Open Collections

UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Discourse particles and the syntax of discourse-evidence from Miesbach Bavarian Thoma, Sonja Christine 2016

Your browser doesn't seem to have a PDF viewer, please download the PDF to view this item.

Notice for Google Chrome users:
If you are having trouble viewing or searching the PDF with Google Chrome, please download it here instead.

Item Metadata

Download

Media
24-ubc_2017_february_thoma_sonja.pdf [ 23.64MB ]
Metadata
JSON: 24-1.0340686.json
JSON-LD: 24-1.0340686-ld.json
RDF/XML (Pretty): 24-1.0340686-rdf.xml
RDF/JSON: 24-1.0340686-rdf.json
Turtle: 24-1.0340686-turtle.txt
N-Triples: 24-1.0340686-rdf-ntriples.txt
Original Record: 24-1.0340686-source.json
Full Text
24-1.0340686-fulltext.txt
Citation
24-1.0340686.ris

Full Text

 DISCOURSE PARTICLES AND THE SYNTAX OF DISCOURSE- EVIDENCE FROM MIESBACH BAVARIAN by  SONJA CHRISTINE THOMA  B.A., Western Washington University, 2005  A THESIS SUBMITTED IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF  DOCTOR OF PHILOSOPHY in THE FACULTY OF GRADUATE AND POSTDOCTORAL STUDIES (Linguistics)  THE UNIVERSITY OF BRITISH COLUMBIA (Vancouver)   December 2016  © Sonja Christine Thoma, 2016   ii  Abstract  This dissertation is concerned with the form, function, and distribution of discourse particles in Miesbach Bavarian. These elements are commonly considered in either semantic, pragmatic, or discourse analytic terms. This current investigation explores the interaction between form, meaning, and distribution of discourse particles, their syntax. I show that discourse particles in Bavarian are constructed, and discourse particles therefore should not be considered as a primitive. ‘Discourse particle’, as I show in this dissertation, is the effect of a unit of language with an invariable core meaning (among them scalar and deictic core meanings) when it associates with a discourse functional syntactic layer that represents the discourse participants’ epistemic states.  The claims of this dissertation are empirical at the core; I show conversational data from the Miesbach Bavarian dialect of German that provides the need to distinguish three classes of discourse particles (DPRTs); speaker oriented, addressee oriented, and other oriented DPRTs. I present an analysis that proposes these three classes to be the result of an association with different discourse participants (speaker, addressee, or other). This association serves to ground propositions. In order to model this grounding function of those items interpreted as DPRTs, I make use of the Universal Spine Hypothesis, a framework proposed by Wiltschko (2014). I extend Wiltschko's Universal Spine to include the participant anchor with the projection GroundP.   iii  Preface  Except for some independent and collaborative projects presented at conferences and/or submitted for publication, the work in this dissertation is based on original, unpublished research by the author, Sonja Thoma. Whereas the analysis and theoretical implementation of the analysis is my own, many aspects of the framework is indebted to the fruitful discussions from lab meetings in Martina Wiltschko’s ‘Eh-Lab’ (http://syntaxofspeechacts.linguistics.ubc.ca).   Part of the analysis for fei (Chapter 4) was published as To p or to¬ p; The Bavarian Particle fei as Polarity Discourse Particle in Sprache und Datenverarbeitung, 33(1-2):139-152 (Thoma 2009). The analysis of some of the discourse particles examined here was previously published as Thoma (2013) Bavarian discourse particles-at the syntax pragmatics interface in Proceedings of NWLC 2013, University of British Columbia Working Papers in Linguistics 38:41-58; the analysis presented here differs slightly in implementation.   The development of the syntactic framework of the Extended Universal Spine used for the analysis in Chapter 6 is due to discussion, and collaboration with colleagues of the Eh-Lab. This framework is elaborated in a publication I co-authored with other members of the Eh-Lab: ‘Intonation and Particles as Speech Act Modifiers: A Syntactic Analysis’, in Studies in Chinese Linguistics as Heim Johannes, Herman Keupdjio, Zoe Lam, Adriana Osa-Gómez, Sonja Thoma and Martina Wiltschko (2016).   All fieldwork undertaken for this dissertation is covered under UBC ethics approval for the project “The Syntax of Speech Acts”, certificate # H12-01864, awarded to Dr. Martina Wiltschko.     iv  Table of Contents  Abstract	...................................................................................................................................................................	ii	Preface	...................................................................................................................................................................	iii	Table	of	Contents	................................................................................................................................................	iv	List	of	Tables	..........................................................................................................................................................	x	List	of	Figures	.....................................................................................................................................................	xii	List	of	Abbreviations	.......................................................................................................................................	xiii	Acknowledgements	.........................................................................................................................................	xiv	Dedication	..........................................................................................................................................................	xvii	Chapter	1:	 Introduction	.......................................................................................................................................................	1	1.1	 Introduction	...........................................................................................................................................................	1	1.2	 	Data	and	Methodology	.....................................................................................................................................	4	1.2.1	 	Miesbach	Bavarian	......................................................................................................................................................	5	1.2.1.1	 	Dialect	......................................................................................................................................................................	5	1.2.1.2	 Syntax	.......................................................................................................................................................................	7	1.2.2	 Why	dialect?	.................................................................................................................................................................	13	1.2.2.1	 Methodology	and	data	presentation	........................................................................................................	13	1.2.2.2	 Dialect	as	L1	........................................................................................................................................................	17	1.2.2.3	 Dialect	as	spoken	language	phenomenon	..............................................................................................	18	1.2.2.4	 Variation	in	DPRT	inventory	and	use	......................................................................................................	19	1.2.3	 Summary	........................................................................................................................................................................	27	1.3	 	Theoretical	assumptions	and	background	...........................................................................................	28	  v 1.3.1	 Syntax	..............................................................................................................................................................................	28	1.3.2	 Pragmatic	model	........................................................................................................................................................	29	1.4	 	Proposal	...............................................................................................................................................................	32	1.5	 	Roadmap	.............................................................................................................................................................	35	Chapter	2:	 Discourse	particle	properties	..................................................................................................................	38	2.1	 	Introduction	.......................................................................................................................................................	38	2.2	 	DPRT	properties	and	the	questions	they	raise	...................................................................................	40	2.2.1	 	Two	types	of	multi-functionality	........................................................................................................................	40	2.2.2	 	Obligatoriness	............................................................................................................................................................	45	2.2.3	 	Distribution	.................................................................................................................................................................	48	2.2.4	 	Heads	or	phrases	.......................................................................................................................................................	48	2.2.5	 	Propositional	scope	.................................................................................................................................................	53	2.2.6	 	Sentence	type	restriction	......................................................................................................................................	55	2.2.7	 	The	semantico-pragmatic	properties	of	DPRTs	..........................................................................................	59	2.2.7.1	 	DPRTs	and	illocutionary	force	...................................................................................................................	61	2.2.7.2	 	Expressions	of	speaker	attitude	................................................................................................................	64	2.2.7.3	 	DPRTs	and	presuppositions	.......................................................................................................................	65	2.3	 	Conclusion	..........................................................................................................................................................	66	Chapter	3:	 Ingredients	for	an	analysis	of	DPRTs	...................................................................................................	69	3.1	 	Introduction	.......................................................................................................................................................	69	3.2	 	The	test	cases	....................................................................................................................................................	70	3.2.1	 	Jetz	...................................................................................................................................................................................	72	3.2.2	 	Eh	.....................................................................................................................................................................................	75	3.2.3	 	Ja	.......................................................................................................................................................................................	77	3.2.4	 	Doch	................................................................................................................................................................................	83	3.2.5	 	Fei	....................................................................................................................................................................................	87	  vi 3.3	 	Belief,	commitment	and	epistemicity	.....................................................................................................	89	3.3.1	 	Grounding:	separating	speaker	from	addressee	.........................................................................................	91	3.3.2	 	Context:	the	relation	between	an	individual	and	p	....................................................................................	94	3.4	 	Clause	types,	illocutionary	force,	and	speech	acts	............................................................................	98	3.4.1	 	Syntactic	form	as	context	....................................................................................................................................	102	3.4.2	 	Intonation	...................................................................................................................................................................	103	3.4.3	 	Declarative	.................................................................................................................................................................	105	3.4.4	 	Interrogative	.............................................................................................................................................................	108	3.4.4.1	 	Wh-interrogative	...........................................................................................................................................	108	3.4.4.2	 Polar	interrogative	.........................................................................................................................................	109	3.4.5	 Imperative	...................................................................................................................................................................	112	3.5	 Conclusion	.........................................................................................................................................................	113	Chapter	4:	 Speaker-orientation,	addressee-orientation	and	other-orientation	...................................	116	4.1	 Introduction	......................................................................................................................................................	116	4.2	 A-orientation	....................................................................................................................................................	119	4.2.1	 A	believes	p	.................................................................................................................................................................	123	4.2.2	 Soliloquy	......................................................................................................................................................................	126	4.2.3	 A	believes	that	not	p	...............................................................................................................................................	129	4.2.4	 A	doesn’t	believe	that	p	.........................................................................................................................................	131	4.2.4.1	 Out	of	the	blue	context	.................................................................................................................................	132	4.2.4.2	 Answer	to	a	question	....................................................................................................................................	135	4.3	 S-orientation	.....................................................................................................................................................	138	4.3.1	 A	belief	is	irrelevant	................................................................................................................................................	139	4.3.2	 S	belief	is	relevant	....................................................................................................................................................	143	4.3.2.1	 S	committed	to	p	due	to	discourse	context	.........................................................................................	143	4.3.2.2	 S	committed	to	p	due	to	form	....................................................................................................................	146	  vii 4.4	 Predictions	so	far	............................................................................................................................................	148	4.5	 O-orientation	....................................................................................................................................................	152	4.5.1	 Jetz	..................................................................................................................................................................................	156	4.5.2	 Eh	....................................................................................................................................................................................	160	4.6	 Conclusion	.........................................................................................................................................................	163	Chapter	5:	 Deriving	the	functional	range	of	DPRTs	...........................................................................................	166	5.1	 Introduction	......................................................................................................................................................	166	5.2	 Formalizing	grounding	................................................................................................................................	170	5.3	 The	normal	course	of	a	conversation	....................................................................................................	172	5.4	 DPRTs,	inferences	and	discourse	coherence	......................................................................................	178	5.5	 The	functional	range	of	ja	...........................................................................................................................	181	5.5.1	 	Shared	knowledge	..................................................................................................................................................	183	5.5.2	 	Surprise	.......................................................................................................................................................................	186	5.5.3	 	Emphasis	....................................................................................................................................................................	190	5.5.4	 	Reason	.........................................................................................................................................................................	191	5.5.5	 	Summary	.....................................................................................................................................................................	193	5.6	 The	functional	range	of	doch	.....................................................................................................................	194	5.6.1	 	Contrast	.......................................................................................................................................................................	196	5.6.2	 	Reminding	..................................................................................................................................................................	200	5.6.3	 	Backchecking	............................................................................................................................................................	205	5.6.4	 	Shared	knowledge	..................................................................................................................................................	207	5.6.5	 	Summary	.....................................................................................................................................................................	210	5.7	 The	functional	range	of	fei	..........................................................................................................................	211	5.7.1	 	Newness	......................................................................................................................................................................	213	5.7.2	 	Emphasis	....................................................................................................................................................................	216	5.7.2	 	Summary	.....................................................................................................................................................................	218	  viii 5.8	 DPRTs,	presuppositions,	and	accommodation	..................................................................................	219	5.9	 Conclusion	.........................................................................................................................................................	222	Chapter	6:	 The	syntax	of	discourse	particles	.........................................................................................................	225	6.1	 Introduction	......................................................................................................................................................	225	6.2	 Lexical	hypothesis	..........................................................................................................................................	226	6.2.1	 Problem	#1:	Clause	type	restrictions	are	not	feature	specification	..................................................	227	6.2.2	 Problem	#2:	DPRTs	are	multifunctional	........................................................................................................	230	6.2.3	 Problem	#3:	Orientation	correlates	with	context	......................................................................................	233	6.2.4	 Summary	......................................................................................................................................................................	236	6.3	 Syntactic	hypothesis	.....................................................................................................................................	237	6.3.1	 	The	Universal	spine	hypothesis	........................................................................................................................	238	6.3.2	 	Extending	the	spine	...............................................................................................................................................	241	6.3.2.1	 Evidence	from	agreement	...........................................................................................................................	248	6.3.2.2	 Evidence	from	confirmationals	................................................................................................................	250	6.3.3.3	 Evidence	from	particle	order	....................................................................................................................	251	6.4	 DPRTs	and	the	universal	spine	.................................................................................................................	252	6.5	 Evidence	for	propositional	scope	of	DPRTs	........................................................................................	258	6.5.1	 DPRT	and	the	vP	boundary	.................................................................................................................................	259	6.5.2	 DPRTs	and	adverbs	.................................................................................................................................................	261	6.5.3	 DPRTs	and	negative	concord	..............................................................................................................................	265	6.6	 Evidence	for	the	association	of	DPRT	with	GroundP	......................................................................	268	6.6.1	 Co-occurrence	and	ordering	restrictions	......................................................................................................	269	6.6.2	 Scope	.............................................................................................................................................................................	276	6.6.2.1	 DPRTs	and	adverbs	.......................................................................................................................................	277	6.6.2.2	 DPRTs	and	confirmationals	.......................................................................................................................	283	6.6.3	 Accent	on	DPRTs	......................................................................................................................................................	285	  ix 6.7	 The	linearization	problem	..........................................................................................................................	290	6.8	 Conclusion	.........................................................................................................................................................	298	Chapter	7:	 Conclusion-	summary	and	open	questions	......................................................................................	301	7.1	 Introduction	......................................................................................................................................................	301	7.2	 DPRT	properties	revisited	..........................................................................................................................	302	7.2.1	 Multifunctionality:	orientations	and	functional	range	............................................................................	303	7.2.2	 Optionality	..................................................................................................................................................................	304	7.2.3	 Categorial	status	.......................................................................................................................................................	305	7.2.4	 Propositional	scope	.................................................................................................................................................	306	7.2.5	 Sentence	type	restriction	......................................................................................................................................	307	7.2.6	 Contribution	to	an	utterance	..............................................................................................................................	307	7.3	 Grammaticalization	.......................................................................................................................................	308	7.4	 Lexicalization	patterns	.................................................................................................................................	311	7.5	 Conversation	as	the	basic	setting	............................................................................................................	313	7.6	Further	research	.................................................................................................................................................	314	References	.................................................................................................................................................................................	318	    x  List of Tables Table 1: Clause typing and the topological field .......................................................................... 10	Table 2: DPRTs as heads or phrases ............................................................................................. 