over TP for obvious reasons.



Discourse marker = discourse particle = thetical = modal particle?  

 251

(16) [ TP [ Emph/AssP [ Asp1P [ Asp2P [ MoodP [ ModalityP ]]]]]] The essential insight (16) provides is that the syntactic scope position of Assertion features derive from below in the clausal structure raised to TP and then IP above TP.6 This echoes our previous claim that MPs are free grammatical categories serving the illocutive notions of Sp (and Addr) deixis with maximal scope over the clausal structure (Abraham 2015). Ass(ertion) eventually lands higher up in the vicinity of IP drawing on the fact that bound grammatical affixes hook up to Comp0 to attach to the required lexical predicate – a process that free grammatical categories need not share.

4 Contrastive accentuation, verum focus, tense, and truth valuation On our proposal to consider MPs as free grammatical morphemes, we come to terms with where exactly modality is projected syntactically. When MPs appear in the German clause, truth-value and illocutionary force are projected relatively low (below TP) in clausal phrase structure. Duffield (2013) suggests that this low position is more highly articulated than previously supposed, involving at least two dissociable functional categories (NegP) and Assertion as belonging to Mood. Going back to our distinction of Mood and Modality (Abraham 2012a, 2012b), we connect truth valuation, and thus Assertion, with Modality. Drawing on Höhle’s (1982, 1988, 1992) notion of verum focus (VF), Klein (1998, 2006) decomposes finiteness into two more primitive concepts: Tense and Assertion. A key piece of empirical evidence for this de-composition involves ambiguities found in contexts of clausal contrastive intonation. Klein observes that in declarative sentences like (17a) containing auxiliaries other than English do, emphasis on the auxiliary (see also Lohnstein 2012, among others) triggers two distinct interpretations: either, as in (17b) below, it is interpreted as bearing on the correction of TENSE on the prior utterance in a dialogue; alternatively, where the accentuated correction is on tense, as in (17c), what is emphasized is understood as what Höhle and Klein have termed the Assertion Validity of the utterance. Notice that negation bears on the conventional or

6 Clearly, as one reviewer points out, this concords with the assumption that truth values are only accessible vis-a-vis certain time/world indices, and illocutionary force cannot be inspected without inspecting properties of the C domain.

252 

 Werner Abraham

conversational informational evidence in that it always appears under pragmatic accentuation: (17) a. Das Buch WAR/LAG auf dem Tisch. emphasis on tensed assertion as ‘The book WAS on the table.’ contrasted to unmarked present or generic tense b. Das Buch ist/liegt auf dem Tisch. – Nein, das Buch WAR/LAG auf dem Tisch. ‘The book is/lies on the table. – No, the book WAS on the table.’ c. Das Buch war/lag NICHT auf dem Tisch. Falsch, das Buch IST/LIEGT auf dem Tisch. ‘The book was NOT on the table. No, that’s incorrect, the book IS on the table.’ VF thus triggers alternatives to (truth) Assertion in terms of tense and mood or modality (realis-irrealis as well as illocutive modality). We might want to describe the difference between (17b) and (17c) in the following sense: The finite element WAR/LAG // WAS/LAY carries the two distinct meaning components: 1. as a tense component, it marks past, in contrast to present or future; 2. as an assertion component, it marks the assertive valuation of the claim − the fact that the situation described by the utterance indeed holds, in contrast to the opposite claim as in the first clause in (17c). As an alternative, the comparison of (17a) and (17c) appears to allow emphasis not on the tensed assertion, but on the assertion itself.7 It was noted by Höhle (1982, 1988, 1992) (and likewise by Klein 1998, 2006; Lohnstein 2012, and others) that, contrary to German, this sense cannot be borne by the lexical predicates in languages such as English, but only by auxiliaries; consequently, in emphatic contexts in English where no other tense auxiliary is available, do must be inserted. This is exemplified by the contrasts in (18) below: (18) a. John LIKED Mary (but he didn’t LOVE her). ?John doesn’t LOVE Mary, but he LOVED her (then). b. c. John DIDN’T love Mary. Yes (No), he DID love Mary. (#Yes/No, he LOVED Mary). (Duffield 2013: Exx. [4a]–[4c]) As Duffield (2013) analyzes it, emphatic stress on the main verb in (18a) can be used to contrast lexical meaning (like vs. love) or, marginally, to contrast tense, as in (18b). However, to contrast assertion validity – to indicate that the situation

7 As was pointed out by one reviewer.



Discourse marker = discourse particle = thetical = modal particle?  

 253

described by the predicate in fact obtained – do/did insertion is required, as in (18c). The use of a stressed lexical verb is unacceptable in this context, as (18b) shows. It follows from the alternations in (18) that a key function of the do-support in Modern Standard English – in addition to, but quite independent of, its morphological role in hosting tense affixes – is to express truth function by assertion valuation. This conclusion has immediate theoretical implications: The structural implication of Höhle’s and Klein’s proposal is that Assertion (Ass), like Tense, should be considered an autonomous functional category, projected externally to the predicate phrase. Notice that this restriction has been postulated quite some time ago for German on the evidence of the demise of the simple past in the dialects and regiolects of South (“Upper”) German: Comp (the “left verbal bracket” in more traditional terminology of German grammar writing) is reserved for the expression of finiteness and tense – and thus truth valuation – in grammatical terms only. The fact that lexical predicates fill this slot in addition to the auxiliaries sein, werden, and haben is due to the constraint of the lexical integrity requirement of grammatical items: “grammatical morphemes have no PF-representation unless hosted by a lexical morpheme” (Abraham 1999). Thus, the interim conclusion is that MPs may very well be attributed the status of free grammatical morphemes in that, on account of their certaintyuncertainty qualifying status, they are assertion and truth valuators. They appear in the vicinity of sentence adverbials, i.e. in the middle field. It remains to be seen whether MPs are grammatical morphemes also in terms of heads, and which scope relations they are subject to in relative linearization with true sentence adverbials. The following illustrations from Italian and French appear to support our syntactic conclusions about German verum emphasis. They feature parenthetic material which is emphatic and features the lexical parenthesis yes (or) no. As Poletto and Zanuttini (2011) claim, such “polarity particles sì/no-oui/non” are located in FinP, which is the lowest projection of the left periphery. Similarly in (19c), oui ou non appears to the right of the French tense auxiliary as (though the alternant in [19d] shows that preverbal placement is also acceptable): que oui/non. (19) a. Je crois I think.1sg that yes/no ‘I think so/not.’ [German: Ich glaube ja/nein.] (Poletto and Zanuttini 2011: ex. [28a]) b. Credo [FinP [ di [ sì/no.]] believe.1sg of yes/no ‘I think so/not.’ [German: Ich denke ja/nein.] (Poletto and Zanuttini 2011: Ex. [28b])

254 

 Werner Abraham

c. Est-ce que [TP tu as [ oui ou non [ sorti les q-part you have yes or no taken-out the poubelles]]]? garbage ‘Did you or did you not take out the garbage?’ [German: Hast du oder hast du nicht den Müll rausgetragen?] (Poletto and Zanuttini 2011: Ex. [29b]) d. Est-ce que oui ou non tu as sorti les poubelles? q-part yes or no you have taken-out the garbage ‘Did you or did you not take out the garbage?’ [German: Hast Du ja oder nein den Müll rausgetragen?] (Poletto and Zanuttini 2011: Ex. [29a]) Furthermore, according to Battlori and Hernanz (2011), an even lower functional category exemplifying two emphatic free grammatical particles is found in Catalan, i.e. bé vs. ben (also pas), see (20).8 ben enganyat. (20) a. En Joan m’ha the Joan cl.dat-has indeed deceived ‘John has deceived me indeed.’ (Battlori and Hernanz 2011: Ex. [34c]) b. Bé ha arribat tard la soprano. Indeed has arrived late the soprano ‘The soprano has indeed arrived late.’ [German: ABER JA/SICHER ist der Sopran spät da gewesen.] (Battlori and Hernanz 2011: Ex. [37a]) c. La soprano s’ha ben enfadat (quan li ho the soprano cl-has really got.angry when cl.dat cl.acc han dit) have said ‘The soprano really got angry when they told her.’ [German: Na, der Sopran ist JA zornig geworden (, als man ihm das vorwarf.)] (Battlori and Hernanz 2011: Ex. [39a])

8 Notice that it is just under the cartographic approach that (20c)/(20d) receive an appropriate explication under scoping relations and the tautness between conversational and conventional implicatures. The fact that items relating to polarity can also appear sentence-internally in other languages is taken as an argument that modal particles can relate to similar semantic properties. The validity of carrying these distributions over to German lies in the fact that modal particles interact with properties of these types. Notice that when the German wohl lit. ‘well’ is adduced as a link to German, its categorical status as an MP (cf. Zimmermann 2004) may be questioned as wohl, contra to MPs generally, can appear in SpecCP.



Discourse marker = discourse particle = thetical = modal particle?  