53	Table 3: Distribution of ja and doch across sentence types .......................................................... 57	Table 4: Multi-functional uses of jetz ........................................................................................... 74	Table 5: Multi-functional uses of eh ............................................................................................. 77	Table 6: Multi-functional uses of ja .............................................................................................. 82	Table 7: Multi-functional uses of doch ......................................................................................... 87	Table 8: Intonation associated with declaratives and resulting functions ................................... 107	Table 9: Intonation associated with wh-interrogatives and resulting functions ......................... 109	Table 10: Intonation associated with V1 clauses and resulting functions .................................. 112	Table 11: Intonation associated with imperatives and resulting functions ................................. 113	Table 12: Summary of clause types, intonation, functions, and commitments .......................... 115	Table 13: Orientations arising from epistemic states .................................................................. 118	Table 14: A-orientation and epistemicity ................................................................................... 137	Table 15: S-orientation and epistemicity .................................................................................... 147	Table 16: Participant epistemicity and DPRT orientation .......................................................... 165	Table 17: Epistemicity matrix for assertion ................................................................................ 172	Table 18: Epistemicity matrix for shared knowledge ................................................................. 184	Table 19: Epistemicity matrix for surprise ................................................................................. 186	Table 20: Epistemicity matrix for emphasis ............................................................................... 190	Table 21: Epistemicity matrix for reason .................................................................................... 192	Table 22:The functions of ja ....................................................................................................... 194	  xi Table 23: Epistemicity matrix for contrast ................................................................................. 196	Table 24: Epistemicity matrix for reminding .............................................................................. 200	Table 25: Epistemicity matrix for backchecking ........................................................................ 205	Table 26: Epistemicity matrix for shared knowledge ................................................................. 207	Table 27: Functions of doch ....................................................................................................... 211	Table 28: Epistemicity matrix for newness ................................................................................ 213	Table 29: Epistemicity matrix for emphasis ............................................................................... 216	Table 30: Functions of fei ........................................................................................................... 218	Table 31:The functions of ja ....................................................................................................... 223	Table 32: Functions of doch ....................................................................................................... 223	Table 33: Functions of fei ........................................................................................................... 223	Table 34: DPRT multi-functionality ........................................................................................... 230	      xii  List of Figures Figure 1: Map of Germany ............................................................................................................. 6	Figure 2: Weyarn and Miesbach County in a map of Germany ..................................................... 7	Figure 3: The topological field aligned with the clausal structure .................................................. 9	Figure 4: The relational functions of the Universal Spine ............................................................ 34	Figure 5: Wohl as modifier of sentence type ................................................................................ 64	Figure 6: Ja as modifier of force ................................................................................................... 64	Figure 7: CG Content .................................................................................................................... 92	Figure 8: Discourse contexts ......................................................................................................... 95	Figure 9: Grounding an assertion ................................................................................................ 171	Figure 10: Grounding an assertion .............................................................................................. 181	Figure 11: Grounding with ja ..................................................................................................... 182	Figure 12: Grounding with doch ................................................................................................. 195	Figure 13: Grounding with fei ..................................................................................................... 212	Figure 14: Categorial identity is derived from place of association in the spine ........................ 240	Figure 15: The extended spine, associated domains and abstract functions ............................... 248	Figure 16: DPRT order based on the order of associating heads ................................................ 275	Figure 17: DPRTs surface order and place of interpretation is crossed ..................................... 276	     xiii  List of Abbreviations Abbreviations used in glosses 1,2,3 1st person, 2nd person, 3rd person DET determiner DPRT  discourse particle NEG  negation PL  plural PRT particle SG  singular  Abbreviations used in the text A  Addressee ERP Epistemic reference point f  Function MB Miesbach Bavarian O  Other S  Speaker SG  Standardized German UoL Unit of Language    xiv  Acknowledgements Just like my particles are largely shaped by the context they occur in, this dissertation itself did not get its shape outside of one. So I would like to give sincere thanks to the people that comprised my particular context on this journey. I know I won’t come anywhere close to being able to express the gratitude I feel toward all who helped shape my work and me over the dissertation years (and before), so this is but a feeble attempt at an acknowledgement.   First and foremost I want to thank my parents Hans und Hanni Thoma- für eure Geduld, auch wenn Ihr euch bestimmt oft gedacht habts was der ganze Schmarrn mit dieser ewigen Fragerei über fei und doch und ja eigentlich soll. Danke dass Ihr mir so ein unglaubliches Durchaltevermögen mitgegeben habts, und mich trotz der Vorbehalte mit diesem Doktor immer voll unterztützt habt- von den Arbeitsmonaten in Weyarn, über eure Reisen nach Kanada zum Babysitten, und so viel mehr. My brother Hansi and my late sister Alex, you too contextualize this accomplishment.  How to thank my academic parent, my Doktormutter Martina Wiltschko?- so much more than ‘just’ a supervisor, you have always supported me to the fullest, have been, and continue to be so generous with your time. You always believed in me and my little hunches, and my ability to do this. Your academic and personal rigor, calm guidance, and invaluable advice along the way not only helped me immensely academically, but also inspired me on a personal level. Gassho!  My committee members Hotze Rullman and Michael Rochemont both have contributed so much to my context. Hotze, although this research has gone very different avenues from what you may have expected from this topic, you have supported me and my academic development from the first day on- I still remember when you handed me the letter of acceptance to UBC   xv linguistics. You helped me get started on fei, always asked good and important questions, and reigned me in when I was playing  “lose and fast”. Michael, thanks for always commenting on my often painfully slow progress, always asking, always encouraging. Especially at the final stage your questions helped me focus in on important details that I would not have recognized otherwise.  Henry Davis, Raffaella Zanuttini and Caroline Rieger, the other members of my defense committee, thank you all for asking tough but important questions at my defense, and for reviewing my defense draft with such care and insight. The end result is undoubtedly better for it. Henry, although I didn’t work with you in recent years, special thanks for your role in shaping me as a wee baby-linguist when I just started at UBC. I am glad this came full circle with you as my department external.  The faculty and staff at the Department of Linguistics at UBC all provided a wonderfully enriching, stimulating and caring environment, with special thanks to Edna Dharmaratne and Shaine Meghji who always made sure the wheels turned smoothly. We’d all be lost without you!   The eh-lab has significantly shaped and helped my progress toward the middle and end of my dissertation stage, and the weekly meetings and discussions were not only incredibly fun, but invaluable in helping me to figure this all out. Thanks to Anne Bertrand, Johannes Heim, Zoe Wai-Man Lam, Marianne Huijsmanns, Oksana Tkachman, Hermann Keupdjio, Adriana Osa-Gómez, Emily Sadlier-Brown, Erin Guntly, Merlin Yang, Yifang Yuan, and Tiffany Ho.  An integral part of helping me weather the ups and downs of PhD life was Thesis Anonymous and my brothers and sisters in arms there. Research, wine, treats, and friendship, what more do you want?! Olga Steriopolo, Atsushi Fujimori, Peter Jacobs, James Thompson,   xvi Christiana Christodoulou, Solveiga Armoskaite, Heather Bliss, Anne Bertrand, Johannes Heim, and Elise McClay, thank you all.   Of course there are all the other friends I made at the Department, but the list of my fellow MA and PhD colleagues who all were an important part of my particular context is too long to list; a testament to how long it took me to finish this all…   I don’t even know where to begin with you, George Moseley. Here’s to Tomodachi Zen, my friend. Wine! Poetry! Philosphy! Life, the great adventure!  My friends in Germany Alex Thamm, Alexandra Leykauf, Dani Schönauer, Isabel Hartmann, Mikko Kahkonen, Niki Ruf, Veronika Ottl, and Martina Effaga, hardly ever asked about the ‘thing that shall not be named’! You all continue to be close to my heart. Even if we don’t see each other as much as we’d like to, when we do it counts.   Sahjia, Mike, Lily, and Jonathon, you continue to make Canada home for me.   The Terminal City Rollergirls provided a great outlet for my frustrations, and allowed me to cultivate the strength and grit to get to the end of this. You all taught me what it means to channel Red Sonja, on the track and off.    My family on the Schlichting side, especially Carl, Dorothea and Simone graciously went along this crazy ride I whisked Andreas onto- and continue to do so.  Elias, Luzia and Marinus. You continue to teach me so much about life, and about what is important, my little Zen masters! Yes, and of course my rock, Andreas. I love you so.    It takes a village to raise a child, it takes a lot more than that to raise three and accomplish something like this. Everyone along the way, people I mentioned and those of you I didn’t, you all played an important part, and for that I am deeply grateful.    xvii  Dedication  Für die Luzia, den Elias, den Marinus und den Andreas      1 Chapter 1: Introduction 1.1  Introduction Natural language and specifically its use in dialogue does much more than simply to express a specific content. The utterance of a sentence with a specific content often also conveys the interlocutors’ epistemic stance, that is, their epistemicity; the commitments, attitudes, and beliefs of the interlocutors regarding the (propositional) content encoded in the string of sound. Epistemicity is encoded differently across the languages of the world, with a varied set of linguistic units that are specialized in the expression of epistemic stance. Discourse particles are among them, and in recent years, the formal study of discourse particles has received rising attention. They are also the topic of this dissertation.   It has become clear that, as linguistic items that straddle the syntax-pragmatics interface, discourse particles can provide valuable insights into the formal mechanisms of discourse organization. German discourse particles specifically are probably among the best researched among the languages of the world. Typically, a group of 16 items is considered to comprise the ‘core’ discourse particles of German; Thurmair’s (1989) ‘classical 16’ includes aber, auch, denn, doch, eben, eigentlich, einfach, etwa, halt, ja, mal, nur, ruhig, schon, vielleicht, and wohl. However, the total number of items that are considered discourse particles is not clearly delimited, and ranges up to 40 particles (Diewald 2013). This difficulty in establishing what is to be considered a discourse particle, and what is not, still poses one of the overarching questions in current discourse particle research. This dissertation aims to address aspects of this difficulty, and proposes analytical tools in response. I also speak in detail to the specific function of discourse particles (DPRTs).    2 Consider the following data:1  (1) a.  Da Marinus is fei drei. DET Marinus is fei  three  “Marinus is fei three (years old).”      b. Da Marinus is  ja  drei.  DET  Marinus  is   ja  three   “Marinus is ja three (years old).”    c. Da Marinus  is eh drei.   DET Marinus  is eh   three    “Marinus is eh three (years old).”  Each of these utterances has the same propositional content, namely the assertion that Marinus is drei: ‘Marinus is three years old’. However, the ‘flavor’ of each utterance in (1) a-c is different, due to the contribution of the DPRTs fei, ja, and eh. Each individual DPRT contribution restricts the use of the utterances to specific contexts in a specific way. It is a well documented property of DPRTs that they do not contribute to the truth-conditional content, but add restrictions on felicitous contexts of use (Gutzmann, 2008, 2012, 2013; Kratzer, 1999; cf. Weydt, 1969). One of the contributions of this dissertation is to present the functional range, that is, the various interpretations of a given DPRT, and the corresponding contexts for the five Miesbach Bavarian discourse particles (DPRTs) fei, doch, ja, eh, and jetz.  It has been established that the general function of DPRTs is to relate the content of the utterance they occur in (i.e. the host utterance) to the context, which includes                                                 1 I will develop a detailed approach to the contribution of each particle I investigate here in Chapter 3. For now, I will simply use the form of the DPRT itself for both the gloss and the translation.   3 the conversational background shared by the interlocutors (Diewald, 2013; Zimmermann, 2011, Kaufmann and Kaufmann, 2012; Zeevat, 2006). In this dissertation, I provide evidence that DPRTs can be further classified. Based on contextual evidence, I show that DPRTs can be classified into three classes:  i) DPRTs which relate the host utterance to the speaker,  ii) DPRTs which relate the host utterance to the addressee,  iii) DPRTs that relate the host utterance to a contextually determined discourse participant (henceforth other).   DPRTs thereby serve to convey fine-grained participant epistemicity. The term epistemicity here refers to the epistemic stance of the interlocutors: the epistemic attitude the speaker has about the proposition expressed in the host utterance within the discourse context. This sensitivity to an individual discourse participant’s epistemicity is reflected in strict contextual constraints for the felicitous use of any given DPRT. These are discussed in Chapters 4 and 5 respectively.   Based on the observation that DPRTs are associated with a range of functions, (henceforth referred to as functional range) I hypothesize that the DPRTs of MB are decomposable. In particular, we call a particular Unit of Language (UoL)2 a DPRT if it occurs in a particular context (abbreviated as Cx) and has a particular function fDPRT. In other words, the function of a discourse particle derives from the contribution of the UoL and its context, as in (2).                                                  2 The term ‘Unit of Language’ UoL is introduced in Wiltschko (2014), and will be adopted here. It refers here to a basic sound π meaning Σ bundle < π, Σ >. I follow Wiltschko (2014) and will not call  <π, Σ > bundle simply a morpheme, as it can comprise units as large as phrases and utterances, which arguably should not be considered as mophemes.   4 (2)   fDPRT = UoL + Cx  I argue that the key to understanding the multi-functionality of DPRTs and their function is to consider their syntax. This argument will be presented in Chapter 6.  The remainder of this Chapter is organized as follows. This dissertation focuses on a specific dialect of German, namely the Middle Bavarian variety spoken in the Miesbach area of Upper Bavaria. I justify this methodological choice in section 1.2 I provide an overview of the Miesbach Bavarian dialect and its syntax. I present a rationale for choosing a dialectal variant of German to study DPRTs, as opposed to using data available from academic literature on DPRTs, or from written corpora. In 1.3 I introduce some of the main theoretical assumptions I use to frame the analysis. The main claims of this dissertation are introduced in 1.4. In 1.5 I present a roadmap of this dissertation.  1.2  Data and Methodology The language of investigation in this dissertation is Miesbach Bavarian (MB). Auer (2004) and Weiss (1998, 2004) give extensive arguments for the preference for data from a spoken regional variant over written data for linguistic investigation. Most importantly, a standard language, such as Standard German (SG) is subject to regulations and normative pressure. Dialects, Auer (2004) argues, display a higher level of consistency due to natural language change, which is often absent or highly constrained in a standard language. Weiss (2004) contends that prescriptive rules, arising from the normalization of a unified, written standard, arise in some of the grammatical properties of that standard.   5 This, in turn, results in data that are not as reliable for formal generative investigation, as standardized languages are subject to arbitrary normative rules not grounded in a conception of grammar. He gives examples from negative concord, and the obligatory presence of a determiner with proper nouns; both of these are grammatical features of Bavarian, but are not phenomena observed in SG. Also, the existence of clitic pronouns, which lead to typical dialect phenomena such as inflected complementizers or pro-drop, are features of Bavarian, but not of SG. All these points will be illustrated in the brief sketch of MB in 1.2.1. So why dialect? First, children acquire dialect as L1. L1, as the first language of a speaker, is the preferred object of theoretical investigation (cf. Auer 2004; Weiss 2004, 1998). This will be discussed in more detail in 1.2.2.1. Second, a dialect is primarily a spoken language (Auer 2004). Since DPRTs are predominantly used in conversation and dialogue, they constitute a spoken language phenomenon (Thurmair 1989) and hence they should be explored in spoken language. This will be discussed more in 1.2.1.2. Third, the inventory of DPRTs, and the range of their uses, varies across the varieties of German. This presents another reason to delimit the study to a specific dialect, and will be discussed in 1.2.2.4. I begin this section by providing an overview of where Miesbach Bavarian is spoken, and its basic syntactic properties.   1.2.1  Miesbach Bavarian 1.2.1.1  Dialect Bavarian itself is not a single dialect, but has distinct northern (Nordbairisch), middle (Mittelbairisch), and southern varieties (Südbairisch) (Renn and König, 2005). This study is limited to a specific Middle Bavarian variant spoken in Weyarn, a village   6 located in in Miesbach County just south of Munich, in the state of Bavaria. Bavarian is an East Upper-German dialect spoken in the German state of Bavaria, as well as in Austria. Small language islands where Southern Bavarian is spoken exist in Italy as well (Ethnologue, 2016). The map in Figure 1 shows Bavaria in the south of Germany. In the map below that (Figure 2), you find Miesbach County highlighted in dark grey and the location of Weyarn within Miesbach County highlighted in red. Miesbach County shares a border with Austria (Österreich) to the south.   Figure 1: Map of Germany  (©World Sites Atlas, 2008, reproduced with permission from World Sites Atlas)   7  Figure 2: Weyarn and Miesbach County in a map of Germany  (Hagar66, 2010) reproduced in accordance with CC-BY  1.2.1.2 Syntax In this section I give a brief grammatical sketch of Miesbach Bavarian (henceforth MB). In many respects, the clausal syntax of this dialect is similar to that described for SG (Weiss 1998). Most notably, it shares the verb second constraint (V2) for matrix clauses. Descriptively, V2 refers to the phenomenon where the finite verb can be preceded by one and only one phrase, as schematized in (3).  (3)    [XP  V …[VP …V]]   8 In matrix clauses, the finite verb moves from its underlying position within the VP to a higher position, generally assumed to be C in matrix clauses (den Besten 1983). In declarative clauses, a phrasal constituent (XP) precedes the verb. This constituent is often the subject, but can be any phrase, giving rise to the impression of relatively free word order around the verb in second position. In subordinate clauses, C hosts a complementizer, and the verb stays in its base position within the VP.  (4) a. [CP D’Martina  [C fahrt [IP mi'm    Radl [vPd’Martina [VP [V fahrt]]]]]]    DET  Martina      drives with.DET bicycle   “Martina is going by bike.”     b. …[C  wei [IP d‘Martina   mi'm  Radl [vP d’Martina [V  fahrt]]]]            COMP          DET.Martina  with. DET   bicycle       drives   “…because Martina is going by bike.”   Holmberg (2010) summarizes German V2 as the result of the two distinct properties in (5).  (5) a. A functional head in the left periphery attracts the finite verb.     b. This functional head wants a constituent moved to its specifier position.  Holmberg (2010:77a,b)  The second property, according to Holmberg, can be formalized as a generalised EPP-feature: It triggers movement and re-merge of any XP to the specifier of the head. This feature also blocks movement of any other category across the constituent that satisfies the EPP feature (Relativised Minimality, Roberts 2004). The result is V2 order.    9  The structure of German sentences is often presented within the ‘topological field model’ (the topologisches Stellungsfeldermodell Drach (1963). This model divides clauses into three areas: the Vorfeld ‘prefield’, Mittelfeld ‘middle field’, and the Nachfeld ‘post field’. The Mittelfeld is delineated by the span defined by the finite verb in C, which serves as the ‘left bracket’, and the verb in its VP-internal position, which serves as the ‘right bracket’. The commonly assumed mapping of the linearly defined topological fields onto the syntactic spine is illustrated in Figure 3.  Figure 3: The topological field aligned with the clausal structure  In clause types other than the declarative, the prefield, i.e. SpecCP, is either occupied by a wh-phrase as in wh-questions (6a), or remains empty as in polar interrogatives (6b), V1 exclamatives (6c), and imperatives (6d). This is summarized in Table 1, with examples (6a-d) below (more detailed discussion about clause types and their formal features will follow in Chapter 3).    10 (6) Clause type Vorfeld Mittelfeld Nachfeld 	 SpecC C IP, vP, VP   a.  Wh-interrogative 	 Wea who	 gibt gives   	 da  Luzie  a    Bussl DET Luzie DET kiss	 --	“Who is giving Luzie a Kiss?”	 	b. Polar interrogative    -- Gibt give da   Marinus  da    Luzie   a        Bussl DET  Marinus  DET  Luzie  DET     kiss -- “Is Marinus giving Luzia a Kiss?”  c. V1 exclamation -- Schaugt looks     des KIND3  scho     wieda   aus DET child    already again   out -- “(My), the way this child looks like again!”  d. Imperative -- Gib give am   Marinus a    Bussl   DET  Marinus DET Kiss -- “Give Marinus a Kiss.”  Table 1: Clause typing and the topological field  Material occupying SpecCP, as well as left dislocated constituents define the Vorfeld, whereas right dislocated material following V define the Nachfeld, the area after the final verb, as gestan in example (7).  (7) [SpecC  I [Cº hob [an Elias  beim Schlecka   dawischt hob]]] gestan     I have    DET  Elias at.DET eating.sweets   caught         yesterday    “I caught Elias eating sweets yesterday.”  Note that positioning a constituent in SpecC has effects on information structure.  Word order within the Mittelfeld is also fairly free due to the availability of scrambling.4 Scrambling, which I assume to involve (leftward) movement of XPs,                                                 3 A pitch accent will be indicated by CAPS throughout this dissertation. V1 exclamatives have an obligatory extra high pitch accent, usually utterance initially. This will be discussed in Chapter 3.  4 Several formal accounts of scrambling exist, with roughly two different approaches: (i) movement approaches and (ii) base generation approaches. See the introduction in Corver and Riemsdijk (1994) for an overview of both. Analyses making reference to movement are e.g. Krifka (1998), Büring   11 presents one of the main reasons for the difficulty of establishing the syntactic position of DPRTs. Specifically, it is unclear whether a given XP in the Mittelfeld is in its base position or in scrambled position. DPRTs in MB (like in SG) occupy the Mittelfeld, as illustrated in (8), where all possible positions for the DPRT fei are identified.  (8) Da  Hansi  hod  (fei)  gestan     (fei)  am   Basti (fei)    auf  da…     DET Hansi  has   fei    yesterday  fei   DET  Basti  fei      on   DET…      …oidn Wiesn   (fei) a Maß  spendiad   … old Oktoberfest  fei     DET litre.beer pay.for         “Yesterday Hansi bought Basti a beer at the old Octoberfest”   In addition to scrambling, MB displays several phenomena unique to the Bavarian dialect that do not occur in SG. One of these is negative concord, referring to multiple negative morphemes in one clause, which together express a single negation semantically. These multiple negations, introduced by a negative particle ned in (9) and a negated indefinite neamd, do not cancel each other out, but are interpreted as one single, sentential negation (Weiss, 1999, data from his Bavarian dialect).   (9)  I  han neamd   ned  gseng I  have  NEG.someone  NEG  seen  “I didn’t see anybody.”    (Weiss 1999: ex 2)  I will use negative concord as one of the diagnostics for the relative position of DPRTs in Chapter 6.                                                                                                                                        (2008), Neeleman and de Koot (2008). For accounts which do not assume movement in scrambling, i.e. which assume base generation, see e.g. Bayer and Kornfilt (1994), Fanselow (2003).    12  Another property of MB, less relevant for this dissertation, yet noteworthy, is the doubly filled Comp phenomenon. This refers to the possibility of having both the head and the specifier of CP filled in embedded clauses in Bavarian, which is banned in SG. I refer the interested reader to Bayer (1984), or Weiss (1998) for in-depth analyses of this phenomenon.  Also not directly relevant yet worth mentioning is complementizer agreement in Bavarian, where the complementizer shows agreement with the subject.   (10)  …wenn-sd  moang wieda  gsund  bist       if.2pSG  tomorrow again   healthy  you.are    “… if you are healthy again tomorrow.”   (Weiss 2002:2)  This morphology on the complementizer has been analyzed as inflection (Bayer 1984), agreement morphology (van Koppen 2005 for Dutch dialects, Fuß 2005), or as the spellout of a functional addressee feature on the subject pronoun (Gruber 2008).   A final property of MB that can be observed throughout the data presented here is the obligatory occurrence of determiners on proper nouns.   Since most of the features of MB discussed so far (except negative concord) are not directly relevant for the analysis of DPRTs, and since the clausal syntax, specifically the V2 constraint, the possibility of scrambling, and the distribution of DPRTs within the middle field are the same as in SG, the question arises about the reason for the particular focus on MB DPRTs. The reasons for this choice will be discussed in the next section.     13 1.2.2 Why dialect? Many linguistic investigations dealing with DPRTs in German (with the notable exceptions of Grosz 2005 and Bayer 2012) tacitly assume their data to represent a unified, more or less standard variety of the language. For example, Cardinaletti (2011:496) observes, “…[t]hat German MPs [DPRTs] have a restricted syntactic distribution is a well-established generalization… Some examples are German denn, doch, eben, etc.” Zimmermann (2004:543) focuses on “the German discourse particle” (all emphases added by ST). This idealization of the linguistic landscape does not capture the true state of affairs, however.   After introducing the methodology of data collection and data presentation conventions in section 1.2.2.1, I show reasons to restrict the investigation to one particular dialect, here MB.  First, a German dialect (as opposed to SG) is, so Weiss’ argument, acquired by children as L1, and hence constitutes their first language (Weiss, 2004) (1.2.2.2). Second, DPRTs are a spoken language phenomenon (Thurmair, 1989) and dialect is the natural language used in conversation for native Miesbach Bavarians. Hence dialect presents the most natural ‘habitat’ for DPRTs (1.2.2.3). Third, I present evidence for variation in DPRT inventories and use across dialects 1.2.2.4.   1.2.2.1 Methodology and data presentation Unless otherwise noted, all the data come from the author’s primary fieldwork with native speakers of MB, aged 50 and above. I also transcribed dialogues from Bavarian dialect movies, particularly M. H. Rosenmüller’s films Wer früher stirbt ist länger tot and Die Perlmutterfarbe. Rosenmüller, the screenwriter and director, is a Miesbach   14 native from the community of Hausham. In addition, data comes from Irgendwie und Sowieso, a popular Bavarian series of the seventies, written and directed by Bavarian native Franz-Xaver Kroetz. In addition, I based data on dialogues in Lena Christ’s Memoiren. All data from movie dialogues and books have been re-elicited and verified with a male and female speaker (both 61), residents of Weyarn and lifetime residents of Miesbach county to ensure consistent judgments from the MB dialect. Additional judgments from speakers from the communities of Weyarn and Holzkirchen were also taken into consideration. Data collection also included excerpts of naturally occurring discourse from conversations at family gatherings. Unless otherwise noted, all data are from MB.   An interesting observation is in order here; for one of the parts of this corpus, I had the opportunity to compare the movie script to the actual movie dialogues. The screenwriter of Die Perlmutterfarbe, Marcus ‘Hausham’ Rosenmüller, provided me with the original movie script, with the explicit comment that most DPRTs were added in the actual enactment of the scene.5 Preliminary findings showed that many more DPRTs occurred in the corresponding spoken sequences as compared to the written sequences of the script. This confirms the hypothesis that DPRTs serve to ground utterances within an immediate discourse context: they are interactive, and manage the common ground between interlocutors (in the sense of Krifka 2008). Hence they are expected to be a ‘live’ phenomenon, which cannot be encapsulated in a script.   Challenges posed by this approach, and a reason why only few scenes were investigated this way, were due to the difficulty in comparing the script with the                                                 5 M. H. Rosenmüller (pc): "Meistens ist es aber so dass so die "fei" und "hehs" und "ohs" erst beim Inszenieren dazukommen, aber schau mal".  ‘Mostly the ‘fei’ ‘hey’ and ‘oh’s are added in the enactment, but have a look.’   15 actual scenes, controlling for exact word choice, word order, and particularly important, controlling for intonation between the script and the movie dialogue. As will be discussed in more detail in Chapter 3, intonation plays an important role, which is not represented or controlled for in writing; this is also a reason why corpus data that is not thoroughly annotated for intonation cannot provide sufficient indications about the status of the host utterance within the discourse context. I now turn to the conventions for presenting the dialect data. Since DPRTs are highly context sensitive, I present most data with the appropriate discourse context. In cases where syntactic or other considerations are the primary purpose of presenting the data, and where the actual interpretation of the DPRT is secondary, a context may not be provided.   In order to indicate the (un)acceptability of any piece of data, I adopt the following convention: * indicates that an utterance is ill-formed. This is in line with common linguistic convention. However, I consider ill-formedness across a variety of contexts, taking into account not only narrow syntactic contexts, but also discourse contexts and dialectal variation. Therefore, to indicate dialect variation I will not use %, but instead I indicate a sentence-dialect pair such as <*, MB>, which should be read as ill-formed in Miesbach Bavarian. In contrast, <✔, MB> indicates that the sentence is well-formed in MB.   Similarly, to indicate ill-formedness within a discourse context, (i.e., infelicity) I will not use #, but instead I will use a sentence-context pair such as <*, Cx1>, which should be read as ill-formed in a particular context Cx1 (which is given along with the example sentence). In contrast, the same form may be acceptable in another context   16 Cx2. The latter sentence-context pair is indicated as <✔,Cx2>. I summarize this notation below.  (11) <*, MB>  ill-formed example in Miesbach Bavarian       à replaces %   <✔, MB>  well-formed example in Miesbach Bavarian    <*,Cx1>   ill-formed example in Context 1          à replaces #    <✔,Cx2.> well-formed example in Context 2         Lastly, I adopt the convention of placing the contribution of a DPRT in square brackets […] in a separate line below the translation line to give the reader a better sense of the English equivalent. DPRTs are often not directly translatable with corresponding UoLs in English (often the function of DPRTs is picked up by intonation in English). Take the example below with the DPRT fei (CAPS indicate accent).  (12) Des  is  fei ganz  schee vadrackt  DET  is  fei  whole  nice  tricky   “That is PRETTY TRIcky”   ‘[I believe you don’t believe that] that is quite tricky.’   This line with the DPRT contribution will only be provided in those cases where the function of the DPRT is under discussion. In other cases, I will resort to providing only a free translation in quotes “…”, as in (13).     17 (13) Des  is  fei ganz  schee vadrackt  DET  is  fei  whole  nice  tricky   “That is PRETTY TRIcky”  As for the orthographic rendition of the data, since there is no orthographic standard for written Bavarian, I use my own system for all data. In this dissertation I transcribe data based on my own intuition, using Standard German orthography as a base. I try to be phonetically as true as possible to the actual pronunciation, but also ignore many details that are irrelevant for the points I am making here.  1.2.2.2 Dialect as L1 Weiss (1998, 2002, 2004) argues that it is preferable to use dialectal data for formal linguistic investigation (cf. Auer 2004). He questions the suitability of standardized languages, such as Standard German (SG), for the investigation of natural language phenomena. Dialect, he argues, is the language acquired by children without special instruction, i.e. it is the first language (L1) for these children (Weiss 2004:1). Thus a dialect such as MB represents L1 for speakers growing up in areas and households where MB is spoken.  SG, as a standardized language can, like other standardized languages, be considered a natural language. However, due to normative, prescriptive pressures, it is also “partially invented” (Chomsky, 1995:51). The same applies of course to English.  Many prescriptive, ‘standard’ rules of English are artificial and do not reflect the reality of spoken language. Famous examples are for example the ‘no split-infinitive’, and preposition stranding rules. This is, of course, also a matter of register: the spoken variety adopted in formal settings is a different register than the spoken variety adopted in informal, familiar settings (cf. Auer 2009). The former is often a lot closer   18 to the written standard, whereas the latter is generally not represented in writing.6 The point here is that the formal spoken variety tries to emulate the written variety, which is subject to normative pressures. Weiss (2004) argues, that SG thus has properties that cannot be just ascribed to natural language change. As an example he refers to the disappearance of negative concord in SG due to normative pressure. However, in most German dialects including MB, negative concord is very much alive.  Additionally, SG was not acquired as a first language before the second half of the 20th century (Weiss 2004). It is in many respects to be considered an L2 for speakers, since it was not until formal schooling started at around age 6 that speakers were instructed in SG; this is also true for my consultants. Historically, SG was primarily intended for a standardized writing system. The Bavarian expression for speaking SG is telling in this regard: Noch da Schrift reen is to “speak like writing” in MB.  In sum, Weiss presents evidence that dialects provide a more privileged access to Universal Grammar, the subject of formal linguistic investigation (Weiss 1998, 2002, 2004). The dialect that serves as the empirical basis for this dissertation, Miesbach Bavarian, is my native language.   1.2.2.3 Dialect as spoken language phenomenon Dialect is predominantly spoken. If it is written, the writing is often a direct representation of a dialogue (as in novels by e.g. Lena Christ).  As such, dialect presents an ideal testing ground for DPRTs, which are predominantly a phenomenon of spoken language (Thurmair 1989). Conversation, as a back and forth between interlocutors, constantly builds, amends and revises the Common Ground (CG) (Stalnaker 1978, 2002). Since DPRTs are used to signal the epistemicity of the                                                 6 A point in case here is the lack of orthographic standard for written Bavarian.   19 discourse participants (i.e., the individual states of knowledge regarding the utterance they occur in), they are a means of Common Ground (CG) management in the sense of Krifka (2008). DPRTs are an aide in the establishment of mutual understanding, and therefore, they primarily feature in spoken dialogue. This is another reason for the preference of predominantly spoken data, as instatiated in dialect, over written (and often standardized) data.   1.2.2.4 Variation in DPRT inventory and use A final argument in support of delimiting the data to a dialect is the great variation in the use and inventory of DPRTs across German dialects. I focus on two main points in this section:   (i) Judgments regarding the use of DPRTs differ depending on the dialect spoken by the consultants, and  (ii) the lexical items used as DPRTs vary across dialects.   I discuss each of these points in turn. The following data illustrate the problematic assumption that German is a homogenous language. The judgments my consultants gave for this example differs from the judgments reported in the original source. Gutzmann (2010) presents JA in (14a) as well-formed (note that Gutzmann does not specify the variety of German the data stems from). In MB, accented JA is consistently judged ill-formed in declaratives with assertive force (14b).     20 (14) a.  A:   David riecht wie ein Zombie     “David smells like a zombie.”      B: <✔, SG>  David ist JA ein  Zombie        David  is   ja   DET zombie       “David IS a zombie.” (Gutzmann, 2010:15 ex 40)       b. A:  D’ Elli   schaugt aus wiar’a  Saubea    DET.Elli  looks    out  like.DET  male.pig     “Elli looks like a pig.”    B: <*, MB>  D’Elli   is  JA  aa    a     Saubea      DET.Elli  is  ja   aa    DET male.pig        B’:<✔, MB> D’Elli     IS   ja  aa  a Saubea       DET.Elli is   ja aa  DET male.pig        “Elli IS a little pig”    Note that JA is not ill-formed in all MB contexts. However, its use is restricted to commands with special emphasis, either expressed in an imperative as in (15a) or as a declarative with directive force, as in (15b).  (15)  a. Mach JA as  Fensta   zua!  make JA  DET  window  close   “Make sure to close the window!”  ‘[DO] Close the window!’    b. Du  machst  JA as Fensta  zua!    you  make   JA DET window  close     “You’re going to make sure to close the window!”   (15a) is an imperative clause type with directive force, whereas (15b) is a declarative   21 clause type with directive force.7 Crucially, however, in MB, the use of JA is ungrammatical in declaratives with assertive force.  Another example of dialectal variation in DPRT use comes from wohl. This DPRT is not used with the same function across all German dialects. According to Zimmermann (2004, 2011) the use of wohl in (16) expresses uncertainty on the part of the speaker, i.e. it has a dubitative meaning. The data presented to support this claim are not identified for the dialect, but presented as ‘German’.  (16)  wohl: <✔, SG >        Max  ist   wohl   auf  See   Max  is   wohl    on   ocean    “Max is probably at sea.”        (Zimmermann, 2011: 2013)  MB also has a version of wohl, realized as woi. It cannot be associated with a dubitative interpretation in the same context (17a).   (17) Heini:  Wo is’n da Maxl heid? I how’n no gar ned g’seng   “Where is Max today? I haven’t seen him yet”   Karl:  I woass  aa  ned   so recht.   I know  also      NEG  so right.   “I’m not so sure…"   a. <*, MB >    … dea is woi   an          Sää    ausse gfahrn    DET  is wohl on.DET  lake  out     driven                                                 7 For details on the connection between clause type, illocutionary force and DPRTs see the discussion in Chapter 3.    22   b. <✔, MB >    … dea  is wahrscheinlich  an        Sää ausse  gfahrn  DET   is probably         on.DET  lake  out      driven          “… he probably went out on the lake.”   c. <✔, MB >     … dea kannt an       Sää ausse gfahrn sei    DET could on.DET lake out    driven  be     “…  he might have gone out on the lake.”  Speaker doubt cannot be expressed in MB with the DPRT (17a), but is expressed with the adverb wahrscheinlich (‘probably’) (17b) or via a modal verb, as in (17c).    If wohl is used in MB, the speaker does not indicate doubt. This is illustrated in (18), a context where Tina is sure about her assessment that riding without a helmet on a motorcycle is dangerous, even for short distances.   (18) Cx: Tobi proclaims that he is going for a motorcycle ride. He is not    wearing a helmet. His sister Tina thinks this is reckless and dangerous   behavior, and urges him to wear a helmet. He responds that he only   wants to go for a short spin around the block, and it will be ok.   Tina: Du spinnst  woi total!           you spin      wohl totally.   “You are totally crazy!”  An example from my corpus further corroborates the claim that in MB wohl is not used to express doubt. Example (19) shows co-occurrence with ja. Ja, as I will argue in detail in Chapter 4, expresses speaker commitment to the propositional content of the host utterance. This interpretation would be incompatible with an expression of doubt. Nevertheless, ja can co-occur with wohl.     23 (19)  Cx: Students are visiting a radio station to learn about broadcasting. Alfred,   the radio host, mentions Jimmy Hendrix to the class.   Student:  Ähh wer is da Jimmy Hendrix?        “Um, who is Jimmy Hendrix?”   Alfred:        Jimmy Hendrix?!?   Na jetz glab  e's awa…    Jimmy  Hendrix PRT   jetz believe I.it aber   …ia  werds ja woi    no an  Jimmy Hendrix  kena oda?!  …you will  ja wohl still DET  J. H   know CONF    “Jimmy Hendrix!? Well, now it's too much! You must surely know J. H., right?”  We have seen now that the use of DPRTs is not uniform across all German dialects. This illustrates the need for a better understanding as to which specific dialect allows for which DPRT reading. To my knowledge this has not been undertaken up to this point.   A second argument for considering a specific dialect of German comes from the fact that some dialects have DPRTs that are not part of the inventory of other dialects. Bavarian, in general, and MB in specific, have the DPRTs fei and ebba8, not used in SG, or northern dialects of German (cf. Weiss 1998). Observe the following example of fei from an online version of a Munich based daily paper, the TZ.    (20) Cx: As of March 2014, the TZ offers a Bavarian written site. The editor   solicits submissions on the Internet homepage.                                                  8 I am not going to consider ebba in this dissertation. The following example from Weiss (1998) illustrates its use, which is close in function to the DPRT vielleicht. Note that Bavarian has an indefinite pronoun ‘somebody’ with the exact same form.  (i) Hod ebba   ebba   ebbs     gseng? has   ebba  somebody   something seen      “Did somebody see something?”     [Could it be that] somebody saw something?  (Weiss, 1998:97 ex 19a)   24   Unsa  neie  Bairisch-Seitn:   De      woi’ma   fei   ned alloa   macha…              our     new  Bavarian site      that   want.we  fei    NEG alone   make          “Our new Bavarian language site: we fei don’t want to do it by ourselves.”  …mia daadn  uns gscheid   gfrein,     wenn aa   Sie   a     bissl  middoa   daadn    we   would  us   properly  be.happy if       also you DET bit     with.do   would   “…we would be very happy if you’d help, too.”  We could speculate that these additional DPRTs simply add other meanings that cannot be expressed in other dialects, and that the meanings of DPRTs shared by both dialects are the same. This speculation is not quite right, as we have seen above. Moreover, there is preliminary evidence that the specific inventories of DPRTs in a given dialect may affect their interpretation. An example comes from an Upper Austrian-Bavarian variety, which has the DPRT leicht (Burton and Wiltschko 2015, cf. Grosz 2005 for Viennese), which does not exist in MB.9  When a speaker has current evidence to believe the proposition expressed in the host utterance is true, Upper Austrian (UA) speakers can use the DPRT leicht. Crucially, the speaker did not know that this proposition holds (i.e. was true) prior to the utterance situation. This is illustrated by (21), which is used if Maria has evidence on the spot, i.e. a current reason to believe that Paul has a dog now. In this context, the sentence-peripheral confirmational particle geu is ill-formed  (Burton and Wiltschko 2015; cf. Heim et al.                                                 9 leicht is likely grammaticalized from the DPRT vielleicht, which does exist in MB as well. However, note that vielleicht is ill-formed in rising declaratives in MB (i), whereas leicht is well-formed in this context, as illustrated by (ii).  (i) <*, MB > Du   host   vielleicht an  neia   Hund?                       you have   vielleicht DET new dog  (ii) <✔, MB> Host du    vielleicht  an  neia  Hund?                       have you  vielleicht  DET new  dog                     “Do you have a new dog?”                    ‘[Is it possible that] you have a new dog?’    25 2016, cf. Wiltschko and Heim 2016).  (21) Cx1:  Maria is going on a walk, and thinks of her friend Paul, whom she    hasn’t seen in a while. A few minutes later she runs into Paul, who has   a dog on a leash beside him. After saying hello, Maria says:   a.  Du   host leicht  an neichn  Hund?    you have leicht  DET new     dog?     “You have a new dog?”   b. <*, UA> Du   host an  neichn Hund,  geu?    you have  DET  new     dog,      CONF     “You have a new dog, eh?”  The existence of the DPRT leicht blocks a particular interpretation of the confirmational particle ‘geu’, which is used in UA to express only that the speaker has previous reason to believe the proposition expressed in the utterance. This is illustrated in a context Cx2, where Maria is told that Paul has a new dog, providing her with previous reason to believe that p.    (22) Cx2: Maria hears from her neighbor that her friend Paul has a dog now. A   few hours later she runs into Paul on the street, who has a dog on a    leash beside him. After saying hello, Maria says:   a.  Du   host   an  neichn  Hund,  geu?    you have   DET    new      dog     CONF     “You have a new dog, eh?”   b. <*,UA> Du host    leicht  an neichn   Hund?      you have   leicht  DET new       dog?  In MB, the equivalent of Upper Austrian geu is gä. Gä is well-formed in both contexts, Cx1 and Cx2. This suggests that the existence of leicht in Upper Austrian covers part of the functional use spectrum of geu, which is covered by gä in MB. In   26 essence, the existence and function of leicht in this dialect appears to block a function of geu/gä that is reported for MB.   Lastly, variation in DPRT inventory also includes the existence of clitic DPRT forms in MB. Notably, MB has clitic versions of some DPRTs, specifically ja and its clitic counterpart ‘a and denn with its clitic counterpart ‘(a)n (Weiss 1998). These are much more restricted in their distribution than their full counterparts (see Grosz 2005 for an in-depth discussion of the Viennese German equivalent ‘dn).   (23) a. How’e’n   da   des ned  gsogt?       have.I.(den)n  you DET NEG said      “Didn’t I tell you that?”    b. How’e’ da denn des ned  gsogt?       have.I.you denn DET NEG said    c. *How’e denn da   des ned  gsogt?          have.I.denn  you DET NEG said    d. *How’e’ da’n  des  ned gsogt?         have.I.you.(den)n  DET NEG said    e. *How’n’e’ da  des ned  gsogt?         have.(den)n.I.you DET NEG said   Clitic versions of DPRTs are not reported for SG. Whereas the full forms can follow a full DP/pronoun in the middle field (24a), the clitic forms have to cliticize onto C, following clitic pronouns (24c).10  (24) Mia is  grod  eigfoin … me is  just    fall.in                                                      10 I will not deal with clitic data in this dissertation, but see Grosz (2005) for a discussion of the Viennese interrogative DPRT clitic ‘dn, and Bayer (2012) for the Bavarian counterpart ‘n.    27     a. … dass  d’Katl  ja  scho lang Geburtstag ghabt  hod   … that   DET.Katl  ja  already long  birthday has had    “I just remembered that Katl already had her birthday a while back.”     b….dass’a d’Katl     scho  lang  Geburtstag ghabt hod     … that (j)a DET.Katl  already  long   birthday  has had    "I just remembered that Katl already had her birthday a while back."      c.*… dass  d’Katl’a  scho lang Geburtstag ghabt hod        …that   DET.Katl.(j)a already long birthday has had  To summarize, the observation that different regional versions of even one general dialect (here Bavarian) show variation in inventory and use of DPRTs, as illustrated by wohl, fei, and Upper Austrian leicht, as well as the existence of clitic versions with different distributional restrictions, indicates the need to restrict data to sources from a delineated dialect area.   1.2.3 Summary In this section I discussed reasons why I chose to restrict the data in this dissertation to the MB dialect of German. First, I summarized arguments from Weiss (1998, 2004) who points out the preferred status of dialect for formal linguistic investigation, since it is in many regards L1, the first language acquired by speakers. Second, dialect as a spoken language phenomenon is ideally suited to the investigation of DPRTs, which also feature predominantly in spoken language. Third, I argued that there is a great variety in the inventory and use of DPRTs across dialects. All these arguments point to the preference of a delineated regional dialect. I refer the interested reader to Weiss (2004) and Auer (2004) for further in-depth arguments in support of this point.     28 1.3  Theoretical assumptions and background In this section I introduce some of my assumptions regarding the models of grammar and pragmatics. I begin with a sketch of my syntactic assumptions.  1.3.1 Syntax I locate the claims presented in this dissertation within a framework that takes as its point of departure the assumption that linguistic competence is based on basic principles provided by a Universal Grammar, which all human languages share. Differences in languages arise via differences in certain language specific parameters. This Principles and Parameters approach to language and linguistic variation has undergone various instantiations since its initial conception in Chomsky (1981). Two such current instantiations of the Principles and Parameters framework are the Minimalist Program (Chomsky 1995) and Cartography  (Cinque 1999, Rizzi 1997). The former assumes bare phrase structure, eliminating functional categories as primitives. Clauses are built from the bottom up, by binary merge of linguistic items. The latter approach, Cartography, assumes a wide range of functional categories, which are often introduced to capture the function of a linguistic item. Clauses in this framework are built by merging linguistic items within a pre-existing syntactic structure. Critics of this approach admonish the proliferation of function-based projections. I refer the interested reader to Ramchand and Svenonius (2014) and Wiltschko (2014) for in-depth arguments and comparisons of the two models.  An instantiation for an in-between model is the Universal Spine Hypothesis, developed by Wiltschko (2014). I adopt this model for the analysis presented here. The Universal Spine Hypothesis, as the name suggests,   29 recognizes a universal syntactic spine, which consists of a series of hierarchically organized, functional categories κ. Each of these functional categories κ fulfills intrinsic abstract grammatical functions, such as linking, anchoring, or classifying. The abstract functions of the spine are based on general cognitive functions (cf. Ramchand and Svenonius 2014). This model also captures the long observed parallelism between the nominal and verbal domains. I will introduce the Universal Spine Hypothesis in more detail in section 6.3, since a discussion of the syntactic analysis of DPRTs will be delayed to Chapter 6. Much of the syntactic discussion in Chapter 6 is inspired by the seminal work of Speas and Tenny (2003), who take Ross (1970) as a point of departure. They argue for the need to introduce a syntactic representation of the speech act and its participants. This approach is further pursued and formalized in much subsequent works from which I draw here, most notably Hill (2007), Haegeman and Hill (2013) and Haegeman (2014).   1.3.2 Pragmatic model The DPRTs discussed in this dissertation allow interlocutors to position an utterance within a context while providing information about a specific mental model; they reflect a speaker’s or addressee’s mental model of each other (cf. Green 2006). This is the reason why I investigate DPRTs in the context of a syntax pragmatics interface. DPRTs provide pragmatic information, which in Green’s (2006:407) terms “is ultimately indexical information, that is, related to indices for speaker, hearer, time, and location of an act of uttering something […]”.   