 255

d. Bé s’ha enfadat la soprano (quan li ho indeed CL-has got.angry the sopano when cl.dat cl.acc han dit) have said ‘But the soprano really got angry when they told her.’ [German: ABER JA/SICHER wurde der Sopran zornig, als sie ihm das sagten.] (Battlori and Hernanz 2011: Ex. [39b]) Such examples led Battlori and Hernanz (2011) to propose a structural distinction between “high EPM [emphatic polarity marker]” and “low EPM”. Specifically, their claim is that low EPMs are located in a functional projection FP between IP and VP, possibly above the vP field, as sketched in (21): (21) [CP … [FOCUSP EPMi [POLP ti [IP … [FP low EPM [vP …]]]]]] Note that, for one, these Catalan data appear to support our conclusion that emphatic assertion may be associated with more than one functional projection. Second, they make explicit the assumption that the lower of the two functional projections, low EPM, is positioned very low indeed, immediately above vP. This is exactly the position where we located MPs, arguing that this syntactic slot must  have something to do with Mood/ Modality and thus with Assertion and Agreement/IP. MPs are free grammatical elements. Their surface position is in the left middle field immediately left/on top of vP, which is indicative of the lowest syntactic mood-modality slot. This appears to be supported by a number of other investigations. With regard to the focused assertion itself, Duffield (2013) has shown that features to this effect are projected rather low in Vietnamese phrase structure, immediately to the left of the predicate phrase. Duffield further claims on the basis of evidence from imperative, interrogative, and modal constructions that this low structural position hosts many other illocutionary features in Vietnamese. The evidence usually brought forward for Romance and Germanic languages, which seem to support a much higher position for such features on the left periphery of the clause, may thus have to be reconsidered under novel distributional tests as for German in the present discussion.

5 Contrastive focus on German MP Let us reappraise what the last section yields for MPs. Take a declarative clause with contrastive focus/VF on what may function as a German MP, ja, i.e. JA. Let us try to decompose utterances like (22a) and (23a) into their constitutive parts (cf. Abraham 2010, 2012b: 7–8; Gutzmann 2010: 19–20). Then let us compare what the

256 

 Werner Abraham

pragmatic-semantic consequences are on unfocused ja as opposed to focused JA. Compare (22a)/(22b) and (23a)/(23b): (22) Declarative clause with ja: a. Paul will Maria ja einladen. Paul wants Maria MP invite b. [Sp presupposes/believes that Addr knows/believes]ASSERT that p [=Paul wants to invite Maria] and that [Addr knows p]ja. (23) Declarative clause with JA = ja + VF:9 a. Paul will Maria JA einladen.10 b. [Sp presupposes/believes that Addr knows/believes]ASSERT that p [=Paul wants to invite Maria] and that [Addr possibly knows/believes the ­opposite of p]ja and that [p should be accepted by Addr as CG]VERUM-FOCUS. In other words, Sp asserts with (22a) and (23a) that Paul wants to invite Maria and that Sp modifies this Assertion vis-à-vis Addr in such a way that, with (22b), he assumes shared presuppositions (= CG) with Addr on the appreciation of p, and in (23b), the Sp, by refuting Addr’s counter appreciation of p, insists on his prior claim p.11

9 Clearly, ja vs. JA in imperatives have a different pragmatic-semantics in Komm *ja/JA gleich nachhause ‘Do come MP home at once’ as there is no imperative with unfocused ja: cf. the syntactic analysis [Sp presupposes/believes that Addr knows/ believes]ASSERT that p [=Paul wants to invite Maria] and that [Addr possibly knows/believes the opposite of p]ja and that [Addr has no option but to accept p]VERUM-FOCUS. Comparison with (23b) yields that the constituting component in (23b), p should be accepted by Addr as CG, is replaced by the no-option differential. 10 Context: A: Paul should not invite Maria. B: Paul will Maria JA einladen. DOCH would be an alternative MP in this case. 11 JA does not mean ‘as you know’ (as its unstressed variant often does), but it turns against the Addr’s prior NEIN ‘no’, much in the sense of DOCH. One might ask what the independent semantic contribution of JA ‘apart from’ or ‘on top of’ has to do with the verum reading? Notice that stressed JA and VF on the predicate cannot co-appear: *Paul WILL Maria JA einladen. It is either focus on WILL or on JA. What this reveals is that VF cannot necessarily be equated with contrastive focus (CF): In one reading, VF links with a truth statement counter to another claim to this effect. By contrast, CF has narrower scope in that it functions as a claim to an alternative word or clausal part. Moreover, with JA working like an operator in front of a CP in its own right, its import, syntactically as well as semantically and pragmatically, is quite different from that of MPs. What all of this leaves us with is that the generalization that MPs function as functional subcategories of VF remains intact. Compare Figure 1. If VF were to be understood as contrastive (as one reviewer points out), the alternative set would encompass grammatical options. I leave this issue open for the time being. Notice that this does not touch upon our conclusion that MPs, once in collocation with VF, operate on top of and modify VF.



Discourse marker = discourse particle = thetical = modal particle?  

 257

As to the scopal relations between the constituting components in (23a) and (23b), VF operates epistemically on declarative p, VerumFoc(p), and the illocutively functioning MPs, ja and JA, both operate on top of the verum-focused p, but they do different things. See (24a)/(24b): (24) a. ja(Mood(p)) b. JA(VerumFoc(Mood(p))) Since the MP ja is a free (grammatical) morpheme and as such is not embedded under the clause mood operator, VERUM stays outside of the clausal mood operator. This results in JA(p) = ja(VERUM(p)) as in (24) (cf. also Gutzmann 2010: 20). Notice that the MP- and verum-modified version of the proposition p, of the content of the clause, is not suspended under clausal negation. Rather, all that is Neg-operated on is p: (25) a. ja(Mood(Neg(p))) b. JA(VF(Mood(Neg(p))))  c. A: Paul will Maria ja einladen. ‘P. wants to invite M. alright’ B: Paul will Maria gar nicht einladen. ‘P. does not want to invite M. in the first place.’ d. A: Paul will Maria JA einladen.  ‘P. indeed wants to invite M.’ B: Nein, Paul will Maria NICHT einladen. ‘No, P. does not want to invite M.’

illustrated by (26c) illustrated by (26d)

(see footnote 11)

Negation (contradiction, refusion, withdrawal) of p, thus, never operates on any of the functional components of p. The B-parts in (25c)/(25d) hold irrespective of the presence of the MPs.

6 Free grammatical MPs in German and in other languages We continue to address typological questions. One open question is whether MPs can follow or precede arguments and adverbials in the scrambling middle field and what their serial relation is with negation. Again we turn to Vietnamese for the reasons presented above (topic language, VO, great number of MPs). For Vietnamese, Duffield (2013) presents what we think is a parallel to MPs in German in the sense that they are free grammatical morphemes with lexically highly

258 

 Werner Abraham

underspecified, but illocutively forceful and constrained meanings. Vietnamese is a radically isolating language with a fairly rich inventory of free grammatical morphemes with clear illocutive functions. German, by contrast, has a fairly rich inflectional morphology tending toward greater morphological fusion on the whole, which obscures underlying illocutive relationships. Thus, what best serves our purposes methodologically identifying the free illocutive MP-functors in German is a high degree of similarity with the unusually transparent view of functional phrase structure in Vietnamese. Duffield presents a set of pre-verbal functional categories expressing assertion, negation, and mood (illocutionary force), most specifically, on the varying interactions between the “assertion marker” có and the negation marker không. Compare (26) for the structural position of emphasized assertion and question, taken from Duffield (2013), who explicitly builds on Höhle (1992). Note the functional and distributional parallels between Vietnamese có and the English “auxiliary do” as well as negation and VF in German. (Capital letters indicate contrastive emphasis.). (26) a. Anh CÓ mua sách! prn ass buy book ‘He DID buy the book!’ [German: Er HAT das Buch gekauft!]  VF (Duffield 2013: 260, Ex. [1a]) b. Anh KHÔNG ( có) mua sách! prn neg ass buy book ‘He did NOT buy the book!’ [German: Er hat das Buch NICHT gekauft!] negation focus (Duffield 2013: 260, Ex. [1b]) c. Anh ( có) mua sách không? prn ass buy book neg ‘Did he buy the book or didn’t he?’ [German: HAT er das Buch (jetzt) gekauft (oder NICHT)?]VF (Duffield 2013: 260, Ex. [1c]) From the negator không preceding the optional assertive có in (27b) and không occurring in sentence final position in (26c), Duffield (2013) concludes that không need to be reconstructed in front of vP for the required scope relations: (27) a. CÓ ( chứ)! ass excl ‘(He) did (indeed)!’ [German: Gewiss doch/HAT er!]  (Duffield 2013: 260, Ex. [1d])

VF



Discourse marker = discourse particle = thetical = modal particle?  

b. KHÔNG ( có)! neg ass ‘No, he didn’t!’ [German: (Nein,) Hat er NICHT!] (Duffield 2013: 260, Ex. [1e])