Syntax plays a role on the one hand, in providing the relationship between the index (e.g. discourse participants) and the content of the act of uttering something (i.e.   30 the content of an utterance, which I will call p across all clause types for exposition). The relation between the index and its content can be viewed as a predicative relation. On the other hand, the aspect placing DPRTs into the realm of pragmatics has to do with their relationship to the index. As we will see in the course of this dissertation, the model alluded to in the previous section and developed in detail in Chapter 6, advocates for the ‘syntacticization of discourse’ (Haegeman and Hill 2013), and places some of this pragmatic burden onto the syntactic spine.   A part of the analysis presented in this dissertation will be framed within speech act theory and is based on a model of conversations and discourse advanced in Clark (1996) and Clark and Schaefer (1989). In particular, I assume that utterances (and their contents) do not enter a common ground shared between a speaker and addressee by virtue of being uttered; rather, they need to be grounded. Grounding, introduced in Clark and Schaefer (1989: 262) is based on the condition that a speaker and addressee mutually believe that the addressee has understood what the speaker meant with an utterance, in a way sufficient for the purposes of the conversation. This leads to a two-step process, where a speaker presents some utterance, and the addressee accepts the utterance. I formalize this separation further, by adopting terminology from Farkas and Bruce (2010) and Malamud and Stephenson (2015), which in many ways builds on the proposal in Clark and Schaefer (1989) and Clark (1996). In particular, the two-step model of grounding utterances presupposes a separation of a formal component tasked with representing the speaker’s belief, and the addressee’s belief. This will be introduced in detail in Chapter 3.  DPRTs are highly context sensitive. However, it is not a priori clear what context sensitivity means. Bach (2012) argues for the need to ask whether it is the content of the host utterance that varies from one context of utterance to another, and whether it   31 does so because of specific features of the context itself. In the case of indexicals, that is, with expressions that have no fixed reference, such as ‘I’ and ‘today’, context can play a more direct, semantic role. Indexicals are context sensitive in that their function depends on the context in which they are used. DPRTs, as I argue in this dissertation, are context sensitive in that their functional range (which is what I will call their individual, sometimes changing interpretational contribution to a host utterance) can be attributed to context. However, DPRTs do not depend on context. A given DPRT can be compatible (or not) with a context, and context can influence the function of a compatible DPRT, but context does not determine this function.  DPRT have received a variety of names that reflect insights about their pragmatic function. DPRTs are known in German as Modalpartikeln ‘modal particles’. Under a definition of modality as 'relational', establishing a connection between what is known by the interlocutors and the proposition expressed in the utterance, we can understand DPRTs as modal in the sense that they relate the proposition to the epistemic state of each discourse participant. I show in this dissertation that a fine-grained distinction can be established, corroborating the idea of DPRTs as modal particles: DPRTs can be classified into those that relate the utterance to an addressee, a speaker and a contextually defined discourse participant O.   Another term used to refer to DPRTs is Abtönungspartikeln ‘toning particles’ (e.g. Weydt 1969). This can be understood as another core property of discourse particles that will be explored here, namely their ability to amend (i.e. tone down or emphasize) the illocutionary force associated with a given utterance. This brings us back to pragmatics, into the realm of speech acts and their associated conditions (Searle 1969). These will be further contextualized in Chapter 5.     32 1.4  Proposal The main empirical goal of this dissertation is to establish the existence of three types of DPRTs, each expressing a different discourse participant's attitude. I propose that each type of DPRT is associated with a different syntactic layer. This presents the analytical contribution of this dissertation. That is, I follow recent syntactic proposals that argue for the need to expand the clausal spine to include projections representing the speech act participants S and A and their epistemic states (e.g. Haegeman and Hill 2013, Haegeman 2014, Heim et al. 2016, Wiltschko and Heim 2016).   My initial research on DPRTs was spurred by the commonly made observation that these UoLs are pervasively multifunctional. A main question triggered by multi-functionality is whether DPRTs are somehow inherently specified as such: can DPRTs be analyzed as a unique sound-meaning pairs with specific categorial labels that do justice to their function? Categorized as such, DPRTs would differ from other, homophonous particles that would be analysed as fully distinct sound-meaning pairs with other category labels. For example, the UoL ja can be used as a response marker (‘yes’), as a discourse particle, and as a discourse marker. Hence it appears to have three functions that may be attributed to three distinct but accidentally homophonous UoLs, as schematized (25).11  (25)   f1:  UoL1    f2:  UoL2    f3:  UoL3     etc…                                                  11 These UoLs of course don’t necessarily have to be completely different. But even under the assumpotion of principled polysemy, most researchers assume DPRTs and their counterparts in other word classes to be distinct items, and featurally specified differently. I discuss this in more detail in Chapter 2 (2.2.1).   33 The Universal Spine Hypothesis provides us with an alternative to positing massive accidental homophony (and even to the idea of related yet distinct items), which I explore here. In effect, I adopt a hypothesis that endows a UoL with a meaning that is general enough to allow more fine-grained functions to arise in the relevant contexts. Specifically, I argue that the notion of a discourse particle is a construct (at least in MB). I talk about a DPRT when we refer to a (language specific) function f of a UoL in a specific syntactic context CxSYN. Other language specific functions (e.g. as response particle) are represented as another function fresp  of a UoL in another specific context CxSYN2.  (26)   fDPRT = UoL +CxSYN1   fresp = UoL +CxSYN2  This dissertation focuses on the first function, the DPRT function. I establish that DPRTs are not homogenous in their property of expressing attitude; specifically, I establish in Chapter 4 a more fine-grained approach, showing that DPRTs do not only express speaker attitude. Rather, I show that they can be separated into expressing three epistemic stances; into those that express (i) addressee belief (A-orientation), (ii) speaker belief (S-orientation), and (iii) those which relate the host utterance to a contextually determined participant other (O-orientation). I show in Chapter 6 that the fine-grained orientations of DPRTs are conditioned by syntax (Cx:syn), that is, by the specific functional projections with the abstract category label κ that these DPRTs associate with (27).     34 (27)   A-orientation= UoL+ κ: addressee  S-orientation= UoL+ κ: speaker  O-orientation= UoL+ κ: other  The functional range of a single DPRT, i.e. the variety of interpretational nuances which can arise from one single DPRT in different contexts, are shown to arise from the contribution of the context. This includes the discourse context (Cx: disc) as well as the situational context (Cx: sit), and is discussed in detail in Chapter 5.   The overarching function of Miesbach Bavarian DPRTs is uniform in that they share the property of being relational (cf. Diewald 2013); they establish a relation between the host utterance and the larger discourse context by grounding the proposition p to a discourse participant. This relational property likens them to predicates, which also relate two arguments. To accommodate this grounding function in the spine, I adopt recent proposals to extend the Universal Spine, introducing an Extended Universal Spine in Chapter 6. Each of the domains introduced by the Extended Spine has a relational function as illustrated in Figure 4. Figure 4: The relational functions of the Universal Spine   35 I show further that MB DPRTs display variation in how grounding proceeds, that is, they differ in how they relate the host utterance to the discourse situation. With this extended spine we can understand the classification of DPRTs introduced here into DPRTs that are addressee-oriented (A-oriented), those which are speaker-oriented (S-oriented), and those which are other-oriented (O-oriented).  This subdivision of DPRTs into these functional groups suggests that the grounding layer GroundP needs to be subdivided, too, into a projection representing A and a projection representing S. I argue for this in Chapter 6, where I also propose that O-oriented DPRTs associate with CP.    1.5  Roadmap The dissertation is organized as follows. In Chapter 2, I give a brief background of the properties that have been uncovered so far in the extensive literature on DPRTs. These properties, however, do not allow us to establish DPRTs as a natural class, since not all items considered DPRTs conform to all of them. One well-known property of DPRTs, and certainly a necessary property, is that DPRTs are attitudinal in some ways. Other properties, such as the obligatory occurrence in the syntactic middle field, their propositional scope despite the syntactic integration, and their sentence type dependence will also be discussed.  In Chapter 3, I introduce the background on the two necessary ingredients that allow me to derive DPRT function: the context, and the UoL serving as the basis for deriving DPRT function.    36 Chapter 4 provides evidence for the need to further distinguish the notion of attitude. My empirical basis draws on case-studies of five MB particles: fei, doch, ja, eh, and jetz. I show that MB DPRTs express epistemicity, i.e. the epistemic stance of a discourse participant. Further I show that we can distinguish three classes of DPRTs, each relating the host utterance to a different discourse participant, thereby indicating their respective epistemicity. fei and doch show that DPRTs can relate the host utterance to the addressee; they are A-oriented DPRTs. ja shows that DPRTs can relate the host utterance to the speaker; it is S-oriented. eh and jetz show that DPRTs can relate the host utterance to a contextually determined discourse participant 'other'; they are O-oriented. Chapter 5 shows that the wide range of interpretations associated with DPRTs, their functional range, can also be derived on principled grounds (cf. Abraham, 1991a,b). A UoL with DPRT function can derive functional range due to the context. I show that it is the larger context (Cx: disc and Cx: sit) of the DPRT that contributes the fine-grained interpretational differences. DPRTs are compatible with a variety of contexts which all have in common that they include aspects of their unique core function. In this way, the larger context contributes to the meaning just like the DPRT (and its host utterance) contributes meaning. Chapter 6 aims at providing a syntactic account for the three orientations uncovered in Chapter 4. I adopt previous proposals to expand a core syntactic spine to include a ‘speech act’ level projection (e.g. Hill 2007, Abraham 2012, Coniglio and Zegrean 2012, Zu 2013, Thoma 2014, Lam 2014, Haegeman 2014, Heim et al. 2016, among others). I integrate this with Wiltschko’s (2014) Universal Spine, and analyze the case studies from Chapter 4 within this extended Universal Spine. I propose that   37 the function of the particles in the three classes is derived, and that DPRTs are best analyzed as syntactic constructs.   Chapter 7 summarizes and concludes with a brief look at open issues and further research.   38 Chapter 2: Discourse particle properties  2.1  Introduction Until the 1960s, DPRTs were largely considered superfluous, unnecessary filler words, as Reiners (1944) puts it: Läuse Im Pelz unsure Sprache ‘lice in the fur of our language’. Based mostly on normative approaches to grammar, which took written language as the ideal object of investigation, DPRTs, as a predominantly spoken language phenomenon, did not receive much attention. This changed during the 1960s and 1970s, when the field of linguistic pragmatics began to emerge, and led to theoretical approaches such as speech act theory, discourse analysis, and conversation analysis. All of these came with their respective approaches to DPRTs, leading to the manifold ways in which DPRTs have been described and analyzed over the last 50 years (cf. Kwon 2005 for more details on some of these approaches). The first investigations into DPRTs, such as Krivonosov (1977), Weydt (1969, 1979), and Helbig (1988), were restricted to describing the general meaning of individual DPRTs. One of the first, more formal approaches to DPRTs is that of Doherty (1987), who argued that the primary function of DPRTs is found in the expression of epistemic attitudes of the speaker. I will follow in Doherty’s footsteps in this dissertation and present an analysis of DPRTs that sees as their main function the expression of discourse participant epistemicity.  More recently, researchers from the generative tradition have approached and investigated DPRTs from the point of view of their syntax and semantics. For syntactic approaches see Grosz (2005), Coniglio (2006, 2007, 2009), Bayer (2008, 2012), Bayer and Obenauer (2011), Cardinaletti (2011), Coniglio and Zegrean (2012),   39 Struckmeier (2014) among others. For semantic analyses see Kratzer (1999, 2004), Zimmermann (2004, 2011), Grosz (2005, 2010a), Gutzmann (2008, 2009), Kaufmann and Kaufmann (2012), among others.  In addition, Werner Abraham has been at the forefront of formal discourse particle research since the early eighties, offering various analyses both in functional and generative grammar frameworks (e.g. Abraham 1986, 1988, 1991a. 1991b, 1991c, 1996, 2000, 2001, 2009, 2010, 2012). Many aspects of the analysis presented here are indebted to and build on his insights, as well as the insights offered by the above-mentioned particologists.  Despite the extensive research on DPRTs, we do not, however, have an exhaustive list of necessary and sufficient conditions to classify DPRTs to date. It is debated whether they constitute an (emergent) word class, with proponents like Meibauer (1994), Diewald (2006, 2013) and Stuckmeier (2014), though others, like Helbig (1988) and Thurmair (1989), argue against this view. With this unresolved debate on their categorial status, it also comes as no surprise that there is no real consensus on the number of linguistic objects that should be considered DPRTs in German. Proposed numbers range from 15 up to 40 (Diewald 2013), with new DPRTs continuously being added to this inventory (e.g. for SG wieder Pittner 2009, for Bavarian fei Thoma 2009). Their multi-functionality, their apparent optionality, sentence type restriction, distribution, and their categorial status, are all well documented, yet analytically unresolved properties. This dissertation aims to contribute to our understanding of these properties. In what follows, I discuss these properties in more detail, and illustrate them on the basis of MB.   40 2.2  DPRT properties and the questions they raise DPRTs are characterized by a variety of empirical properties. These properties raise a host of analytical questions, which will be presented in the next section. The problems addressed include multi-functionality (2.2.1) and their reported optionality in a sentence (2.2.2). In 2.2.3 I show their linear distribution within the clause. Their phrasal status is addressed in 2.2.4. In section 2.2.5 I address the propositional scope of DPRTs. Sentence type restriction will be discussed in 2.2.6, whereas the meaning contribution of DPRTs is addressed in 2.2.7 I conclude this Chapter in section 2.3.   2.2.1  Two types of multi-functionality  Probably one of the most mentioned, but least discussed properties of DPRTs is their pervasive multi-functionality (Abraham 1991a,c). DPRTs are multifunctional in two ways. First, DPRTs have counterparts in other word classes, with different functions. This will be discussed as multi-functionality across categories. Second, DPRTs are multi-functional within their DPRT use. That is, a DPRT such as doch can have a functional range, with interpretational differences delimited by context. Examples are the so called ‘corrective’ use of doch versus the ‘reminding’ use of doch, and the ‘shared knowledge’ use of ja versus the ‘surprise’ use of ja. This will be further illustrated and discussed below as the functional range of DPRTs.   Multi-functionality across categories:  DPRTs in MB are generally multi-functional. That is, DPRTs have counterparts in other categories.12  This is illustrated                                                 12 Of the five particles discussed here in detail, all besides fei are multi-functional. fei along with one other DPRT (hoid; SG halt) in MB is mono-functional. The mono-functionality of fei will be briefly addressed in Chapter 7, where I deal with grammaticalization of DPRTs.   41 below based on jetz, a UoL, which can function as a DPRT, as a temporal adverb, as an adjective, and as a discourse marker.13  (1) UoL Function f  jetz f1:  discourse particle f2 :  temporal adverb f3:  adjective f4:  discourse marker  Multi-functionality is pervasive with DPRTs, and can be treated two ways; in terms of meaning maximalism or in terms of meaning minimalism (cf. Abraham 1991a, Zimmermann 2011 and references therein). According to a meaning maximalist approach in its most extreme instantiation, multi-functionality arises due to accidental homophony. UoLs with differing functions are assumed to be synchronically unrelated – each function is associated with a distinct lexical entry, with different distributional, phonological and semantic properties. This is illustrated in (2) based on jetz.  (2) Meaning maximalist approach to multi-functionality of jetz UoL1 jetz: f1   à 	discourse particle UoL2 jetz: f2     à 	temporal adverb UoL3 jetz f3     à 	adjective UoL4 jetz f4    à 	discourse marker                                                    13 The differentiation between DPRTs and discourse markers is not particularly clear, since there is a wide range of naming conventions in the literature (see the introduction in Fischer (2006), as well as Diewald (2013) for ways of differentiating). The assumption I adopt here is that DPRTs are distinct from discourse markers due to (i) their unique distribution in the middle field and (ii) their function. These criteria are specific to MB (German), since discourse markers and discourse particles differ in their distribution across languages.   42 The most prominent advocate for this approach is Helbig (1988). But more recent approaches toward the syntax and semantics of DPRTs can also be seen to fall under this category. Bayer and Obenauer (2011), for example, consider the interrogative DPRT denn to be endowed with special syntactic features, which account for its restriction to polar and wh-interrogatives (see Coniglio and Zegrean 2012 for a similar proposal for Italian and Romanian DPRTs). Denn can also function as a conjunction, however. The conjunction denn under this proposal is presumably not equipped with the same syntactic features as the DPRT denn, since their distribution and interpretation is different. Two UoLs with two separate feature specifications are the result.  Coming from the semantic side, Grosz (2010a) approaches DPRTs similarly to how Bayer and Obenauer (2011) approach their syntax. Instead of proposing special syntactic features that distinguish DPRTs from their homophones, Grosz (2010a) argues that the DPRTs JA, ruhig, and bloß should be considered as special modals, each with their own lexical entry. These lexical modal features then restrict their distribution and interpretation. He states: “Note that JA  (with obligatory stress) is homophonous with ja ‘yes’ and the unstressed discourse particle ja discussed by Kratzer (1999); bloß is homophonous with bloß  ‘only’ and ruhig with ruhig  ‘quietly’. These alternative readings are irrelevant for the purpose of this paper.” (Grosz, 2010: Fn2). This quote captures the general approach to DPRT multi-functionality, even under a meaning maximalist hypotheses; variable readings are seen as homophonous, and often considered unrelated, irrelevant or are simply excluded from analysis.  Meaning maximalist approaches contrast with meaning minimalist approaches, according to which differences in interpretation can be derived from an underlying   43 core interpretation, provided by a sound π meaning Σ bundle < π, Σ> (which I call UoL, following Wiltschko 2014). Under this hypothesis, it is the syntactic context (Cx SYN) that DPRTs appear in that is responsible for the differences in function.   (3) Meaning minimalist approach to multi-functionality for jetz  UoL:   jetz     Cx SYN 1   f1   à   discourse particle    Cx SYN 2   f2    à   temporal adverb    Cx SYN 3   f3    à   adjective    Cx SYN 4   f4    à   discourse marker  There are several attempts that seek to derive (at least a subset of) the possible functions of some DPRTs from one underlying core meaning (Abraham 1991a, 2001; Krifka 2013; Thurmair 1989). I will follow in this vein here, and whereas I make suggestions, I will not be able to address in detail this aspect of multi-functionality in this dissertation. The next multi-functionality aspect, however, is at the core of Chapter 5, and will be introduced next. Note, in passing, that I view both types of multi-functionality as instantiating the same phenomenon: the function of a given UoL is largely contextually derived where context can be viewed as the narrow syntactic context or the larger discourse and situational context.  The functional range of DPRTs:   Meaning maximalism originally intends to refer to the fact that a DPRT such as ja and doch can have a variety of interpretations depending on the host utterance, discourse context, and accenting pattern (unaccented or accented). Thus, there is a functional range associated with a given DPRT. For the particles discussed here (fei, doch, ja, jetz, and eh), a variety of functions have been reported (cf. Franck 1980,   44 Helbig 1988). I will address in detail the functional range of ja, doch, and fei in Chapter 5, but introduce the general problem below by illustrating with a subset of the functional range of ja.  (4)  a. Cx1: Martl asks who is coming to Dani’s party. Alex responds:     D’Sonja konn  ned kemma…   DET Sonja  can NEG come.        …de  muass’se  ja  um  ihre Zwilling kimman.   …she  has.self  ja  for  her  twins  care    “Sonja can’t come.  She ja has to care for her twins.”   (Modeled after Kratzer 2004: ex 10)     b. Cx2: I am sitting in the coffee shop, working on my thesis. I check the time,   and realize how late it is. I mutter to myself:   I muass ja in 5 Minuten in da Abteilung sei!  I  must  ja  in  5  minutes in DET  department be   “(Gosh), I gotta ja be in the department in 5 minutes!”       (Modeled after Kratzer and Matthewson, 2009: ex 10)  In (4) a, the function of ja, according to Kratzer (2004) is to refer to mutually shared knowledge, often framed as a Stalnakerian presupposition (cf. also Kaufmann 2004, Kaufmann and Kaufmann 2012). In (4) b, however, the function of ja cannot refer to mutually shared knowledge; there is no addressee present in the discourse context, as observed by Kratzer and Matthewson (2009). In this context, ja expresses surprise (Kratzer 2004, Lindner 1991). The (partial) functional range of ja is summarized in (5).     45 (5) Functional range of ja UoL: ja    Cx 1   f1   à   mutually shared knowledge    Cx 2   f2    à   surprise   The question that arises here is whether the functional range of ja is lexically encoded or derived by contextual factors. I take up this question in Chapter 5.   2.2.2  Obligatoriness DPRTs are not considered obligatory for the grammaticality of a sentence in MB (cf. also Thurmair 1989, Kwon, 2005 for SG). This is illustrated below. The host sentence by itself, without a context, is well-formed without a DPRT (6a), with the DPRT eh (6b), and with the DPRT ja (6c), and with both DPRTs (6d).  (6) a. Es findts   nix.  you find.2PL nothing   “You're not gonna find anything.”  b. Es findts   eh nix.  you find.2PL  eh nothing  “You're eh not gonna find anything.”  c. Es  findts    ja nix.   you  find.2PL  ja nothing  “You're ja not gonna find anything.”   d. Es  findts    ja  eh  nix.  you  find.2PL ja  eh nothing  “You're ja eh not gonna find anything.”    46 However, consider what happens if we place these examples in a particular context. Take (6d), where both DPRTs co-occur. In the context below, the presence of these DPRTs is obligatory. Without them, the utterance is not possible in this context.   (7)  Cx1:  Alexander is being accused by his classmates of stealing a special kind   of paint. His classmates interrogate him, he denies. At the end, one of   them suggests to look into his bag. Alexander reacts:    Vo mir aus…   “As far as I’m concerned…  a.  <✓, Cx1>… schaugts in mein Schuiranzn, findts      ja eh nix.            look.2PL  in my    school.bag  find.2PL  ja eh nothing    “…look into my bag ‚you're ja eh not gonna find anything anyways.”  b. <*, Cx1>… schaugts in mein Schuiranzn,  findts   nix.       look.2PL in my  school.bag  find.2PL nothing   This suggests that, once we take context into consideration, DPRTs are no longer optional. Furthermore, there are several sentence types, such as insubordinate clauses, where DPRTs, according to Thurmair (1989) are  “almost obligatory”, (cf. Zimmermann, 2004, Kwon 2005, Struckmeier 2014). Insubordinate clauses are typically characterized by two seemingly contradictory properties:  (i) they are used as independent utterances (i.e., what appears to be a root-context), and  (ii) they have the form of a subordinate clause: they are introduced by a complementizer.     47 Interestingly, insubordinate clauses require the presence of a DPRT. This is shown by the contrast in (8). The subordinated clause (dass’d brav bist bei da Oma) is well-formed when it is embedded under a matrix clause (8a) and when it is used with a DPRT (8c); but not in the absence of the DPRT. That is, a clause introduced by a complementizer requires a DPRT to be well-formed as a stand-alone utterance, which in this case serves as a command.   (8) a.  (I mecht)  …dass’d   brav bist bei da  Oma   I want        …that.you good are at DET grandma    “I want that you’re good at Grandma’s.”  b. *  Dass’d  brav  bist  bei da  Oma   that.you  good  are  at   DET grandma    “…that you’re good at Grandma’s.”  c. ✔   Dass’d    fei   brav bist bei da  Oma    that.you  fei   good are  at   DET grandma     “Be good at Grandma’s.”   The data in (7) and (8) raise two questions:    (i) Why are DPRTs considered optional?   (ii) Why are DPRTs obligatory in some contexts?  I provide answers to these questions in Chapter 6.      48 2.2.3  Distribution As shown in (9), DPRTs like fei occur in the middle field.  (9) (*fei) Da  Hansi (*fei) hod (fei) am  Basti (fei) a     Maß      (?fei ) spendiad (*fei)     fei  DET Hansi    fei   has  fei   DET Basti  fei   DET litre.beer  fei    pay.for  “Hansi bought Basti fei a beer.”   We have seen in Chapter 1, that in German, scrambling can alter the base position of constituents. Hence, the surface position of a UoL, i.e. the way it is linearized, is not always indicative of its interpretive position, the place it associates with the spine (details in Chapter 6). While the linear position of DPRTs in the middlefield is clear (cf, (9)), the question about their syntactic association, that is, structural position in the spine is less clear. Nevertheless, the position of DPRTs in the syntactic string has been used as structural diagnostic (e.g. Diesing 1992).   