 259

negation focus

Assertion is thus shown to be projected low in the functional phrase structure in Vietnamese – it is in a position below and scoped over by that of sentential negation (Duffield 2013: 260). Since, as in German and probably universally, sentential negation in Vietnamese is rhematic information, it carries sentential stress by default. In German this is deep down in VP, usually in front of V. Any other position of negation is no longer clausal, but scopes over argument-DPs. (# signals requirement of special dialogic contexts.) See (28): (28) a.   Sie hat das den Eltern NICHT gesagt.   she has that the.dat.pl parents not said  ‘She did NOT say that to her parents!’ a.’  Sie HAT das NICHT gesagt.  she has that not said  ‘#She DID not say that to anyone!’ a.’’ Sie LIEBT dich! – ((Das) Wird sie) NIEMALS!  ‘She LOVES you! – She NEVER (will)!’ b.  Sie HAT das gesagt!  ‘She DID say that!’ c.  SO hat sie das gesagt! / Hat sie SO gesagt!  ‘She said that in SUCH A WAY!’ d.  HAT sie so gesagt!  ‘She DID say that in such a way!’ The (28a’) and (28a’’) versions are verum focused as in (28b)/(28d), whose syntactic locus is high in the clause structure. Recall (16). But presumably there is  a wide gap between the near-IP slot for (28b)/(28d) and the verum-like ­Assertion/Mood slot in (26a). Consequently, both versions’ positions must be taken to relate to signaling the function of Assertion declarative – in the latter, lower case as an instantiation of Mood/Modality. See (29b) and compare with (16) as well as with the assumption of two Focus locations, a higher one and a lower one, as conceived by Battlori and Hernanz (2011) in (21), repeated here for convenience in (29a): (29) a. [CP … [FOCUSP EPMi [POLP ti [IP … [FP low EPM [vP …]]]]]]

260 

 Werner Abraham

b. Fin TP XP

T’ T

Emph/AssP

Emph/Ass MP

Asp1P Asp1

Asp2P Asp2

Ass/NegP Ass/Neg

VP V

This suggests a modification of Rizzi’s (1990, 1997) cartographic functional expansion of CP, much in the sense of (30), with the crucial distinction of “high emphasis/focus” and “low emphasis/focus”: (30) Force < Emphasis < Topic < Focus < Fin(iteness) < TP < Emph/Ass < Asp1P < Asp2 < Ass/NegP < VP The essential insight that (29) and (30) provide is that the syntactic scope position of Assertion features derives from below in the clausal structure to raise to IP. Notwithstanding the fact that MPs have nothing directly to do with high and low emphasis/focus, they relate to the illocutive/modal functions, declarative, and clause negation occupying identical functional locations. Recall our assumption that MPs are free grammatical categories serving the illocutive notions of Sp (and Addr) deixis (perspective) with maximal scope over the clausal structure (Abraham 2015). The fact that lower Ass(ertion) eventually lands higher up in the vicinity of IP draws on the fact that bound grammatical affixes hook up to Comp0 to attach to the required lexical predicate – a process that free grammatical categories do not share. Disregarding other functional nodes in (31) but those essential for our topic, MPs are therefore to be seen in a structural span from below T down to the left of the Mood/Modality realm: (31) [FIN …[TP XP [Emph/AssP/MP Emph/Ass/MP [MoodP/AssP Mood/Ass/MP [ModalityP Modality [vP . . .]]]]]]



Discourse marker = discourse particle = thetical = modal particle?  

 261

Note that under the assumption of (31), Klein’s (1998) tenet is met to the extent that Tense marks an arbitrary time span, to which either a specific time point is assigned, or that this time span is also open for assignment of the function of Assert or some other illocutive/modality marker. The expectation, then, is that alongside exponents of assertion features illocutive, modal features, be they low or high, should occupy the same primary merge position in German as they do in other languages. Notice that it would be wrong to conclude that all one can say about MPs is that they can follow or precede arguments and adverbials in the scrambling middle field. They never follow negation. Yet, there are more and far better motivated restrictions of the serialization, not only of MPs as a single class of lexemes, but with respect to differences among single subclasses thereof. We shall return to this presently. The last conclusion appears to bear out when we look at the illustrations in the following section. Note the alternatives between the finite predicate and MP-morphemes.