This position has also prompted analyses that propose that DPRTs occupy the specifiers of functional projections in a highly articulated IP domain (Coniglio 2005, 2008; Grosz 2005, 2007), whereas Bayer and Obenauer (2011), propose that DPRTs are merged in a Particle Phrase in the middle field above vP. This open question about the position in the spine DPRTs associate with (which, as I mentioned, can be different from their overt position), immediately leads to the next open issue, namely the phrase-structural status of DPRTs as heads or modifiers. I turn to this question below.  2.2.4  Heads or phrases It is an open issue in DPRT research whether DPRTs are best considered as heads or phrases. Below I discuss some of the empirical problems associated with this issue.    49 DPRTs cannot be modified. In this respect, they contrast with adverbs. Unlike the adverb bestimmt, the DPRT ja cannot be modified with ganz as shown in (10).   (10) a. Da  Hans is  ganz   bestimmt  scho     do       DET  Hans  is  completely surely       already there      “Hans is certainly already there.”    b.  Da  Hans is *ganz  ja   scho     do       DET Hans  is   completely ja  already  there   c.  Da Hans is ja scho    do       DET Hans  is ja  already  there       “Hans is ja already there.”  This resistance to modification has led researchers to assume that DPRTs are heads (Abraham 1995, cf. Cardinaletti 2011, Coniglio 2005, Grosz 2005, cf. Poletto 2004 for Italian dialects).  DPRTs cannot be moved to SpecCP. Another possible indication of the head status of DPRTs is the fact that they cannot be questioned. This has been taken as an indicator that DPRTs lack constituent value (Weydt 1969).   (11) a.  Des  is fei schwaar.   DET  is fei  heavy    “That is fei difficult.”   b. A: Wia is des so?          “What's it like?”         B: Schwaar / *fei        “difficult” /   fei   50 Also unlike phrasal constituents, DPRTs cannot occur in Spec C; this again can be taken as an indication of their head status.   (12) a.  Do is fei   mei Mama.   there  is fei   my  mom              b. Mei Mama is  fei do   my  mom   is  fei  there    c. *fei is do mei Mama   DPRTs can be combined but not coordinated. DPRTs, unlike other phrases (for example APs), cannot be coordinated.  (13) *Do     is fei und eh mei  Mama   there  is  fei and  eh my  mom   This inability to coordinate, unlike adverbs, is often taken as an indirect indication of their head status (cf. Grosz 2005, Abraham 2000). However, DPRTs can combine by stacking (14a). Stacking is not coordination (14b) and not considered a phrasal property, but a property of heads. The most common combination is that of two DPRTs like in (14a).   (14) Cx1: Luzia is complaining to grandma about her brother.   Luzia:  Oma, da Elias nimmt ma imma de Lego weg         “Oma, Elias is always taking my Lego!”       51  a. Oma:    <✓, Cx1> Du host  doch eh scho     so vui…      you have doch eh already so many    “You already have doch eh so many, just leave them to him.”    b. <* , Cx1>  Du  host doch und eh scho so vui…     you have doch and eh already so many   c. <*, Cx1>  Du  host doch  scho so vui…         you have doch  already so many       … loss' eam doch nehma.         … leave.them him doch take   “You already have doch eh so many, just leave them to him.”   Thurmair (1989, 1991) observes that sequences of three DPRTs are rare, but can be found; more than two DPRTs are not attested in my corpus however. Also observe that whereas DPRTs can combine, the same DPRT cannot be repeated (14c) to achieve an emphasis effect, for example. Under analyses that consider DPRTs types of expressives (Kratzer 1999, 2004; Potts 2007), repeatability could be expected, since this is a typical property of expressives (Potts 2007). However, the same DPRT such as in (14c) cannot be repeated, such as an intensifier like ‘very’ could. I address a possible reason behind this stacking restriction in Chapter 6.   Some DPRTs can behave as clitics.  Some MB DPRTs have clitic-like properties. For example, ‘a and ‘n, the clitic counterparts of ja and denn, respectively, cliticize to the verb, or, if clitic pronouns are used, to the verb-pronominal clitic complex (cf. Bayer 2012, cf. Grosz 2005). So far, no clear pattern has been established about when a clitic version of a DPRT is used versus when the full particle counterpart is used. Observe (15a), with the full form of the DPRT denn, which follows the pronominal   52 clitic complex. (15b) shows that the clitic version of denn, ‘n, can attach to the pronominal clitic complex as the last element. However, it cannot directly cliticize directly onto the verb, preceding pronominal clitics (15c).   (15) a. Wos  hob’e’da  denn  do?   what have I  you  denn  done?    “What have I denn done to you?”   b. Wos hob’e’da’n do?   c. *Wos hob'n'e'da do?   Since clitics behave inconclusisively with regard to their behavior as syntactic heads or phrases (cf. Zwicky 1985), therefore this particular property is inconclusive with respect to their phrasal status. DPRTs don’t block Verb 2. A possible indication that DPRTs are phrasal is the fact that they do not block head movement. For example, in a V2 clause the finite verb moves from V to C via INFL. If DPRTs were generated as heads between vP and C, they should block verb-movement. That is, according to the Head Movement Constraint head-movement cannot skip any head position (Travis 1984). With this in mind, Bayer and Obenauer (2011) and Bayer (2012) argue that DPRTs are merged as ‘minor deficient heads’ that don’t project, and therefore don’t block verb movement (cf. Struckmeier 2014).  The table below summarizes the findings of the properties discussed in this section, and indicates for each property whether DPRTs behave more like heads (X) or Phrases (XP).   53  properties DPRTs can be coordinated X can be modified X can occupy SpecC X can behave as clitic X/XP blocks head movement XP Table 2: DPRTs as heads or phrases Despite the seemingly overwhelming evidence in favor of DPRTs as heads, many authors (e.g. Cardinaletti 2011; Coniglio 2007; Grosz 2005, 2006) adopt the view that, due to the adverbial nature of many of their counterparts in the other word classes, they are deficient maximal projections in the sense of Cardinaletti and Starke (1999). As such, they are proposed to be merged as adjuncts in the specifiers of a highly articulated IP domain. In sum, it is unclear, and debated in the literature, whether DPRTs are heads or modifiers. The challenge of their status will be taken up in Chapter 6.   2.2.5  Propositional scope DPRTs are interpreted with propositional scope (e.g. Gutzmann 2008, Meibauer 1994, Weydt 1969, Thurmair 1989, Zimmermann 2011). This is surprising because DPRTs appear inside the propositional structure, while scope is assumed to be assigned under c-command (e.g. Hinzen 2006). Hence we might expect that DPRTs would c-command the proposition. This might be related to another well-known property of DPRTs:  they cannot be negated (Thurmair 1989). Crucially, it is the overt syntactic position of negation, and not the semantic scope of negation, which matters for acceptability. The following data show that a DPRT can occur in a negated   54 sentence, where sentential negation occurs low, but scopes semantically over the whole proposition.  (16) a.  Vo dia loss’e  mi  fei  ned  dablecka!       from you  let 1SG  me  fei  NEG  mock       “I won't fei let you mock me.’    (Lena Christ: Memoiren)    b. *Vo dia loss’e  mi ned  fei  dablecka!    from you  let 1SG me  NEG  fei mock  The ordering restriction between DPRTs and negation suggests that DPRTs cannot occur in the semantic scope of negation. They have to take scope higher than negation (cf. Thurmair 1989). The next examples illustrate the same phenomenon with other DPRTs. (17a-b) show that doch also has to precede, and thus scope above, negation. (18a) illustrates that jetz is interpreted as a DPRT when it precedes negation. In contrast, when jetz follows negation, as in (18b) it has to be interpreted as the temporal adverb “now”.   (17) a.  I geh doch  ned   auf’s    Seefest      I     go     doch NEG on.DET  lake.fest      “I'm doch not going to the lake festival.”            b.       * I geh ned doch  auf’s Seefest    I  go   NEG doch  on.DET lake.fest  (18) a.  Fahrts’s’  jetz  ned  in’  Urlaub?        drive.2PL  jetzt  NEG  in.DET  vacation        “Aren’t you jetz going on vacation?”           55  b.<*, fDPRT> Fahrts’s ned  jetz in’ Urlaub?    drive. 2PL NEG  now in. DET vacation      “Aren’t you going on vacation now?”   A similar pattern is observed between DPRTs and other operators, such as modals, quantifiers, question formation, etc. (Gutzmann 2008:33).  They have to outscope all of these sentence level operators semantically. Under the assumption that scope is assigned under c-command the propositional scope of DPRT is a puzzling.   2.2.6  Sentence type restriction Another property of DPRTs is their sentence type restriction. That is, not every DPRT can occur in every sentence type. Consider, for example, the contrast between ja and doch illustrated in (19).  (19) declarative14  a. I geh doch  in'd Uni            I go doch  in.DET  uni                 “I go doch to University.”    b. I  geh  ja in'd Uni              I  go  ja in.DET Uni              “I go ja to University.”                                                    14 I use ‘declarative’ for exposition purposes here. It will become clear that it is neither the formal clause type (i.e. declarative) nor the illocutionary force conventionally associated with this clause type (i.e. assertion), which licenses the DPRT. Aspects of this complex interaction between form and force are discussed in Chapters 3 and 4.    56 (20) imperative   a. Geh doch in'd  Uni!              go  doch in.DET Uni                         “Go doch to University!”    b. *Geh ja  in'd  Uni!15     go ja  in.DET Uni   (21) interrogative    a. *Gehst  doch  in’d  Uni?         go. 2SG  doch  in.DET Uni    b. *Gehst  ja  in’d  Uni         go.2SG  ja in.DET Uni  (22) exclamation16     a. *DU17  bist  doch  gwaggsn!        you  are  doch  grown   b.  DU  bist ja gwaggsn!      you  are ja grown      “How you've ja grown!”       As summarized in Table 3, doch is possible in declaratives and imperatives, but not in interrogatives or exclamations. In contrast, ja, is possible in declaratives, and exclamations, but not in imperatives or interrogatives.                                                 15 Note that accented JA is acceptable in imperatives, however. See Chapter 6 for a discussion on accent on DPRTs.  16 There is a crucial difference between (sentence) exlamations such as the examples here, and exclamatives. Both express the illocutionary force of exclamation, however, their syntax differs. A throrough discussion is beyond the scope of this dissertation, but see Rett (2011) for the differentiation, and the motivation for it. In this dissertation I only consider sentence exclamations that are based on V2 declaratives. Further discussion on these is provided in 3.3.  17 Accent on a word is indicated by CAPS throughout this dissertation. I will address the formal features of exclamations as (21) in 3.3.    57  DPRT  Sentence type      ja doch declarative ✔ ✔ interrogative ✗ ✗ imperative ✗ ✔ exclamation ✔ ✗ Table 3: Distribution of ja and doch across sentence types  The question arises as to what is responsible for this sentence-type restriction of DPRTs. In particular, are we dealing with a restriction on formal clause-type? Or are we dealing with a restriction on illocutionary force? The data in (19)-(22) show that ja and doch are both allowed in declaratives with assertion force, whereas only ja is allowed in exclamations based on the declarative clause type. This suggests that it is the illocutionary force of an utterance that determines the restriction. Observe the following contrast in (23), which shows that this conclusion cannot be correct; the DPRT doch, which was shown to be acceptable in declaratives with assertive force (19a) and imperatives with requesting force (20a), cannot occur in a V1 structure that with requesting force (23b) (indicated by the falling intonation \ and second person inflection on the verb, unlike in a true imperative such as (20a)). This suggests that the distribution of DPRTs is not directly sensitive to illocutionary force, either.  (23) a. Machst       des  auf  \.  make.2SG  DET  open  “(You) open this!”    b. *  Machst des doch  auf  \  make.2SG DET doch open    58  Thurmair (1993) for example argues that DPRTs are dependent on the formal syntactic features of the clause, not the illocutionary force associated with it. She doesn’t, however, provide a motivation for how this dependency comes about.  Similarly, Bayer (2008) also argues for a dependence of DPRTs on formal clause typing features. According to him, a feature lexically associated with the DPRT is responsible for the clause type restriction (Bayer 2008, cf. also Bayer and Obenauer 2011, Bayer 2012, Coniglio and Zegrean 2012). Following this, Bayer and Obenauer (2011) propose that the DPRT denn carries an uninterpretable feature that restricts it to interrogatives (represented as QForce). They propose the feature specification for denn given in (24).   (24) denn [uQForce]  Following in this vein, one could posit similar lexical entries for ja and doch, and endow them with a specific force feature, akin to (24).  However, positing a force feature is inherently problematic, and raises several concerns. First, it presupposes that the illocutionary force of a sentence is syntactically encoded (Rizzi 1997). This is not uncontroversial, and rejected by many researchers on conceptual grounds (cf. Brandt et al. 1992, Reis 1999, Zanuttini and Portner 2003, Portner 2004, a.o.). Second, clause type and illocutionary force do not map onto each other in a one to one fashion, as will be further explored in 3.3. For example, the exclamation in (22) is realized by a declarative clause, which is typically used for assertions as in (19) (cf. Rett 2011). If illocutionary force and the speech act were indeed encoded syntactically, we might expect some syntactic reflexes of this difference. Alternatively, one could say that an exclamation like (22) is not a clause type; but   59 then we have no way to account for the distribution of ja and doch in this example based on syntactic grounds. A third problem for a feature specification account is that many DPRTs (including ja and doch) can occur with a variety of clause types. This means that a given particle would have to be specified for a variety of clause types. This raises the question of how multiple sentence type dependency can be modeled. Presumably we would have to posit a series of unvalued features associated with distinct, but possible related lexical entries. There is another possibility besides the syntactic feature specification hypothesis, namely that the clause-type restriction is semantically conditioned (Grosz 2010a). To evaluate this hypothesis, we have to explore some of the semantico-pragmatic properties of DPRTs. I turn to this in the next subsection.  2.2.7  The semantico-pragmatic properties of DPRTs  It is generally agreed that due to their non-truth-conditional status, DPRTs contribute ‘not at-issue’ content in the sense of Potts (2007). Non truth-conditional status is ascribed to DPRTs in particular due to their perceived optionality for the grammaticality of an utterance. As I will argue, however, DPRTs are not truly optional (see section 2.2.2, and Chapter 6) but do not, however, contribute to the truth-conditions of an utterance, as illustrated below.   (25) a. Da Marinus is fei drei          DET   Marinus  is fei  three             “Marinus is FEI three (years old).”   à true iff Marinus is three      60  b.   Da  Marinus is drei    DET   Marinus  is three    “Marinus is three (years old).”        à true iff Marinus is three   Both utterances are true in a context where Marinus is indeed three years old. The additional meaning fei (like other DPRTs) operates on a different level of interpretation than adding direct lexical content to the propositional content of the host utterance. However, from an analytic point of view, there is little agreement on exactly what kind of meaning DPRTs contribute. Types of meaning that have been suggested to be at the core of DPRTs include presuppositions (e.g. Kaufmann 2004, Kaufmann and Kaufmann 2012), expressive meaning (Kratzer 1999, Zimmermann 2011), and conventional implicatures (Gutzmann 2008).  Although I do not couch my analysis in a specific semantic model, the proposal made here follows closely a proposal in Rett (2013). She argues that there is a distinction between content that is traditionally thought of as at issue (e.g. propositional content), not-at-issue (e.g. implicature or presupposition) and further argues for the need to distinguish content that is neither of these two, but is distinct illocutionary or speech act content (e.g. declarative force). DPRTs, I argue here, modify illocutionary content. This is also how DPRTs have been seen in some of the literature, where they are analyzed as speech act or illocutionary force modifiers (e.g. Jacobs, 1991; Karagjosova 2004; Lindner 1991; Zimmermann, 2004, 2008; Zeevat 2004). A stronger stance takes them as indicators of illocutionary force (Doherty 1987, Jacobs 1986).   In the rest of this section I discuss the strong claim that DPRTs are indicators of illocutionary force, and the weakened version that DPRTs are modifiers of   61 illocutionary content (2.2.7.1). In 2.2.7.2 I discuss the general observation that DPRTs indicate speaker attitude, a claim that will be refined in the course of this dissertation. Section 2.2.7.3 briefly addresses the idea that DPRTs are presupposition triggers.  2.2.7.1  DPRTs and illocutionary force DPRTs have been argued to be indicators of illocutionary force (Doherty 1987; Jacobs 1986, 1991). Consider for example insubordinate clauses, i.e., clauses that have the form of a subordinate clause in that a complementizer introduces them, but at the same time can be used in isolation (i.e., without an embedding matrix clause). However, to be fully felicitous, some insubordinate clauses have to contain a DPRT. For example, the addition of fei renders the formally subordinate clause well-formed in isolation, i.e. it allows it to stand as an insubordinate clause. In particular, it is interpreted as a command (26c).   (26) a. I mecht dass’d    brav  bist  bei  da   Oma.    I want  that.you  good  are  at    DET  grandma   “I want that you’re good at Grandma’s.” b. * Dass’d  brav   bist   bei  da  Oma. that.you good  are  at    DET  grandma   “…that you’re good at Grandma’s.”      c. ✔ Dass’d    fei   brav  bist  bei da  Oma!   that.you  fei   good  are   at    DET  grandma    “Be  fei good at Grandma’s!”   In the example above, (26c) is an insubordinate clause- it can stand by itself, and here does not have the illocutionary force of an assertion, but that of a request/command,   62 and it contains the DPRT fei. Note that the same insubordinate without fei cannot function as an insubordinate clause.18 Similar examples have been taken as evidence for the status of DPRTs as indicators of illocutionary force, allowing a clause that otherwise cannot stand by itself to function like an independent matrix clause (Thurmair 1989). But how does fei contribute to the illocutionary force of the clause? Since fei is also compatible with assertions, we have to conclude that it does not always trigger the same illocutionary force. Hence any approach that proposes that DPRTs supply or ‘match’ semantic or syntactic illocutionary force features of the clause will have difficulty in accounting for data like (26c).  A slightly weakened version of the claim that DRPRTs are indicators of illocutionary force is pursued in Jacobs (1986, 1991) (cf. also Linder 1991), who argues that DPRTs modify the illocutionary force of a sentence. Accordingly, a clause-type is associated with a particular illocutionary type. For example, declaratives are typically associated with the illocutionary force of an assertion. The addition of a DPRT like ja, however, derives a special illocutionary type (i.e., “ja-assertions” Jacobs 1991). According to Jacobs (1986:103) ”Wenn Verbstellung, Verbmodus, Intonation etc. einen bestimmten Illokutionstyp X festlegen, so wird                                                 18 Special intonation on BRAV  can render the example in (26b) felicitous, such as (i)  (i)  Dass’d BRAV  bist   bei  da  Oma!  that.you good are  at    DET  grandma   “Be  fei good at Grandma’s!”  The point here is that without a DPRT or special intonation, the clause cannot function as an independent sentence.    63 daraus durch Hinzunahme eines Abtönungsmittels ein anderer Illokutionstyp X’, der […] in ihren Anwendungsbedingungen eine eingeschränktere Version von X ist.”19  There appear to be several problems with this account. First, it is not clear how an operator for illocutionary type, that is, a ‘force typer’ is assigned to a sentence, and how it interacts with, or can be attributed to clause typing. It resembles the force feature proposed under syntactic proposals, and suffers from the same problems (see 2.2.6). Second, DPRTs can occur in embedded clauses (Coniglio 2009). Hence, embedded clauses would have to be associated with their own illocutionary operator. Whereas this is an assumption made for embedded clauses under verbs of saying (e.g. Krifka 2013), it is unclear whether this applies to all embedded clauses. Last, it is unclear how many illocution types should be considered, and how a DPRT selects its ‘matching’ illocution type.   However, despite these problems, the proposal put forth in this dissertation is a version of, and builds on the illocutionary force modification approach. In particular, I analyze DPRTs as modifiers of the default commitments of the discourse participants. This approach will be developed in more detail in Chapter 5.   Zimmermann (2004, 2008, 2011), also a proponent of a modification approach, argues that DPRTs do not form a homogenous semantic class in this respect. The findings in this dissertation corroborate Zimmermann’s claim. Whereas the analysis presented here differs, I also propose the need for a differentiation of classes of DPRTs from one another. This is argued for in Chapter 4.                                                  19 When verb position, verb mood, intonation etc determine a specific illocution type X, it turns, via the addition of a ‘toner’ [DPRT-added by the author] into another illocution type X’. This type is more restricted in its conditions of use than X is. (Translation ST).    64 Zimmermann specifically shows that wohl in the German variety he discusses is a modifier of clause type (as schematized in Figure 5), whereas ja is a modifier of force (as schematized in Figure 6).20   Figure 5: Wohl as modifier of sentence type   Figure 6: Ja as modifier of force  It is unclear at this point which exact syntactic projections are associated with ‘force’ or ‘clause type’. I take up this question in Chapter 6 where I build on Zimmermann’s insight, and show how a semantic split of DPRTs can follow from a syntactic approach.   2.2.7.2  Expressions of speaker attitude Most approaches to the semantics of DPRTs consider them to be indicators of the speaker’s attitude towards the host utterance (e.g. Doherty 1987, Kwon 2005, Meibauer 1994, Weydt 1969). For example, according to Cardinaletti (2011:493) DPRTs “...are elements that express the speaker’s mental attitude toward his/ her utterance”. Zimmermann (2011:2012) refines this attitudinal function by including the                                                 20 Note that I do not intend to map illocutionary force or sentence type onto a functional projection. It serves exposition purpose only. wohl clause-typeclause-type’ja ForceForce’  65 addressee: DPRTs “are used to organize discourse by expressing the speaker’s epistemic attitude towards the utterance, or by expressing a speaker’s assumptions about the epistemic states of his or her interlocutors concerning a particular proposition”. Kaufmann and Kaufmann (2012) specifically choose the term ‘epistemic particles’ to reflect one of the main functions of DPRTs, namely that of indicators of epistemicity (cf. Doherty 1987). Despite a clear intuition among researchers that DPRTs have a core attitudinal function, it is unclear how exactly DPRTs encode this speaker attitude. Is it directly encoded in the lexical entry, along with the features that restrict it to sentence types? Or does it arise via other means? This, too, is a challenge taken up in this dissertation. In Chapter 6 I will argue that (at least in MB) the attitudinal properties of DPRTs are derived from syntax.  2.2.7.3  DPRTs and presuppositions DPRTs such as ja, are often considered to indicate what is in a mutually shared Common Ground (CG) (e.g. Gutzmann 2012, Kratzer 1999). The CG is assumed to be the (continuously changing) body of public information. It keeps track of what has happened in the conversation (Stalnaker 1978, 2002). In other words, the CG refers to the information that is mutually known to be shared by interlocutors. Under the assumption that presuppositions also refer to mutually shared beliefs (Stalnaker 2002), DPRTs are often considered to be presuppositional (e.g. Grosz 2010b, Kaufmann 2004, Kaufmann and Kaufmann 2012).  Presuppositions are encoded in the semantics of natural language sentences (von Fintel, 2008). If presuppositions are not (yet) satisfied in the CG, addressees can accommodate. Accommodation refers to “the process by which the context is adjusted quietly and without fuss to accept the utterance of a sentence that imposes certain   66 requirements on the context in which it is processed.” (von Fintel, 2008:137). I will argue in this dissertation that MB DPRTs are not presuppositional. On the basis of a case study, I establish in Chapter 4 that some DPRTs can refer to A’s belief on the one hand  (fei, doch), and are compatible in contexts where S and A share knowledge about p on the other hand. In that sense, DPRTs like ja and doch can have presuppositional effects, as will be discussed in Chapter 5, but these derive indirectly from the compositional effects of the discourse context, the host utterance, and the core semantic contribution of the UoL that derives the DPRT function.   2.3  Conclusion In this Chapter I introduced some of the core properties that have been observed for German DPRTs. I showed that these properties raise several questions, which to date have received unsatisfactory or incomplete answers. The open questions concern the following.   (i) DPRTs are multifunctional in two distinct ways. On the one hand, they have counterparts in other word classes such as adverbs, conjunctions, or response particles. On the other hand, they also are multifunctional as DPRTs: they display a functional range. Whereas the first aspect will not be directly addressed in this dissertation, I will provide an answer for the second facet of DPRT-multi-functionality in this dissertation.  (ii) DPRTs are described as optional, since they contribute neither truth-conditional content nor grammatical content. However, I showed that DPRTs are not optional when considered within a larger discourse context as well as in some cases of insubordination. This raises the question of whether propositional structure is the only   67 point of evaluation for the ‘grammaticality’ of an utterance, and in what way felicity and grammaticality judgments can be seen as parallel. I show in this dissertation that we can extend the domain of evaluation for well-formedness of an utterance within discourse.  (iii) The phrasal status (head of modifier) of DPRTs is unclear. It is also unclear whether DPRTs are to be considered a word class. I argue here that DPRTs are a category in the sense of Wiltschko (2014). Their categorial status is not lexically encoded, however, but derived from syntax. Syntax also derives their mixed behavior with respect to their phrasal status; in Chapter 6 I argue that DPRTs do not merge in a head position, but that they associate with a head (Wiltschko 2014). (iv) DPRTs have high propositional scope, yet are syntactically integrated into the utterance. The question arises as to how DPRTs gain propositional scope. Chapter 6 argues that DPRTs associate with the syntactic spine initially at IP, the clausal level where a situation is fully established with all its arguments and temporal relations.  (v) The sentence type restriction of DPRTs has been treated as a consequence of syntactic/semantic features, or as a correlate of the illocutionary force of an utterance. In Chapter 4 I propose that this restriction is the result of neither of these; the restriction falls out from the compatibility between the orientation (expressed as commitment) of a clause type, and the orientation of a DPRT.  (vi) Regarding the semantic contribution of DPRTs, it has been argued that they indicate or modify the illocutionary force of an utterance, that they supply presuppositions, and express the speaker’s attitude. I will take elements of all these proposals, but will show that some of these observations about their semantics are merely effects of DPRTs in a particular context, but not their main function.    68 Each of these questions summarized here will be addressed in more detail in the course of this dissertation.    69 Chapter 3: Ingredients for an analysis of DPRTs 	3.1  Introduction One of the basic components of human cognition is the recognition and understanding of others’ mental states. Theory of mind, which is the ability to attribute mental states, that is, knowledge, beliefs, and intentions, to oneself and others, is a prerequisite for the ability to have conversations. If a speaker were not able to assess an addressee’s epistemic state, she could not decide, for example, whether or not a proposition is new to the addressee (Hayano 2013). This means that knowledge about the discourse context influences what kind of utterances are used, what form those utterances have, and how the discourse participants are affected. I argue in Chapter 4 that DPRTs are sensitive to discourse participant epistemicity. This will lead me to the proposal that some DPRTs are sensitive to the (perceived) epistemic stance of the addressee; some are sensitive to the epistemic stance of the speaker; and some are sensitive to the epistemic stance of some other discourse participant.  As introduced in Chapter 1, I pursue here the idea that the DPRT function fDPRT is just that; the function f of a UoL in a specific context Cx, as indicated in (1).  (1)  fDPRT = UoL + Cx  Chapter 4 will deal in detail with the distinct discourse particle functions fDPRT of five MB UoLs. This current Chapter is dedicated to providing the necessary background for the tests applied in Chapter 4, which rely heavily on context. I also introduce the core contribution of the UoLs that derive their DPRT function.    70 This Chapter is structured as follows. I begin with the introduction of the UoLs that are at the heart of this dissertation, and show examples of their discourse particle function fDPRT for illustration. Based on diachronic evidence, I propose a core DPRT meaning for the fDPRT of each UoL discussed. With this fDPRT, these UoLs operate over participant epistemicity, in that they indicate something about a discourse participant’s epistemic stance toward the content of the utterance p. This brings us into the realm of Common Ground, beliefs, commitments and context, all of which will be introduced in more detail in the second part of this Chapter. In 3.3 I introduce how I perceive of Common Ground (CG) in this dissertation, and how CG relates to individual beliefs, shared beliefs, commitments and to the context. In 3.4, I discuss the role of clause typing in establishing a discourse participant’s commitment to the content of an utterance. Since DPRTs are sensitive to the participants’ epistemic stance, I argue that they predictably interact with the public commitments established by clause typing (cf. Jacobs 1991), shown in 3.5. Section 3.6 summarizes.   3.2  The test cases The main claim of this dissertation is that discourse particle function is decomposable into a UoL with a core content and a specific context. This is repeated again below.  (2)   fDPRT = UoL + Cx  At the heart of this idea are UoLs, basic sound and meaning bundles, and contexts. This means that a UoL and its core meaning itself contributes toward fDPRT. This contribution has been recognized in many approaches to DPRTs, such as this:   71 “particles […] have a semantic content which they deploy in connection with other elements of the utterance.”(Weydt 2006:207, cf. Abraham 2001). Diewald (2006) also recognizes that each DPRT (used throughout this dissertation as the shorthand for fDPRT) has a diachronically motivated UoL with a basic semantic meaning at its core. This core meaning is generally interpreted within the context of the propositional structure (p-structure), resulting in the ‘other’, lexically accessible interpretations of these multifunctional UoLs. However, if interpreted within the context of the discourse participants’ epistemic stance (see Chapters 4 and 6), the lexical content of the UoL is not directly transparent. Yet, it crucially contributes to deriving fDPRT. In this dissertation I discuss jetz (3.2.1), eh (3.2.2), ja (3.2.3), doch (3.2.4), and fei (3.2.5). For each UoL I address the following aspects in this Chapter:   (i) core DPRT function and epistemicity (ii) diachronic origin (iii) example of fDPRT (iv) previous literature (v) non-DPRT functions (multi-functionality)  I follow Hentschel (1986) and Zeevat and Karagjosova (2007) in considering diachronic data for a synchronic analysis, assuming that diachronic core meaning components are synchronically present as specific meaning components that are shifted in DPRT use. This shift in meaning is referred to as deixis shift in in Hentschel (1986) leading to her concept of metacommunicative deixis.    72 3.2.1  Jetz The UoL jetz derives from Middle High German je and zuo, meaning ‘always’ and ‘to, towards’. These two independent words were conjoined, rendering the Upper German ietzo (DWDS 2016). Echoes of this use are visible in MB until today, where jetz often is phonetically realized as jetza, or etza. Synchronically, jetz is most often used as a temporal adverb, with the interpretation ‘at the present time, now’. Based on this diachronic origin and synchronic use, I propose that the core function of the DPRT jetz in MB is as in (3) below (to be refined in Chapter 5).  (3)    jetz ≈ p is salient in the context now  In particular, I will argue that jetz expresses salience of p for a contextually determined discourse participant O at the moment of utterance, argued for in Chapter 4. Examples of its fDPRT are shown in the following examples:  (4)   a.  Warst      jetz   gestan  bei da  Gerda?       were.you jetz  yesterday  at    DET  Gerda        “Now then, were you at Gerda’s yesterday?”    ‘[It is relevant now] whether you were at Gerda’s yesterday.’  b.  Mia  fahrn  jetz  moang east.       We  drive  jetz  tomorrow  first    “We’re now going tomorrow.”   ‘[It is relevant now that] we’re going tomorrow.’  (4a) shows jetz in a polar interrogative, (4b) in a declarative. In each of the examples, jetz co-occurs with a temporal adverb ‘yesterday’ and ‘tomorrow’, gestan, moang, indicating that jetz cannot be interpreted as temporal adverb in these examples, as the   73 result would be two conflicting temporal references. 21  Under a reading with both as adverbs, they would have to be coordinated, e.g. as jetz und moang ‘now and tomorrow’ in order to result in a grammatical utterance. I take this ability to co-occur with other temporal adverbs as a diagnostic that jetz is not interpreted with respect to the p-structure (where temporal relations are established), but with respect to the discourse structure (to be refined in Chapter 6) (cf. Hentschel 1986).  jetz is not mentioned in the list of ‘core’ DPRTs presented e.g. in Thurmair (1989). Hentschel (1986:35) however argues that jetz is a DPRT in Alemannic varieties of German (cf. Krier 1991). She considers the function of jetz to be equivalent to denn, a DPRT used in interrogatives (1986:35). Hentschel does not discuss cases like (4b), however, which show that jetz can also be used in declaratives, where, like in interrogatives (4a), it loses its adverbial function (cf. Rehbock 2009). I adopt Hentschel’s assessment that jetz is used as a DPRT in southern German varieties, and propose that in the uses shown in (4) it functions as a DPRT.   Besides its use as a DPRT, and temporal adverb meaning ‘now’, jetz also has a use as discourse marker, characterized by its clause initial position. In example (5) below, jetz occurs in a request before the vocative Heini, clearly indicating that it does not occupy SpecC, and therefore is not a moved temporal adverb. In the example in (6), the second occurrence of jetz is as temporal adverb.                                                  21 Munaro and Poletto (2004) show that the Venetian dialect of Italian has a DPRT ‘mo’, which is derived from (and syncronically used as) a temporal adverb meaning ‘now’ in other dialects of Italian. mo, however explicitly cannot co-ccur with other temporal adverbs, since, according to the authors  “[i]n Venetian imperatives mo is sensitive to the time of the utterance, as it signals that the utterance time and the event time must coincide.” This is different from MB DPRTs, which clearly can co-occur with temporal adverbs.  (i)  Ciamime  (*tra un’ora), mo! Call me  (in an hour),  mo  (ii) Lezilo (*doman),  mo! Read it (tomorrow),  mo    74 (5)   Cx: Heiei and Heini, two boys are making a plan together. Heini is    distracted and monkeys around. Heiei admonishes him.   Heihei:    Jetz  Heini,  reiß  di zsamm!   jetz  Heini   pull  you  together    “Now Heini, pull yourself together!”        (Perlmutterfarbe)  (6)   Cx: Heiei and Heini, two boys, are making a plan together. They are    pulling sticks to see who is going to be the person getting strapped into   a contraption their friend made in order to undergo an ‘experiment’.   Heini, who always tends to lose, complains:  Heini: Jetz  konn’e  ma   scho  wieda  denga   wos   jetz  bassiert   jetz  can.I     me    already  again   think    what  jetz  happens    “I can already imagine what is going to happen right now”   In the right syntactic context, namely when it is derived by the adjectival suffix-ig, jetz can also function as an adjective, as the next example illustrates.  (7) Da    jetz-ige Voastand  is  fia  2  Jahr  gwäid     DET  jetz.adj.SG     board   is for  2  years  elected “The current board is elected for 2 years.”   The table below summarizes the various uses of jetz depending on the contexts.  UoL function jetz discourse particle  temporal adverb discourse marker adjective Table 4: Multi-functional uses of jetz   75 3.2.2  Eh The UoL eh derives from the Middle High German temporal marker ē, meaning früher, vormals ‘earlier, before’ (DWDS 2016). Following Hentschel (1986), and based on its diachronic temporal origin, I propose that the function of DPRT eh is that of marking the content p of an utterance as true before the time of utterance (see 4.5).  (8) eh ≈ p was true in the context before the time of utterance  An example for the DPRT use of eh follows.  (9)   Cx: The cabinet-maker offers farmer Schiermoser some furniture.    Schiermoser declines.  Schiermoser: A     Einrichtung  sogst?...    DET  furniture     say.2SG     “Furniture you say?”   ...I  brauch koa       Einrichtung. Mei Haus  is eh eingricht….   …I  need  NEG.DET  furniture         my  house is eh furnished.    …I brauch gar       nix.  …I need  really nothing.  “I don't need furniture. My house is eh furnished.  I don't need anything at all.”  (Lena Christ, Memoiren)  The cabinet-maker didn’t know that Schiermoser doesn’t need or want furniture, otherwise he wouldn’t have offered. However, Schiermoser responds that his house is furnished; eh in the answer indicates that this state of affairs was true before he uttered this. Its use thereby also indicates to the cabinet-maker that he may have   76 known this already.  The DPRT eh has received little attention in the literature (Hentschel 1986, Schlieben-Lange 1979, Thurmair 1989). To date no generative analysis has been provided. Hentschel (1986:53) considers the contribution of eh to be “that what is said was valid and true before, that is, before it became the topic of conversation.” (translation ST). She, as I do here, relates this contribution of eh to the diachronically related temporal adverb counterpart ehe ‘before’. Thurmair provides the most detailed description for DPRT eh, which is often translated as anyways or already. She assigns eh the function of marking p as previously known to S (but not to A) (Thurmair 1989: 136). In the account offered in Chapter 4, eh does not relate p to S, but rather to a contextually defined participant O.  Echoes of the temporal origin of eh can be seen in some Bavarian dialects, which use eh as a temporal conjunction (although no instances for MB were found in my corpus). Merkle (2005:193) provides the following example:  (10) Eh       dass’e  do      naufsteig … before  that.I    there  up.climb     …vazicht’e       liawa  auf  de scheene Aussicht    …go.without.I  rather  on  DET  nice   view     “I’d rather give up on the nice view than climb up there.”  Besides the (rare) use as temporal conjunction, eh can also function as a response particle, as in the example below.22                                                   22 The adverbial use of eh in contemporary German is restricted to a fixed phrase seit eh und je (‘since forever’, ‘since ages’), which attests to the diachronic origin of the word as a temporal adverb (Hentschel 1986).   77 (11) Martin: Kummst  moang  aa  zum  Alex   come.2SG tomorrow  also  to.DET Alex  “Are you coming to Alex’s tomorrow, too?” Hans:   Ja, eh.    “(Yes), eh.”   ‘(Yes), [this was true before I answered this].’   In sum the UoL eh has a variety of uses, shown in the table below.   UoL function eh discourse particle response particle  temporal conjunction Table 5: Multi-functional uses of eh  3.2.3  Ja The UoL ja has as diachronic origin in the proto-Indo-European pronominal stem *i- (a 3rd person neuter demonstrative ‘that’), which in turn derives from a locative (DWDS 2016, Hentschel 1989). From this origin, the UoL derives its synchronic uses as response particle, as well as discourse particle and discourse marker. I propose that as DPRT, ja functions to mark that the speaker S believes the propositional content of the utterance p.    (12) ja ≈ I believe p  The following example illustrates this function (13).     78 (13) Heihei:    Jetz  Heini,  reiß  di zsamm!  jetz  Heini   pull  you  together     “Now Heini, pull yourself together!” Heini:   Is ja scho   guad!    is ja  PRT-already good      “Ok, ok, I got it.”    ‘[I believe that] it’s already good.’  According to Hentschel (1986), SG ja is the most common DPRT in the German language. Besides doch, it is one of the best-described and researched DPRTs. Despite this wealth of literature on SG ja, several problems remain. First, it is unclear which German variety/varieties the claims in the various analyses and descriptions are made for. This is relevant, since the inventory of DPRTs in a given language can affect the individual particle’s range of functions, as I illustrated in Chapter 1. Second, the analyses diverge in the scope of the data covered; most accounts, for example, regard ja and its accented counterpart JA as two separate DPRTs (e.g. Coniglio 2009, Kratzer 1999, Thurmair 1989). Observe the contrast between the two below.   (14) a. Du gehst ja in’d Schui  you go.2SG ja in.DET  school  “You’re ja going to School.”  ‘[I believe that] you’re going to school.’  b. Du gehst JA  in’d Schui  you go.2SG ja+accent in.DET  school   “You’re JA going to School.”  ‘You had better be going to school.’   In (14a), unaccented ja renders a reading that will be discussed in more detail in   79 Chapter 5, I will call it the ‘reason’ reading. In  (14b), accented JA derives a request, with a warning undertone. I will not discuss JA and its function in detail in this dissertation, but I discuss a general approach to accented DPRTs in 6.6.3.   Thurmair (1989) for example bases her assumption about two separate forms, JA and ja, on observations about the relative position of JA and ja with other DPRTs such as auch; ja has to precede auch, whereas JA has to follow auch.   (15) a. Er hat ja  auch immer seine Aufgaben gemacht   he  has ja  auch always his    chores     done   à ja>>auch  b.  Mach auch JA  immer deine  Aufgaben! do     auch  ja+accent    always your   chores          à auch>>JA   (Thurmair1989:103)  I do not assume here that accent is indicative of two separate UoLs.23  It is not a general property of German or its dialects to distinguish lexical items based on pitch accent (Féry 2012). Hence it would be surprising if that was the case for ja or other accented DPRTs like DOCH.  Instead, I follow Meibauer (1994), Gutzmann (2010), and Egg and Zimmermann (2011) in assuming that the accented DPRTs are derived compositionally from its unaccented counterpart. The difference in function follows from the contribution of                                                 23 In the syntactic analysis I present on Chapter 6 I will end up arguing that JA associates with a lower syntactic projection than ja. CP for the fomer, GroundS for the latter.     80 focus accent (Féry 2012).24   Another problem concerns the contribution of ja to an utterance. There are two main approaches in the literature as to how the contribution of ja can be best captured. I refer to them as the (i) shared knowledge approach (e.g. Kratzer 1999, 2004, Thurmair 1989; Zimmermann 2011), and (ii) the factivity approach (Lindner 1991, Ormelius-Sandblom 1997, Bárány 2009, Kratzer and Matthewson 2009, cf. Grosz 2010b). I argue that the factivity approach is on the right track, at least for MB ja.  According to the shared knowledge approach, ja refers to mutually shared knowledge between S and A, that is, to CG (Kratzer 1999, 2004; Thurmair 1989; Zimmermann 2011). The appeal to shared CG knowledge is often reflected in the choice of English translations for ja, such as ‘as you know’. The shared knowledge approach is ostensibly supported by the example below from an unidentified German dialect. No context was provided for the example, which makes it difficult to verify (or falsify) the proposed contribution. In this example in particular, according to Zimmermann (2011:2013) ja "indicates that the speaker takes the hearer to be aware of the fact that Max is at sea." Without context, however, it is impossible to determine what A knows, and what S can reasonably assume that A knows.   (16) Max ist ja auf See     Max is ja  at   sea     (Zimmermann, 2011: ex 1)  In an approximation for the contribution of ja, Kratzer (1999), another proponent of the shared knowledge approach, considers ja appropriate  “…in a context c if the                                                 24 I will return to accented DPRTs in Chapter 6.6.3.    81 proposition expressed by p in c is a fact of wc which - for all the speaker knows - might already be known to the addressee.” (Kratzer 1999: def. 3).  The second approach, the factivity approach to ja characterizes the role of ja to point to the factivity of p in a given situation. In particular, ja is assumed to mark the proposition p expressed in the host utterance as a fact and hence true (Bárány 2009, Lindner 1991, Kratzer and Matthewson 2009, Ormelius-Sandblom 1997). No appeal to shared knowledge between interlocutors (i.e., CG) is made.  I adopt the factivity approach for MB ja as well. In particular, I present evidence in Chapter 4 that ja does not explicitly appeal to shared knowledge. This follows Kratzer and Matthewson (2009), who argue that ja is not presuppositional, i.e. does not refer to A’s epistemic stance (pace, for example Kaufmann, 2004). They consider the contribution of ja to be as follows.  (17) “If p is the descriptive content of a sentence U in a context c, then the use of [...] ja in c indicates that the speaker of c takes p to be an established fact, and therefore doesn't consider the question whether or not p to be an issue for either the current or any future inquiry". (Kratzer and Matthewson, 2009:6) My proposed contribution for ja in (12) follows Kratzer and Matthewson’s proposal in (17). It will be defended in detail in Chapter 4.   Besides the use as DPRT, ja has other functions as well, most prominently as a positive response particle, shown in (18), and as discourse marker, shown in (19).  (18) Cx:  Sebastian’s class is at a radio station, where they learn about radio    transmission of sound. The students get to try a radio booth and    microphones.     82  Bernie:  I her di, herst du mi aa?     “I hear you, can you hear me, too?”  Sebastian:  Ja,  freilich!  ja  certainly   “Yes, of course!”   (Wer früher stirbt…)  (19) Cx: A family dinner. Everybody is getting ready to sit down. Oma is    getting her grandkids to sit, whereas other adults are trying to find    open spots on the bench.     Oma: Jetza- de Luzia sitzst si moi do her. Elias, wo bist denn du?      “Now- Luzia is gonna sit down over here. Elias, where are you gonna be?”   Peter: Ja und SCHO is da Blots bsetzt…     ja and already is DET place occupied.    …Ja dann sitz’e me  do  hea   …ja  then sit.I  me  there here  “Oh well, and the spot is taken just like that. Well then I’ll sit down here.”   Hentschel (1986) in particular argues for the idea that all uses of ja, in particular the DPRT use can be derived from its diachronic origin as a deictic root. The synchronic multifunctional uses are summarized below.   UoL function ja discourse particle response particle discourse marker Table 6: Multi-functional uses of ja   83 3.2.4  Doch According to Karagjosova and Zeevat (2007), doch originates as a marker of questions with which the speaker seeks confirmation of A’s opinion. Hentschel (1986) argues that doch derives from the proto-Indo-European demonstrative *te-/to- and the addition of two emphatic clitics, functioning like an emphatic “this!” (Hentschel 1986:43).25 Doch therefore has a similar diachronic origin as ja, but contains more emphasis.  Taking these diachronic origins of doch into consideration, I suggest the contribution of the DPRT doch to be: this is something that I believe you believe, or shorthand I believe you believe p.   (20) doch ≈ I believe you believe p  This proposed contribution in (20) is corroborated by Kwon (2005), who assumes that when S uses doch, she assumes that A does not consider p at the moment, although p is already known to A (Zimmermann 2011). An example for this use of doch in MB is shown in (21).  (21) Cx1: Franz hears music on the radio, which is played on the intercom in the  supermarket where he is shopping with his dad. His brother Sebastian  is on a field trip at a radio station. When the station identifier is played,  Franz says to his dad:                                                      25 Hentschel (1986:43-44) considers the combination of “all this+emphasis+emphasis” nothing else but a “doubly reinforced, anaphoric ‘this'” (Translation ST)    84  < ✓, Cx1 > Do  is  doch  heid  da  Sebastian.    there  is  doch  today DET  Sebastian   “SeBASTIAN is THERE today!  ‘[I believe you believe that] Sebastian is there today.’    (Wer früher stirbt…)  DPRT doch, alongside ja is among the most widely researched DPRTs of German (e.g. Abraham 1991, Bárány 2009, Egg and Zimmermann 2011, Grosz, 2010b, Hentschel 1986, Lindner 1991, Schmerse et al. 2013, Thurmair 1989; Zeevat and Karagjosova 2007, a.o.). According to Grosz (2010b:1) the function of DPRT doch in SG is to mark its host utterance as “familiar, old, given, uncontroversial or shared”. In addition, he assumes that it also conveys, “some notion of contrast or correction” (cf. Abraham 1991, Bárány 2009, Kaufmann and Kaufmann 2012, Lindner 1991, Thurmair 1989, Zeevat and Karagjosova 2007). Whereas I do not adopt this approach, contrast is widely proposed to be one of the core meaning components of doch (e.g. Bárány 2009, Egg and Zimmermann 2011, Grosz 2010b, Ormelius-Sandblom 1997), and is considered to be encoded in its lexical entry. As an advocate of this hypothesis, Grosz (2010b) suggests that it includes two meaning components, as shown in (22). The first component establishes p as uncontroversial, the second component establishes that p contrasts with another proposition q. Grosz bases the first part of the contribution of doch on Kratzer and Matthewson's (2009) semantics of German ‘ja’, and in effect claims that doch is ja+contrast.       85 (22) Grosz’ proposal for the meaning of doch a.  uncontroversiality:  doch p presupposes that p is an established fact and ¬p can be safely discarded.  b.  contrastiveness:  doch p presupposes that there is a contextually salient proposition q, such that: i.  q is a focus alternative of p ii.  given contextually salient background assumptions, q contradicts p (i.e. if p and q is not a logical contradiction, doch p presupposes that in the current context ¬ [p and q].)  (Grosz 2010b:2)  Grosz (2010b:3) presents the following example to support the contrastive meaning component; according to his analysis, doch is the element introducing the correction.    (23) Mary: Schau mal! Diese Blumen sind so hässlich.  look PRT these  flowers are so ugly   “Have a look! These flowers are so ugly.”    Bill: Was hast du    denn? Diese Blumen  sind doch  schön!  what have  you  denn these  flowers  are  doch  beautiful   “What's your problem? These flowers are [DOCH] beautiful!”  ⇒ [p The flowers are beautiful] is used to correct [q the flowers are ugly].           (Grosz 2010b: ex (3))  Following Karagjosova and Zeevat (2007), who also claim this for other German varieties, I do not consider doch inherently contrastive (cf. Krifka 2013). Rather, I show in Chapter 5 that the contrastive interpretation is composed from the context   86 doch occurs in, and from inferences that arise due to the need to establish discourse coherence. The main reason for this is that doch can also occur in contexts that do not provide any contrastive interpretation. This will be discussed in more detail in 5.6.  A common translation for doch is ‘but’, reflecting this idea of contrast (24a). However, contrast arises (and the same translation can be chosen) in the absence of the DPRT, as is illustrated in (24b). Translation therefore can be misleading, since it is often aimed at verbalizing the elements of the propositional structure (not the discourse structure).  (24) Cx: Hanni is at a family gathering at her aunt Betty’s. Betty usually makes   cake, and also tends to prepare an evening meal, Brotzeit. After coffee   Hanni says goodbye and gets ready to leave. Betty says:   a.  Gehst du    scho?  Es gibt doch no    a     Brotzeit!            go.2SG you   already  it  exist doch still  DET bread.time             “Are you leaving already? But there is still cold cuts and bread”     b. Gehst du    scho?    Es gibt  no    a     Brotzeit!            go.2SG you   already  it  exist  still  DET bread.time             “Are you leaving already? But there is still cold cuts and bread”   Regardless of whether doch encodes contrast or not, all analyses agree that it expresses some aspect of S’s belief about A’s belief. That is, doch relates p to A’s epistemic state.   Besides its use as DPRT, MB doch also functions as response particle, illustrated below.26                                                 26 SG has also a use for doch as conjunction; this function is not attested in MB, however.     87 (25) A: Fahrts es heia  ned  mid zum törggelen?27      drive.2Pl you.pl this.year NEG  with  to.DET wine-tour         “Aren’t you coming along this year to go wine-touring?”  B:  Doch.         doch          “Yes, we are.”  Krifka (2013) analyzes the response particle doch as an IP (propositional) anaphor. He proposes that doch presupposes a contextually relevant proposition p, as well as a another contextually salient proposition p′, such that p′ is a focus alternative to p, and ¬[p and p′], the same way Grosz (2010b) analyzed the DPRT doch. This lends itself to deriving DPRT and response particle use from one another. Both uses of doch are summarized below.   