7 Verum focus and MP-readings We claim that there is an interesting relation between verum focus (VF) and the selection of MPs in terms of contrast intonation varying between VF and lexical MP focus. See the illustrations in (32a)–(32d’). Given that VF bears on truth functionality, one must ask to what extent MP-selection shares this quality. (32) a. Hat sie AUCHADV-FP gesagt!  prefield object ellipsis + declarative V-first ‘She did TOO say that!’ a.’ HAT sie auchMP (nicht) gesagt! ‘She DID too say that!’ b. Hat sie EBENTEMP-ADV gesagt! has she this very moment said b.’ Hat sie EBENADJ hinter sich gebracht. ‘She made it on level ground.’ b.” HAT sie ebenMP gesagt! ‘She DID say that for whichever reason!’ c. Hat sie SCHONTEMP-ADV/MP gesagt! has she ADV/MP said c.’ HAT sie schonTEMP-ADV gesagt! — Hat sie schonTEMP-ADV has she already said has she already GESAGT! said

262 

 Werner Abraham

d. Hat sie DOCHMP gesagt! ‘Has he said all the same.’ d.’ HAT sie dochMP gesagt! — has she though said

Hat sie dochMP GESAGT!

The least specific thing one can obviously say about (32a)–(32d) is that contrast accent on the MP lexeme – to the extent that it can be emphasized in the first place – elicits the lexical reading of the source category, as is the case in (32a), (32b), (32b’), (32c). An exception is (32d), (32d’), which is mitigated by the fact that the lexical doch occurs as an MP both with and without contrastive accent. While this is true for schon as well as in (32c), (32c’), schon shares the swap properties of the other MP-lexicals in (32a)–(32c): Under VF the MP-reading is elicited, while stress on the homonymic lexeme allows only for the lexical source reading (with category status as adverbial or focus particle). Apart from the exception in (32d), there appears to be compensatory (mutually excluding) distribution between focus on a grammatical word (VF) and lexical focus on the MP-homonymic adverbial or focus particle. Specifically, under VF which is in fact focus on the truth value the modal option of the MP-morpheme comes to the fore. See (33): IST ([MP vielleicht]) ein Gauner! (33) a. Der he.dem is perhaps a rogue b. DER ist ([MP vielleicht]) ein Gauner. he.dem is perhaps a rogue c. Der ist [*MP/ADV VIELLEICHT] ein Gauner. he.dem is perhaps a rogue Consider (33a) expressing the illocutive mode of mirative confirmation emerges on the strength of VF on the copula with or without the MP vielleicht. By contrast, the homonym without VF on the copula but on VIELLEICHT, as in (33c), elicits the heteronymic adverbial function of vielleicht. In other words, the operator effect of vielleicht in (33a) must be very different from that in (33c). The illocutive function emerging from MP-slots accompanying VF on Infl must be between IP and vP. Either slot is one relating to (truth) Assertion or its temporal or modal modification (on the scale between certainty and uncertainty). The properties of the assertion phrase may be assumed to equal one feature specification of the general Mood or Modality Phrase. We further assume that many, if not all, illocutionary features are initially associated with this lower Assertion projection as is the case in other languages. Our claim that assertive emphasis (VF) and the illocutive function of MPs have much in common is consistent with Laka’s (1990) general claim that features relating to truth-value and illocutionary force are



Discourse marker = discourse particle = thetical = modal particle?  

 263

projected relatively low (below TP) in phrase structure. We may go even further by showing that this low position can be more highly articulated and involves at least two dissociable functional categories (NegP/PolP and Assertion/Mood). See (34):12 (34) a. Hat er die Katze denn gefüttert? has he the cat MP fed b. Hat er die Katze WIRKLICH has he the cat truly gefüttert? fed c. Ist WAHR, dass er die Katze is true that he the cat gefüttert hat? has fed d. Ist wahr, DASS er die Katze is true that he the cat gefüttert hat? fed has e. Ist wahr, dass er die Katze is true that he the cat gefüttert HAT? fed has f. Er weiß, [c(omp) DASS [ sie he knows that she angekommen ist ]] arrived has g. Was denn/bloß/nur/auch/?schon/*ja/*eben/ what MP/… *vielleicht hat er sich überhaupt MP has he refl anyway dabei gedacht. thereof thought

signaling Sp’s doubt as to Addr’s claim Ass(p) signaling Sp’s strong doubts as to Addr’s claim Ass(p) signaling Sp’s surprise about Addr’s claim Ass(p)

signaling Sp’s surprise given that he did not expect Addr’s claim Ass(p) signaling Sp’s doubts as to Addr’s claim Ass(p) holding for the past C(omp)-VF (signals ­ignorance beyond p)

MP adjoined to wh-word in the wh-question

12 Some of the nearly identical comments for (34a)–(34h) are not meant to gloss over subtle differences between them, but to confirm that VF can be expressed in different, though closely related terms.

264 

 Werner Abraham

h. Was hat-a-si-n denkt? what has-he-refl-denn.cl thought ‘What did he think?’ (34g) illustrates the restricted occurrence of MPs in wh-questions (cf. Thurmair 1989), (34h) demonstrates that in contrast to Modern Standard German, the Southern German dialects do not use wh-questions without cliticized denn, thus making the MP an integral part of wh-questions in ways that resemble subjunction inflection. Without any illocutive operation as in the simple unmarked yes/ no-question Hat er die Katze gefüttert? as in (34a) (without denn), it is merely signaled that Sp is undecided between a positive and a negative assertion, thus . This is called Sp-epistemics with respect to p. The VF variant in (34b), however, signals that the discursive CG on judging whether or not the cat has been fed is to be partially shared between Sp and Addr. However, this has not been decided on as yet: i.e. ?CG(p(Sp) v ¬p(Sp)), or in words: The CG with respect to the proposition (utterance) established by Sp and Addr is still to be aligned.

8 MPs and epistemic vs. root modal scoping Building on work by Coniglio (2011), Gärtner (2013, 2017) undertakes a more systematic exploration of the inventory of modal particles in root infinitivals. The special value of this investigation lies in the fact that the selection of MPs is independent of finiteness. Coniglio adapted Cinque’s (1999) adverbial hierarchies to the domain of MPs. Notice the peculiar affinity of root infinitivals (infinitives without the infinitival preposition zu ‘to’) and the infinitival imperatives in German. This suggests the generalizations in (35) and (36): (35) MP-hierarchy (IMP-compatible) (cf. Coniglio 2011: 80) doch > halt > DOCH > nur (nicht) > bloß (nicht) > ruhig > JA (36) MPs acceptable in root infinitivals (present infinitives and past participles) (cf. Gärtner 2013: 224, 2017: x) doch > halt > DOCH > nur (nicht) > bloß (nicht) > ruhig > JA Evidence is also provided in (37a)–(37c): (37) a. Leg die Bücher doch ( JA, *ja) auf den Tisch!IMPERATIVE lay.imp the books MP MP MP on the table ‘Put the books on the table, will you!’ b. Die Bücher (* doch, OKeben/nur JA) auf den ROOT INFINITIVE the books MP … on the



Discourse marker = discourse particle = thetical = modal particle?  

Tisch legen! table lay ‘Put the books on the table, will you!’ c. Die Bücher (* doch, OKeben/nur JA) auf den the books MP … on the Tisch gelegt! table laid ‘Put the books on the table, will you!’

 265

PAST PARTICIPLE

Taking Cinque’s (1999) universal hierarchy of functional projections, modal epistemicity (as well as evidentiality) can be modeled in terms of distinct dedicated functional projections, i.e. (38): (38) [ frankly Moodspeech act [ fortunately Moodevaluative [allegedly Moodevidential [ probably Modepistemic [once T(past) [then T(future) [ perhaps Modirrealis [necessarily Modnecessity [ possibly Modpossibility [ … Asp … We see reason to argue for an extension of (38) in terms of Sp deixis. The question is how to align the operator status of MP-morphemes on the Sp’s expected, or challenged, degree of shared CG on Addr’s side. See the hierarchies in (39)–(41) relating the invisible discourse structure of p in Cartographic scopal terms to the visible illustration: (39) Invisible Discourse structure of p in Cartographic scopal structure: {[Verum focus]/Sp-Addr >> CG >> Sp-epistemics (EMV) >> Topic (à propos) >> Foc >> Topic (about)} >> CP What are the mutual relations between (39) and (40)? (40) Visible Clause structure of p: where are MP qua scope? CP(/) >> Comp/FIN >> WP >> { } >> VP >> Rhema >> V/FIN The answer to the question is provided by (41)–(42). (41) shows that MPs with their Sp-Addr oriented status outscope all other clause-structural categories. (42) shows the structure below which hosts the Wackernagel position (WP), following MPs and the rhematic domain. (41) Invisible scopal discourse (CP left peripheral) structure of p:  {[VF]/MP/Sp-Addr >> CG >> Sp-epistemics (EMV) >> Topic (à propos) >> Foc >> Topic (about)} >> Fin (42) Visible Clause structure of p with inceptive Comp/Fin: Comp/FIN >> WP >> [Thema {MP> Thema] >>[ Rhema VP Rhema] >> Vo