UoL function doch discourse particle response particle Table 7: Multi-functional uses of doch  3.2.5  Fei Very common in southern German varieties, the MB DPRT fei is popularly seen as a marker of Bavarian identity (Merkle 2005).28 Fei is diachronically derived from the adverb fein, meaning ‘fine, exact, to the highest degree’ (DWDS 2016). In its adverbial use, the final nasal ‘n’ is kept, whereas the DPRT fei drops the nasal at the                                                 27 Törggelen is a term used in South Tyrolia for wine-harvest time trek from mountain side wine-cellar to wine-cellar. Groups of people usually hike together and enjoy wine and local delicacies.  28 Fei, in 2004 and 2010 surveys conducted by the Bavarian public broadcaster “Bayrischer Rundfunk” was voted to be the most popular Bavarian word.  (http://www.sueddeutsche.de/panorama/mein-liebstes-bayerisches-wort-des-is-fei-a-dantschigs-herzipopperl-1.665773)   88 end. Echoes of the diachronically present final nasal can be found in older speakers, who nasalize the diphthongized vowel to render feĩ. The particle fei is frequent in MB. It is one of the only two DPRTs that is not multifunctional, and only used as DPRT, the other one being ‘halt’.29 I propose its contribution as DPRT to be the following.  (26) fei ≈ I believe you don’t believe p  In an attempt to formalize the contribution of the DPRT, I analyze fei as a “polarity discourse particle” in Thoma (2009). In particular, I propose a meaning for fei in terms of polarity focus (cf. Höhle 1992): fei emphasizes the opposite of p (note that this is the generally proposed function of doch). Fei anchors the utterance to the discourse context by accessing A’s beliefs toward p. Like doch, fei expresses the speaker’s assessment about the epistemic state of A. An example of the DPRT fei is shown in (27).  (27) Cx:  Martin is at an Ox race, where Sir Quickly races his Ox Ringo. Ringo   only runs to the tune ‘Mr. Tambourine Man’, but the tape is missing.   This makes Sir Quickly extremely nervous, and he is pacing up and   down. Martin gets impatient with Sir, and says to him:   Martin: Lang  schaug’e  ma  des fei  nimma         oo    long   look      I  me  DET  fei  NEG.always  on     “I am running out of patience!”    ‘[I believe you don’t believe it, but] I am running out of patience!’                 (Irgendwie and Sowieso)  Schlieben-Lange (1979) considers the contribution of the DPRT fei as marking the                                                 29 halt is rendered as hoid in MB. This particle will not be discussed in detail here.     89 propositional content of the host utterance (her ‘argument’) as new. Fei has also been called an emphatic marker in Merkle (2005); its use, according to Merkle,  “strengthens” the utterance. This notion of strengthening or emphasis is often evoked in the description of DPRTs in general. I will turn to an explanation for how emphasis is derived from the interplay between the core meaning of fei, and the various contexts in which it occurs in Chapter 5.  3.3  Belief, commitment and epistemicity The repository of knowledge shared between discourse participants in a conversation is known as the Common Ground (CG; cf. Stalnaker 1978, 2002). Establishing CG requires the discourse participants to build and maintain a complex model of the conversation situation. Any given speaker has to consider her own belief, her interlocutor’s belief, and the beliefs that S assumes to be shared with A. All these beliefs can be accessed by DPRTs. Hence, we need a model of CG which can be relativized to a conversational situation, and which allows one to target any of the discourse participant’s individual beliefs. In order to model the contribution of DPRTs, we need a discourse model that separates the discourse participants’ public beliefs, that is, their commitments (cf. Farkas and Bruce 2010). The concept of grounding (introduced in 3.3.1.) allows us to model just that. I show that an individual’s belief about a state of affairs expressed in an utterance can be established via the context within which this utterance occurs (3.3.2). I propose for all the UoLs introduced above that they function to express the epistemic stance of a discourse participant regarding the content of the host utterance. This stance is dependent on the discourse context, as well as the individual commitments and content (propositional   90 or otherwise) expressed via clause-typing. Whose stance is expressed exactly, and with which UoL, will be argued for in detail in Chapter 4. In order to frame the discussion to follow, I now explain how the notions of belief, epistemicity and commitment relate to each other. In order to frame the discussion to follow, I now explain how the notions of belief, epistemicity and commitment relate to each other.  In a given (discourse) situation, S and A both have their individual sets of private beliefs, that is, the beliefs that are not public. By uttering a sentence, a speaker S conveys her private belief, expressing, by uttering the sentence, a public commitment regarding its content. For example with an assertion with propositional content p S expresses a public commitment to p (Gunlogson 2001). S wants her private belief to become CG, and pending on the addressee’s acceptance or refusal, the content of the assertion will become CG. CG then is the intersecting set of these private beliefs that, if made public and mutually accepted, are the set of mutual beliefs. Following Gunlogson (2001) I will call these mutual beliefs shared public commitments. This is how commitments can serve as proxy for beliefs (Rett 2013). A commitment then is public, whereas a belief is private. However, often beliefs are referred to as public beliefs, but when they are, they are in effect commitments, at least in the sense I intend it here. An utterance, before its content p is grounded, can be understood as straddling the line between the speaker’s private beliefs and the Common Ground (of shared public commitments). Therefore speaker intention plays a role in reaching certain conversational states of affairs; for example, by uttering an assertion, S wants to effect a specific conversational state of affairs (namely that p become CG).   As I discuss in 3.4, clause-type also serves to indicate a speaker’s public commitment (Beyssade and Marandin 2006). For example, the form of a declarative communicates an assertion that expresses S (public) commitment to the propositional   91 content p (Gunlogson 2001). p does not only refer to the content of assertions. I intend for it here to serve as shorthand for the content of any utterance, in the form of any clause-type. Therefore, in this dissertation p also stands as shorthand for the content of imperatives, interrogatives and exclamations. In short, I take p as any (propositional or otherwise) content expressed in a clause type, and accordingly p co-varies with the syntactic form. Whereas this is a simplification of the facts, it allows me to model the contribution and effects of DPRTs.     3.3.1  Grounding: separating speaker from addressee DPRTs are used to indicate the epistemic stance, often in form of a belief, of a discourse participant toward the host utterance. It is therefore crucial for an analysis to be able to track who believes what in a conversation. The discourse component tasked with storing shared mutual beliefs, that is, public commitments, is commonly assumed to be the CG (Stalnaker 1978, 2002). The CG minimally contains propositions (Stalnaker 1978), a partially ordered set of questions under discussion (QUD) (Roberts, 2012) and the latest discourse move (Ginzburg, 1996). I also assume that it contains the individual discourse participants discourse commitments with regards to a to-do-list (Portner 2004). Thus the CG is a model of how discourse participants share what is talked about (propositions, to-do-lists) and, as well, that something has been said (latest discourse move, QUD). Its contents are summarized in figure 7 below.       92                                        Figure 7: CG Content  Crucially, this content only becomes CG after it has been grounded. Grounding, a term introduced by Clark (1985) can be seen as “…the fundamental, moment-by-moment conversational process by which speaker and addressee are constantly establishing mutual understanding” (Bavelas et al. 2012:5, Clark and Schaefer 1989).30  CG building and management is viewed as a dynamic process. CG content is not built and updated automatically. Just by virtue of uttering an utterance with a given content (that co-varies with clause-type, and henceforth is called p for exposition), p and latest move do not automatically become CG; they need to be grounded.31 The process of grounding proceeds in two phases (28) (cf. Clark and Schaefer 1989, Clark and Brennan 1991, cf. Bavelas et al 2012).                                                   30 Michael Rochemont (pc) points out that this also makes deaccenting a grounding device, as well as non-focus accenting.  31 I am only concerned with the grounding of the content of an utterance, not with grounding the fact that something has been uttered- the two are separate processes, as evidenced by the fact that a response particles can react to both (examples due to work on response particles by Erin Guntly p.c. and Martina Wiltschko p.c.) (i)  Elise: Are you coming to the party?  Erin:  Yeah, no.  Here Erin grounds the utterance by accepting that Elise asked a question. She grounds the content p of the utterance by refuting it with ‘no’.   Propositions To-do-lists QUD   93 (28) Phases in grounding:   Presentation Phase: S presents utterance with content p to A  Acceptance Phase: A accepts p  This two-step process is necessary in order to integrate the content of an utterance into the CG (Farkas and Bruce 2010, Irmer 2009, Malamud and Stephenson 2015). This is in many ways an idealization of the state of affairs, since much depends on the form of the utterance, the choice of lexical items, etc. In other words, the contexts that will be presented next play a crucial role in whether p becomes CG or not.  Since utterances do not update the CG by virtue of being uttered, they are then proposals to enhance the CG of a conversation (cf. Farkas and Bruce 2010; Malamud and Stephenson 2015). S, by uttering p, places p on the “conversational table” T. A then can pick up p from T, and accept it, or refute it. Grounding can occur with linguistic as well as non-linguistic cues, such as gestures, facial expression, nods, etc. p can be grounded for example by the absence of a negative response; speakers may also assume that A accepts p, by virtue of a lack of overt linguistic disagreement (cf. Gunlogson 2008).  This two-step model crucially presupposes the separation of the beliefs of the discourse participants: S’s beliefs about p are separate from A’s beliefs about p. In addition, S can have a belief about A’s belief, but does not have to share A’s belief. That is, there are propositions to which a certain discourse participant is not publicly committed, but which are nevertheless in their ‘belief space’ (i.e. for example S knows that A knows that S believes p, but A doesn’t believe p, etc.). In what follows, I indicate S’s belief about p as in (29a) and S’s belief about A’s belief as (29b).    94 (29) a. Bel (S,p)   b. Bel (S, Bel (A,p))   Importantly, the discourse participants can hold beliefs about their interlocutor’s epistemic states. Belief ascription is a necessary precursor to our ability to have conversations; S would not normally ask A a (sincere) question, if S didn’t believe A had the answer. S normally would not make an informative statement via an assertion if she thought A already knew this information. This normal state of affairs is expressed in the constitutive rules associated with the speech act of an utterance (Searle 1969). DPRTs serve to modify aspects of this normal course (cf. Zeevat 2004). They are special kinds of speech act modifiers that do not change the speech act itself, but amend the public commitments (and thereby, epistemicity) indicated by the speech act (cf. Egg 2010, cf. Jacobs 1991, cf. Zeevat 2003). Discourse participants gain support for what (and how strongly) they believe, and for what they believe their interlocutor believes from linguistic, discourse, and situational evidence provided to them (Fischer 2006). The beliefs of S and A can therefore differ, since propositions, to-do-lists, and QUD can be established by means other than linguistic context alone. This is discussed in the next section.  3.3.2  Context: the relation between an individual and p In this section I discuss in more detail the notion of context  (Cx). The core proposal of this dissertation is that the function of a given DPRT is compositionally derived from a UoL and the context it is used in. I introduced the UoLs in section 3.2, whereas now I turn to the role of context with the background on grounding in mind.      95 (30)   fDPRT = UoL + Cx  In particular, we shall see that context serves as an indicator for which discourse participant believes what about p, or about their interlocutor’s epistemic stance. Hence, context restricts the use of DPRTs, as we will see in more detail in Chapter 4.   I take context to be multi-faceted. That is, context is not one, uniform discourse entity, but instead is best viewed as a construct of several sub-contexts, all of which can establish CG content (Clark 1985, 1996; cf. Irmer 2009). In particular, to understand the function of a DPRT (fDPRT) we have to take into consideration several sub-contexts, as listed in (31).   (31) Ways of establishing CG content via context  (i) syntactic form: the host utterance   (ii) discourse context: linguistic knowledge via discourse antecedent   (iii) situational context: specific facts regarding the utterance situation   (iv) world knowledge: nonspecific knowledge of facts and rules in the world   These factors are not separate, but rather are in a subset relationship to each other.   Figure 8: Discourse contexts   96 In order to utter a sentence, S needs to decide on the appropriate form of a sentence, i.e. its clause type and associated intonation contour, which results in a particular function (i.e. illocutionary force). The specific form depends on the discourse context; not every form is appropriate in every context. The immediate discourse context, (i.e., preceding utterances), also delimits the types of forms that are possible. 32 Furthermore, the situational context, which contains the participants’ actions, non-actions, behaviors, looks, gestures, etc., can serve as indicator of the kinds of utterances that are appropriate and felicitous in a discourse context (Fischer 2006).  Lastly, CG content established by world knowledge is the kind of information we assume our interlocutor shares with us, based on general facts about the world that are taken for granted by members of the speech-community at large. For example, as a resident of Canada, I can reasonably assume that people I talk to in Vancouver in late October 2015 are aware that a new Canadian Prime Minister was elected recently. Crucially, discourse participants do not need overt linguistic antecedents in order to gain belief, or share CG about a certain situation or state of affairs. Bartels (1999) frames this in terms of salient propositions, which need not be expressed overtly.33 All contexts can provide evidence, or allow for inferencing, that a certain state of affairs is known to or believed by a given interlocutor (Fischer 2006, Irmer 2009). This is illustrated in the following examples.34                                                  32 The discourse context may also contain information about the social status of participants, their relationship to each other, etc.  33 This can also be framed as QUD (Roberts 2012). Nothing hinges on the choice of salient propositions over QUD for the purpose of the analysis presented here.   34 Whereas these contexts guide inferences, and can proved evidence for certain beliefs about the interlocutors’ epistemicity, contexts do not determine the belief of an interlocutor.     97 (32) Belief inferred from discourse context (via linguistic antecedent)   Cx1: Two friends, Martl and Alex, visit with each other and chitchat. Martl    tells Alex he doesn't have time to stick around for dinner since he’s    going to the movies.   • Martl believes that Alex knows that he doesn’t have time for dinner, since Martl just told Alex so.   (33)  Belief inferred from situational context  Cx2: Two friends, Martl and Alex, visit with each other and chitchat. Alex,    who lives alone, sets the dinner table for 2. Martl assumes the second    plate is for him.  • Martl believes that Alex thinks he is having dinner with him, since he sets the table for two.   (34)  Belief inferred from world knowledge  Cx3: I have a conversation with a colleague. I make reference to “our new     Prime Minister” without mentioning his name.  • I believe that my interlocutor knows I speak about Justin Trudeau since we both live in Canada. I can reasonably assume that anybody living in Canada would have followed the very extensive election coverage, and knows that Trudeau was elected PM in the fall of 2015.  DPRTs, as shown in 2.6, are subject to clause-type restrictions. I argue, that clause-type restriction instantiates a special kind of contextual restriction, namely a restriction to form. (i.e., Cx: form). In particular, clause-type, I argue, provides CG content regarding the interlocutor’s belief states in terms of public commitments. That is, an utterance does not only carry information regarding its propositional content, its syntactic form also indicates who is committed to, and is supposed to be committed to p. This will be discussed next.    98 3.4  Clause types, illocutionary force, and speech acts I take a speech act to be the output of a form and force pair, which can be modified by a variety of means, including DPRTs (cf. Zeevat 2003). What I refer to as form is what is traditionally clause type, whereas force is illocutionary force, the function associated with the form. I make this distinction, since there is sometimes a terminological conflation of the notions of clause type (here form) and the associated illocutionary force (here considered a function). Speech act, under the approach taken here, corresponds to the notion of sentence mood, which is the final output of the combination of form type and function type (Altmann 1993). The insight that form type (= clause type) and function type (=illocutionary type) are two distinct concepts and need to be kept separate is particularly relevant for the question of clause type restriction of DPRTs (cf 2.6).   I assume here that the force of a clause type is compositionally derived from the combination of distinct distributional and morphological features (cf. Lohnstein 2000, cf. Zanuttini and Portner 2003). This contrasts with the growing body of literature on clause typing, according to which sentential force is syntactically encoded via a force feature in the left periphery of the clause (e.g. Cheng 1997, Han 2000, Rizzi 1997, 1999, and most cartographic accounts). I do not adopt a formal feature that determines the illocutionary force of a clause type, due to form-force mismatches shown later in this dissertation.  It is uncontroversial that clause type is conventionally associated with a function, that is, with a specific illocutionary force (Meibauer et al. 2013, Sadock and Zwicky 1985) as summarized in (35).     99 (35)  The conventional association of form and force types      Form   Force  Commitment        declarative   asserting   S     interrogative   asking   A     imperative    requesting   A      exclamative   exclaiming   S  Allan (2006) refers to this conventional association with illocutionary force as the primary illocution of an utterance. Each illocutionary force is concomitant with a discourse participant’s commitment, and therefore, a belief (cf. Rett 2013). Assertives commit S to p, and ask A to commit to p as well. Questions commit S to an open issue that A is supposed to resolve. Directives commit S to an outcome, and ask A to add it to their to-do-list. Exclamatives are exclusively concerned about S’s commitment, and do not involve A (Beyssade and Marandin 2006; Malamud and Stephenson 2015). Each of these commitments will be motivated further when I turn to individual clause types. Clause typing then, as I claim here, can serve to establish ‘who believes what’ in a conversation. I crucially draw from the proposal in Truckenbrodt (2006), who develops a formal theory of the effect of V to C movement in German. Whereas I do not adopt the feature mechanisms proposed by Truckenbrodt (2006), I take his main insight that verb movement to C activates a ‘context index’ situated in the CP layer, representing S’s belief (his Epist).35                                                 35 More formally, Truckenbrodt proposes the following: (i) In a context index <DeontS (x) (<Epist>)> in C   a. Epist is present iff    (i) C contains a finite verb with indicative or subjunctive II, or    (ii) C/CP is marked [+WH];   b. x = A iff C contains a finite verb with person inflection.  In Truckenbrodt’s terms, DeontS refers to the speaker of the utterance. Epist is the shared knowledge between S and A, expressed in Truckenbrodt as ‘it is common ground that/whether..[p].’. Indicative or subjunctive verbal morphology, as well as the person features on the verb provide the necessary   100 Clause types, in sum, are conventionally associated with an illocutionary force, which in turn is tied to a specific discourse participant’s commitment (Beyssade and Marandin, 2006). This commitment can be modified and changed by intonation, however (among other modifying tools, such as DPRTs, as will be shown here). Therefore, intonation is an essential factor, for establishing and also for modifying belief. I assume that every form has a ‘default’ intonation contour, as we simply cannot produce an utterance without any intonation at all. I also assume that there is a second type of modifying intonation that is not usually associated with a specific form, and can serve to modify the primary illocution (cf. Rett 2013). However, the picture is a lot more complicated than what I am able present here; nevertheless, I argue here that clause type and intonation are considered to form a unit, which must be interpreted together, establishing a form. This form (clause type + intonation) also constitutes a special kind of context. It encodes the public commitments of a discourse participant, and positions its content p with respect to the discourse participants. Clause typing this way establishes a commitment for a discourse participant (cf. Gunlogson 2004, 2008). Since form gives an indication of the (public) commitments of the discourse participants, it indirectly provides an indication of who believes what. I crucially adopt Rett’s (2013:20) assumption that public commitments can serve as a proxy for belief.36  Searle (1969), following Austin (1975), refers to the primary illocutions, expressed as the result of the commitments, as speech acts, each with their constitutive rules.                                                                                                                                       features that need to be valued in C. Therefore only verb movemement from V to C can value these features in C. Thus x (referring to A) and Epist in the context index of C can be seen as unvalued features triggering V-to-C. (Truckenbrodt 2006: 262).  36 I refer the reader to Rett (2013, 2014) for further details on how commitment and belief can be considered the same for the purposes I intend here.     101 These will be introduced in Chapter 5, where I discuss how DPRTs mark a divergence from the default commitments expressed in each speech act. He specifies four types of constitutive rules for utterances:  (36) Constitutive rule for speech acts a. content rules   b. preparatory rules c. sincerity rules (addressing what is needed for a speech act to be sincere) d. essential rules (specifying what the speech act counts as).37   The content rules for a speech act corresponds to what I refer to as Cx: form. This, in Austin’s terms, refers to the locutionary act: it is the actual utterance and its meaning, including the phonetic realization, syntactic form, and semantics of the lexical items involved in the utterance.  The sincerity rules concern the situation in which a given speech act is uttered, i.e. they refer to the contextual preconditions for a given speech act. These contexts will (have to) be presented for each utterance containing a DPRT, in order to track the epistemicity and intentions of the interlocutors. The preparatory rules for each speech act will be presented in detail when I investigate the interpretational effects of DPRTs in Chapter 5.    In the next two sections I show in more detail that a form is comprised of two parts: clause type and intonation. Each will be discussed.                                                    37  In the following I ignore the ‘essential rules’, since they do not directly bear on the discussion here. An example for an essential rule for a request for example would be that the utterance counts as an attempt of S to have A perform the requested action.   102 3.4.1  Syntactic form as context The form of an utterance, its clause type and its intonation establishes the commitment of the discourse participant towards p. I refer to this as commitment assignment (cf. Rett 2013)  (37) Commitment assignment: sentence form (including clause-type + intonation)  encodes the discourse participants’ belief toward p, showing their public  commitments.  The notion of commitment assignment is based on the well-established assumption that the features of formal clause typing show S’s relationship, and S’s preference for A’s commitment to p (Byessade and Marandin 2006; cf. Brandner 2010; cf. Condoravdi and Lauer 2011; cf. Zaefferer 2001; Truckenbrodt 2006). This preference is expressed in the constitutive rules associated with each speech act. A declarative for example publicly commits S to its propositional content p. In addition, S wants A to commit to p as well (Bach and Harnish 1979, Zeevat 2003, Searle 1969). When S has a question, A, by default, is believed to have the answer, and so on.  The commitment assignment created by the default form-force pairing is not invariable, however; it can be modified to indicate a change in S and/or A commitment. Clause typing establishes a default force, and modifiers such as peripheral particles (Beyssade and Marandin 2006, Malamud and Stephenson 2015 Heim et al. 2016, Wiltschko and Heim 2016) and intonation (Gunlogson 2004; Truckenbrodt 2006, 2012, 2013; Trinh and Crnic 2011), may modify the intended speech act, more explicitly, the default epistemicity expressed in the speech act. This means that a clause type and the associated illocutionary force do not always map onto each other in a one to one fashion. I show now that intonation is an integral part   103 of Cx: form, and a non-negligible factor when considering discourse participant commitment.  3.4.2  Intonation Intonation is usually not directly taken into consideration when discussing clause types (see Altmann 1984, 1993 for an exception). The role of intonation, however, is crucial in (re-)assigning commitments (Trinh and Crnic 2011; Heim et al. 2016). For that reason Searle (1969) classifies intonation as an Illocutionary Force Indicating Device. I adopt here the idea developed elsewhere that intonation contours are intonation morphemes in their own right (Truckenbrodt 2012, 2013; Pierrehumbert and Hirschberg 1990, Trinh and Crnic 2011; Davis 2011 for Japanese; cf. Altmann 1984, 1993 for German). Therefore, they play an essential role in the syntactic composition of clause types. Simplifying a rather complex picture significantly, I will consider three basic contours here and describe their contribution to specific clause types: sentence final falling intonation (indicated by \), sentence final rising intonation (indicated by /), and sentence initial or medial extra heavy pitch accent  (exclamation intonation; indicated by √).38 I summarize each of their functions in turn.   Studies such as Gunlogson (2004) establish that rising intonation may serve to turn a declarative clause-type into a question. Such sentences are known as ‘rising declaratives’. With /, S raises the issue whether p; it involves A in that it shifts the commitment to p from S to A (Gunlogson 2004, 2008; Truckenbrodt 2006, for German 2012, 2013; cf. Davis 2011 for Japanese). That is, the assertive illocutionary                                                 38 This is a simplification of the picture. However, I base my assumptions on findings from the works cited in this section, which provide a clear indication about the crucial role of intonation. However, more work on its contribution, and on the exact pitch contours needs to be undertaken.     