266 

 Werner Abraham

The following generalizations can be envisioned deriving from the specifics above: (43) MP-positions and functions can be explained in the context of the functions and the constraints of the phenomenon of VF in terms of speech acts (particularly felicity conditions; cf. Egg and Mursell [2017]) and clause type. (44) MP-positions are to be integrated into the net of conditions of the Cartographic expansion of CP into the information structure (discourse dimension) and, consequently, in terms of the Thema-Rhema distinction and (contrastive) focus. (45) The structural areas of integration of VF- and MP-positions are SpecCP, i.e. INFL/first verbal bracket (possibly + prefield), and V-last/second verbal bracket. (46) MP-functions are activated in the context of VF. This is because it is not until under VF, Sp-deicitc, epistemic truth-functional valuations become operative. They are presupposed for the operativity of MPs. (47) If sentential epistemicity may be mapped in terms of the uncertainty of the Sp’s utterance claim p: a. λx:Sp(x) ∧ KNOW(x,[p∨¬p)] then it holds for MP: b. λxλy:Sp(x), Addr(y) ∧ CG(x,y,[p∨¬p]). One may also say: If p-truth valuation/Epistemics relates to Sp-uncertainty about p, then one may say that under VF over p, there is need of negotiating a CG between Sp and Addr wrt the valuation of p. See (48)–(52): (48) MPs signal the need of negotiating CG between Sp and Addr. Thus, MPs are operators about the epistemic state of Sp. The epistemicity operation on VF involving only Sp wrt p is is a one-place/intransitive operation. Once this operation is extended by one predicative place with the Addr as a negotiating (and challenging) partner, one may speak of a two-place/transitive operation (Abraham 2012a, 2012b). (49) Lohnstein and Stommel (2005) claim that VF is a phase structure, which is an extra Focus function in the Rizzi-CP-expansion dominating the higher Topic aboutness function (Abraham 2012b). Against this background, MPs as grammatical functions (not lexical words) have phase status as well. Sp-epistemics (EMV) is lower in the hierarchy in the sense that MP presupposes VF as Sp-epistemics. See the functional phases in (38). (39) arranges



Discourse marker = discourse particle = thetical = modal particle?  

 267

the serial and scope order of the lexical and grammatical constituents and heads of CP in German. Compare (37)–(38) above. (50) The fact that literature on MPs in German focusing on criteria of lubricating dialogical speech and making the interaction smoother and Addr-friendlier has never been analyzed with formal precision. The link between (51) and (52) below is about what is at hand: MPs – as much as grammatical focus at the hands of VF – involve Sp and Addr in a CG-determining fashion – thereby leaving the CP-base in (51) by transferring their semantic weight to the discursive CP-expansion in (52). The result is a far higher scoping effect. The very fact that MPs (as well as the non-lexical type of VF) undergo this scoping transfer appears to be at the bottom of the specific form-function property of grammatical elements as basically and radically different from lexical representatives of truth assignment (such as wirklich-truly/really, vermutlich-presumably, wahrscheinlich-probably etc.). Such lexical representatives are usually taken to be equivalent to the grammatical exponents of truth ascriptions (Diewald and Smirnova 2010a, 2010b; Degand, Cornille, and Pietrandea 2013). However, lexical exponents achieve truth (degree) effects on their purely lexical-semantic potential, but never in suchterms as the structural, scope-inducing and scope indicating, ones as in (48) and (49). Note also that this conclusion corroborates our claim that MPs are grammatical form-function representatives – a status that makes superfluous the need of clarifying the syntactic category status ofMPs, e.g. as (light) adverbials etc. (41)–(42) are replicated here as ­(51)–(52): (51) Invisible scopal discourse (CP left peripheral) structure of p: {[VF]/MP/Sp-Addr >> CG >> Sp-epistemics (EMV) >> Topic (à propos) >> Foc >> Topic (about)} >> Fin The scope relations in (51) are not reflected in the spell-out in (52): (52) Visible Clause structure of p with inceptive Comp/Fin: Comp/FIN >> WP >> [Thema Thema] >>[ Rhema VP Rhema] >> Vo

9 Indismissable topic reference of MPs What the range of MP-positions in the left middle field of German amounts to is that they are constrained to the cartographic field of functions that account for discourse-related, rather than clause-related, sentential properties. The CP-based

268 

 Werner Abraham

constraint in the Logical Form (LF) in (52) also implies that MPs cannot surface in thetic – i.e. purely rhematic (“out-of-the blue”) – sentences. See (53a), (53b). Note that in (53b), the MP-positions before or after the definite DP der homo novus decides over categorical vs. thetic status of the sentence – and, accordingly, about the insertability of MPs. Take (53a) as a thetic stage direction (‘Enters the homo novus’). (53) a. [VF AUF-[Comp tritt [MF * eben/*ja/*schon/*doch ein homo up- steps MP … a homo novus [V AUFtritt]]]] novus b. [VF Es [Comp tritt