104 force typically associated with a declarative clause-type is modified, and yields question force. This shows clearly that it is not clause type alone that establishes illocutionary force; rather, clause type and the associated intonation have to be taken into consideration. In example (38), for instance, a V1 structure with rising intonation / renders a polar interrogative with question force. A declarative with the same intonation / renders a (rising) declarative with question force. The same clause type with falling intonation \ renders an assertion.   (38) Clause type+ intonation = form  a.  Is that a persimmon /?   Polar interrogative = Question  b.  That’s a persimmon /?   Rising declarative = Question   c.  That’s a persimmon\.   Declarative with falling intonation = assertion          (adapted from Gunlogson 2008, ex:3)  Next consider the special intonation that derives exclamations (√). It is typically characterized by extra high amplitude or length on any constituent in the clause, or an intonation peak on the main accent of the utterance (Bolinger 1986, Batliner 1988, Bartels 1999). It is associated with high speaker emotion (Bolinger 1986, Truckenbrodt 2013), and surprise. The surprise can either be about the fact that p, or due to the extreme degree of an element of p (Rett 2011, Truckenbrodt 2013). Regardless of the reason for the surprise, an exclamation does not involve A in any way. Rather it involves only S’s attitude towards p in that it commits S to p (Beyssade and Marandin 2006, Rett 2011).   Consider the example in (39). Here the change in intonation leads to a change in illocutionary force. The same clause type (in the form of a declarative) can receive   105 three different functions, depending on intonation. It can function as an exclamation (39a), as an assertion (39b), or as a question (39c).   (39) Adam can cook steak. Example Clause type Intonation Force Adam can cook steak! declarative √ exclamation Adam can cook steak. declarative \ assertion Adam can cook steak? declarative / question  (modified from Rett, 2011: ex 27a)  According to Truckenbrodt (2013), the falling intonation morpheme \ expresses that S wants to assert p as true.39 This contribution is conventionally associated with assertions.40 I assume here that \ facilitates to establish S commitment to p, as in (38c) and (39b).   In what follows, I discuss the forms relevant for the analysis of DPRTs.41  3.4.3  Declarative A MB declarative matrix clause is characterized by the following properties (cf. Altmann 1993, Brand et al 1992, Reis 1999, cf. Meibauer et al 2013); (i) the finite                                                 39 Mit [\] drückt der Sprecher aus, dass er zu p eine assertive Einstellung hat, i.e .p als wahr darstellen will. ("Ich sage, dass p")”. With \ S expresses that he has an assertive attitude toward p, that is, wants to present p as true. (“I say that p”). Truckenbrodt (2013) (Translation ST).  40 It is difficult to establish whether it is the contribution of \ alone, or whether the syntactic properties of clause types also encode this  41  This list is not exhaustive, but covers the clause types and associated illocutionary forces investigated here.    106 verb is moved to C via V2 .42 (ii) verb mood is indicative or subjunctive (iii) a non-wh phrase occupies SpecC. This is illustrated in (40). In (40a), a subject DP (‘da Marinus’) occupies SpecC, whereas in (40b), an adverbial phrase occupies SpecC. The default intonation associated with declaratives is \.   (40) a. [SpecC Da Marinus [C  gibt   [da  Luzie  heid  a  Bussl]]]    DET  Marinus      give    DET  Luzia   today DET kiss      “Marinus is giving Luzia a kiss today.”     b. [SpecC  Heid [C  gibt   [da Marinus  da Luzie a Bussl]]]  \    today     gives  DET Marinus  DET Luzia DET  kiss  “Today Marinus is giving Luzia a kiss.”     à Bel (S,p)   A declarative with \ conventionally is interpreted as an assertion. The constitutive rules for uttering an assertion include that S is committed to p, and believes p, i.e. Bel (S, p).   If / associates with a declarative, the illocutionary force of the utterance is no longer assertive, but questioning (Gunlogson 2008). / shifts the commitment to A, i.e. shifts Bel (p) to A.43                                                  42 Verb movement is assumed to establish illocutionary force (e.g. Wechsler 1991; Truckenbrodt 2006; Bayer 2010). This is a type of syntactic force feature approach, since it assumes a feature that triggers the verb to move to C to establish force, where matrix clauses have their own illocutionary force, and can stand by themselves as utterances, whereas subordinates lack illocutionary force, and thus cannot stand independently (cf. Haegeman 2002:159).    107  (41) ?? Da Marinus gibt da  Luzie  heid a Bussl  /      DET Marinus  gives   DET  Luzia  today  DET kiss      “Marinus is giving Luzia a kiss today?”  A declarative associated with √ is a sentence exclamation in Rett’s (2011) terms. It is associated with the illocutionary force of exclaiming. Only S belief is involved, and there is no requirement on A’s part to respond, or add p to their beliefs. Rett (2011) notes that since sentence exclamations are always uttered with an intonation distinct from matrix assertions, they should not be analyzed as assertions. The next example provides an illustration; whereas the clause type is that of a declarative, the illocutionary force is that of an exclamation.   (42) Da MaRINUS gibt da Luzie heid a Bussl √   DET Marinus      gives  DET  Luzia   today  DET  kiss     “(Wow), Marinus is giving Luzia quite a kiss today!”  The findings of this section are summarized below.  INTONATION function commitment \ Assertion S / Question A √ Exclamation S Table 8: Intonation associated with declaratives and resulting functions  Next I turn to the interrogative clause type, and the associated illocutionary forces.                                                                                                                                        43 Note that it is my impression that rising declaratives are very rare in MB. They are certainly not used with the same frequency as in English. I believe that the same is true for other dialects of German, but this would have to be confirmed by a larger data sample.     108 3.4.4  Interrogative I consider two types of interrogatives, (i) wh-interrogatives, and (ii) polar interrogatives. Each type of interrogative is conventionally associated with the illocutionary force of questioning. However, formally, and in terms of commitments, both types differ, and both types can be modified by intonation. I discuss each in turn.   3.4.4.1  Wh-interrogative As illustrated in (43), MB wh-interrogatives are characterized by the following properties: (i) the finite verb is in C and (ii) verb mood is either indicative or subjunctive. (iii) SpecCP has to be occupied by a wh-word. Wh-interrogatives typically occur with a rising intonation /. They are conventionally associated with the illocutionary force of questioning.  (43) [SpecC Wea [C  gibt    [da  Luzie a  Bussl ]]]  /         who      gives   DET   Luzia  DET  kiss    “Who is giving Luzia a Kiss?”  Unlike what we observed with declaratives, however, a change in the intonation contour does not correlate with a change in illocutionary force. That is, even if an interrogative clause associates with falling intonation \, the illocutionary force is still that of questioning. I assume with Brandner (2010 for SG) that wh-interrogatives are always interpreted as questions because of their syntax: spec head agreement between the wh-word in SpecC and the verb in C triggers questioning force.  Turning to √, Brandner (2011:88) reports for German that wh-interrogatives, “can never be re-interpreted [as exclamation: ST] – even with the ‘‘best’’ […] intonation and the ‘‘best’’ plausible context”. In MB Bavarian, however, wh-interrogatives can   109 easily be modified with √, and be interpreted as exclamations, as the next example shows.    (44) Cx:  Teenager Lena is dressed up, ready to go out. Her mom looks at her    with a disapproving look, and says:     a. Wia schaugst  denn DU aus √        how  look       denn you out         “(Boy) how you look!”  b. Wia  SCHAUGST  denn  du aus √        how  look             denn  you out         “(Boy) how you look!”  The table below provides a summary.  INTONATION function commitment \ question A / question A √ exclamation S Table 9: Intonation associated with wh-interrogatives and resulting functions  3.4.4.2 Polar interrogative As illustrated in (45), polar interrogatives are characterized by the following properties: (i) the verb occurs in C, (ii) it has indicative or subjunctive morphology, (iii) nothing occupies SpecC. Conventionally, a polar interrogative associates with rising intonation /, resulting in question force.   (45) [SpecC --[C  gibt [ da Marinus da Luzie a  Bussl ]]] /         give  DET Marinus DET Luzia DET kiss           “Is Marinus giving Luzia a kiss?”    110 If an utterance with the verb in initial position is realized with falling intonation \, the interpretation is that of a comment, which is characterized by the absence of S commitment to p (Reis 1999, Önnerfors 1997). The data below show a V1 structure with falling (46a) and rising (46b) intonation.   (46) Cx1: Karl tells a joke to his soccer buddies. He begins:  a. <✓, Cx1> [SpecC --[C  Kummt [a  Estarreicha in’  Himme ]]] \...        comes    DET Austrian in.DET heaven        “So an Austrian goes to heaven …”  b. <*, Cx1>  [SpecC --[C  Kummt [ a  Estarreicha in’  Himme ]]] /...   I follow Reis (1999), who, building on Önnerfors (1997), argues that the filling of SpecC with an XP results in S commitment to p (cf. Truckenbrodt 2006). She observes that so-called V1 sentences are all comment and do not assert. V1 plus \ presents propositional content, but does not assert its truth. These types of V1 are often used to set the scene when opening a narrative, or to begin a joke as in (46a), which supports this view of them being ‘presentative’ (Wiltschko to appear) V1 clauses associated with \ are therefore not assertions and lack S commitment to p.   Finally, V1 sentences can also associate with √. The resulting exclamation is characterized by high speaker emotion and surprise (Bolinger 1986, Truckenbrodt 2013). The surprise, as mentioned above, can either be about the fact that p, as illustrated by (47a), or due to the extreme degree of an element of p, shown in (47b)  (Rett 2011, Truckenbrodt 2013).     111 (47) a. Cx1:  3 year old Marinus is visiting from Canada. He hasn’t seen his great-   Aunt Christa in a long time. He is shy, but she overhears him saying   something in Bavarian to his grandma. Christa exclaims:    Redt  da BUA  BOArisch √ !   speak  DET    boy    Bavarian    “(It’s surprising that) the boy speaks Bavarian!”     b. Cx2: 3 year old Marinus is visiting from Canada. He hasn’t seen his great-    Aunt Christa in a long time. He is shy, but she overhears him saying    something in Bavarian to his grandma. Christa chuckles about his     Canadian influenced Bavarian:      Redt  da BUA  a  Boarisch  √ !44    speaks  DET  boy DET  Bavarian     “The (kind of) Bavarian the boy speaks!”  Like with all exclamations, the use of a V1 does not involve A, or express A commitment to p. Only S is committed to p. If the verb in a V1 utterance is inflected for a 2nd person and occurs with \, the associated illocutionary force is that of a directive. I assume here that it is specifically the 2nd person features on the verb that facilitates this interpretation (cf. Zanuttini 2008, Isac 2012).   (48) Machst      as  Fensta    zua  \  make.2SG  DET  window  closed   “You close the window!”  The clause type typically associated with directive force is the imperative, which will be discussed next. Before discussing imperatives, I show a summary of the findings in                                                 44 The value for the extreme degree reading is due to the determiner, which implies an elided adverb phrase specifying the degree (Brandner 2010). I will not further consider the differences in these two exclamative readings, since the associated commitments and thus illocutionary force are the same.     112 this section.   INTONATION function commitment \ \ + 2nd person inflection presenting (“V1 Declarative”) requesting -- A / question (polar interrogative) A √ exclamation  S Table 10: Intonation associated with V1 clauses and resulting functions  3.4.5 Imperative As illustrated in (49), the imperative clause type is characterized by the following properties: (i) V in C. (ii) verb in imperative mood. Whereas imperatives in German are often considered subjectless (Wratil 2013), subjects are sometimes possible, as discussed in detail by Zanuttini (2008). SpecC remains unoccupied.  (49) [SpecC --[C gib   [am Marinus a   Bussl]]]           give   DET Marinus DET kiss     “Give Marinus a Kiss!”  By uttering an imperative, S places a requirement on A, and gives instructions to add the property expressed in the utterance to A's discourse commitments (to-do list in Portner 2004). With the use of an imperative S asks A to add p to her discourse requirements. Thus the conventionally associated illocutionary force is that of a directive. The associated intonation contour is a final fall (Altmann 1993).45  Neither / nor √ can associate with imperatives.  If the A commitment expressed in an imperative could be either tied to the verb                                                 45 It is my impression that this final fall differs from the final fall associated with e.g. declarative assertions \. Further research is necessary to establish what the precise contours are.  For exposition, I will continue calling this contour ‘final fall’ and represent it with \.    113 mood or the fact that there is nothing in SpecCP, I consider the possibility that it is due to the imperative mood of the verb. This is due to the fact that imperative force obtains even in those rare cases where an XP precedes the imperative verb. The data below suggest that even though SpecC is occupied with XPs in the marked forms below, no S commitment is expressed with those imperatives (Wratil 2013).   (50) a. Jetz     gib     am  Marinus  a    Bussl \   now  give    DET   Marinus  DET kiss    “Give Marinus a kiss now!”    b. A    Bussl   gib     am  Marinus \    DET  kiss give   DET   Marinus      “Give Marinus a kiss!”  I summarize the findings of this section in the table below:  INTONATION function commitment \ request A / n/a n/a √ n/a n/a Table 11: Intonation associated with imperatives and resulting functions  3.5 Conclusion In this Chapter I introduced the necessary components needed to derive the function of DPRTs. In particular, I introduced the UoLs that serve as a base to express DPRT function. I also introduced the types of contexts I will use in the next Chapter to test the epistemic states of the discourse participants.   Discourse, situational and world contexts can set up the ‘who believes what’ by means of preceding utterances, situations, gestures, and the general environment (e.g.   114 if two people are standing in the rain talking, they both are aware of, and therefore can be reasonably assumed to believe that it is raining without overtly saying so). Form is also such a context, albeit a more restricted one. In particular I discussed that the form of an utterance, clause-type and intonation, indicates which discourse participant is committed to, and has a belief about p. The combination of both can therefore determine the illocutionary force of an utterance, and which in turn is restricted due to the discourse context. For example, since a polar interrogative with /, expressing a question, conveys S’s wish to receive an answer from A, this polar interrogative can only be uttered in the appropriate context, one in which the discourse context (e.g. via preceding utterances as in (51)), provides an indication to S that A can reasonably know the answer to the question. This is illustrated below.   (51) a. Cx 1: Andreas tells Heidi about the new house he just bought. She asks him  where it is located. Andreas responds that it is in the interior of BC,   and continues:  <*, Cx1> Are winters snowy there?   b.  Cx2:  Andreas tells Heidi about the new house he just bought in the interior   of BC. She congratulates him and tells him she grew up there. Andreas   responds that he would love to have tips about life out there and    continues:   <✔, Cx2> Are winters snowy there?  Discourse and situational contexts can therefore restrict the possible forms available to discourse participants in their next discourse move; a form and its associated illocutionary force indicates a discourse participant's commitments (and by proxy, beliefs).    115  Below I present a table of the forms shown in this Chapter. The illocutionary force of the clause types under consideration here changes depending on the associated intonation contour (cf. Altmann 1993). Imperatives appear to resist association with intonation contours other than the default, as do wh-interrogatives. The latter, however, allow for modification with √, resulting in an exclamation.   form function Clause type intonation  illocution commitment V2 Declarative  \ assertion S, A / question A √ exclamation S Interrogative -wh  / question A \ question A √ exclamation S Interrogative polar  V1 V1 / question A \ presentation - √ exclamation S Imperative  \ request A / n/a  √ n/a  Table 12: Summary of clause types, intonation, functions, and commitments  In the next Chapter I show how each of the UoLs I introduced above expresses a DPRT function fDPRT in specific contexts. These fDPRT can be organized into three separate groups; those which express the belief of a contextually determined reference point O, those which express the belief of the speaker S, and those which express the belief of the addressee A.   116 Chapter 4: Speaker-orientation, addressee-orientation and other-orientation  4.1 Introduction In this Chapter I show that DPRTs are sensitive to the epistemic states of discourse participants: some are sensitive to the epistemic state of the addressee (A); some are sensitive to the epistemic state of the speaker (S); and some are sensitive to the epistemic state of some other discourse participant (O). I refer to this particular characteristic of DPRTs as their orientation. I show that DPRTs can be classified according to their orientation, that is, to which discourse participant’s epistemic state they are sensitive. As a result, I argue that the MB DPRTs I discuss here fall into three main classes as summarized in (1).  (1)  a. A-oriented DPRT: relates content of an utterance to A   b. S-oriented DPRT: relates content of an utterance to S  c. O-oriented DPRT: relates content of an utterance to a contextually   determined discourse participant O To establish the orientation of a particular DPRT I make use of specific test frames as summarized below.  A-orientation: A-oriented DPRTs can be identified in contexts in which S displays a belief about A’s attitude towards p. This is the case in three distinct contexts as summarized in (2). S believes that A believes p (2a); S believes that A believes ¬p (2b) or S believes that A has no belief about p (2c).   117 (2)  Three contexts to diagnose A-orientation   a. Bel (S, (Bel (A, p)))   b.   Bel (S, (Bel (A, ¬p)))  c. Bel (S, (¬Bel (A, p)))  S-orientation: S-oriented DPRTs can be identified in contexts which require S to display an explicit attitude towards p. This is the case when S believes p (3a) or when S believes ¬p (3b).46   (3)  Two contexts to diagnose S-orientation  a. Bel (S, p))  b.   Bel (S, ¬p))  O-orientation: O-oriented DPRTs can be identified in contexts where neither S nor A, but a contextually defined epistemic reference point (EPR) ‘other’ expresses a belief about p. This is the case in contexts where S doesn’t have grounds to believe that p (4a).47 If S is neutral toward, and has no commitment to p, S does not express her attitude and is not taking any stance with respect to p. Specifically, this property results in a shifting behavior of O-oriented DPRTs; they can occur in forms that have a variable EPR in terms of the commitments of the discourse participants. This has as result that O-orientation is testable in shifting contexts, i.e. across several forms;                                                 46 S can also say something about p without displaying an attitude or an orientation- this is the case, for example in presentatives, i.e. V1 clauses with falling intonation. Also specific markers such as evidentials (e.g. reportedly, allegedly in English) present a context in which S presents some information and reports another’s belief about p, without herself having an attitude about p.   47 Recall that if S thinks that A doesn’t believe p, S still has a belief about A. This is not the case when S doesn’t have a belief; then S truly is not committed to p or an attitude about p.   118 interrogatives, declaratives and imperatives, each of which commit different discourse participants to p (4b). The term shifting is usually used to refer to indexical elements that shift, for example from referring to the addressee of the utterance to referring to the addressee of the matrix predicate. I also chose this term, following Döring (2013), highlighting the similarity between DPRTs with shifting indexicals.  (4) Test contexts to diagnose O-orientation  a. ¬Bel (S,p)  b. Shifting environments: Bel (O, p)   This is summarized in Table 13 below.  A belief Orientation    Bel (A, p) A    Bel (A, ¬p) A    ¬Bel (A, p)/no belief about p A S belief     Bel (S, p) S    Bel (S, ¬p) S O belief   ¬Bel (S, p) O	    Bel (O, p)	 O	Table 13: Orientations arising from epistemic states  I begin the discussion in 4.2 with A-orientation, and show that the DPRTs fei and doch function as A-oriented DPRTs. Section 4.3 shows that ja functions as an S-oriented DPRT. In section 4.4 I take stock and discuss the predictions regarding the distribution of S- and A-oriented DPRTs, across forms as well as across contexts that indicate epistemicity via non-grammatical means. In section 4.5 I show that jetz and   119 eh function as O-oriented DPRTs. I summarize the findings and conclude this Chapter in 4.6.  4.2 A-orientation In what follows, I discuss two MB A-oriented DPRTs: fei and doch. I show that they both are sensitive to what S takes to be the epistemic state of A.   A belief is established here via situational contexts (Cx: sit) and discourse contexts (Cx: disc), although general knowledge about the world, (i.e. Cx: world) can serve to indicate the belief state of A as well. In other words, in certain contexts, what can reasonably be known about the world, can give S an indication about A’s set of beliefs.   As I show now, A-orientation obtains whenever S assesses A’s attitude towards p. This is the case if S thinks that (i) A believes p, (ii) A believes ¬p, and (iii) when A does not have a belief about p. This results in three test frames for establishing A-orientation for fei and doch.   (i) Cx: sit and Cx: disc provide S with evidence that A believes p, (ii) Cx: sit and Cx: disc provide S with evidence that A believes ¬p and finally, (iii) Cx: sit and Cx: disc do not provide S with evidence that A has a belief about p (this is trivially true in the absence of contextual information as well as in contexts where A just asked S about p). These test frames are summarized below. Note that, soliloquy, (i.e., self- talk), is a special instance of (i) and (ii), with the additional requirement that S is the same person as A (cf. Hasegawa 2011).     120 (5)  Test frames for A-orientation  (i) A believes that p: Bel (A, p) due to Cx: disc or Cx: sit   (ii)   A believes that ¬p: Bel (A, ¬p) due to Cx: disc or Cx: sit (iii)  A has no belief about p: Cx: disc or Cx: sit do not indicate whether A believes p or ¬p a. out of the blue (OOB) as a special context of ¬Bel (A, p) b. S answers A’s question about p  (iv) Self talk (Soliloquy): A believes p or ¬p, AND A=S  I show now that fei and doch (introduced above in sections 3.2.4 and 3.2.5) are both A-oriented DPRTs. The former is used in contexts where A believes that ¬p, whereas the latter is used in contexts where A believes p. As proposed in 3.2.4 and 3.2.5, the contribution of fei and doch can be summarized as in (6) (see Chapter 5 for some refinements):   (6)  a. fei  ≈ I believe you don’t believe p    b. doch  ≈ I believe you believe p   The following two examples provide support that these are indeed the contributions of fei and doch, respectively. The context S uses to infer A’s belief is established by world knowledge and situational context in the following way: Sir is in a beer tent full   121 of people, and wins an award (Cx: sit). It is common and generally expected for award winners to say at least a few words of thanks (Cx: world).   (7)  Cx1: Sir Quickly, who is extremely shy, won an Ox race. The award    ceremony is in a beer tent full of people. He is called to the podium to   receive his award, and his friends urge him to go up. He is extremely   hesitant, but gets up to go, while uttering the following:   Des sog’e da    glei…   that say.I   you  soon    “I’m telling you…”   ‘I’m telling you…’   a. <✓, Cx1>  …I sog  fei  nix …    …I say  fei  nothing…     “…I definitely won’t say anything…”     ‘…[I believe you don’t believe] I won’t say anything…’      …wenn’e  wos   sogn  muass geh I glei   wieda obe    …if I    something  say    must     go  I  soon again  down    “…if I need to say something I’ll go down right away.”   ‘…if I need to say something I’ll go down right away.’               (Irgendwie and Sowieso)   b. <✓, Cx1> …I sog nix    …I say nothing     “…I won’t say anything”   c. <*, Cx1> …I  sog  doch  nix    …I  say  doch  nothing     “… I won’t say anything”    ‘…[I believe you believe] I won’t say anything.’    122 Sir, in Cx1 can be reasonably expected to speak, and give a thank you or acknowledgment for the prize he won, when getting up on stage. With the use of fei, he emphasizes that, counter to expectations he won’t speak (7a). The utterance is felicitous in the same context without fei (7b) but here it lacks the ‘counter what you expect’ component. Doch in this context is not felicitous (7c); this is expected under the proposal that it expresses A believes that p.  In the constructed, minimally contrasting context Cx2, where the friends encourage Sir to go to the podium, and explicitly state that he won’t be expected to speak, fei is not felicitous (8a). In contrast, the utterance with doch is felicitous here (8b). An utterance without a DPRT (8c) is perceived as odd by consultants.  (8)  Cx2:  Sir Quickly, who is extremely shy, won an Ox race. The award    ceremony is in a beer tent full of people. He is called to the podium   to receive his award, and his friends urge him to go up, and reassure   him that he won’t have to say anything; he can just get the medal.   He gets up to go, while uttering the following:    a. <*, Cx2> I sog fei nix    I say fei nothing    b.  <✓, Cx2>  I sog  doch nix.     I say  doch nothing  “I won’t say anything.” ‘[I believe you believe] I won’t say anything.’    c. <?, Cx2> I sog nix.48     I say nothing      “I won’t say anything.”   The contrast between (7) and (8) shows that it S’s assumptions about A’s belief which                                                 48 The crucial contrast that consultants noted is the difference between (8a) with fei, which is perceived as a lot worse than (8c) without fei.   123 fei and doch are sensitive to. In (7) Sir believes that he is expected to speak; the use of fei is felicitous, doch is not. In (8) Sir believes that he is not expected to speak; fei is not felicitous, whereas doch is.   The following section introduces further tests to support the proposed function of fei and doch, showing that both DPRTs access S’s assumptions about A’s belief and thus are classified as A-oriented.   4.2.1 A believes p The proposal that ƒei expresses S’s belief that A doesn’t believe p makes predictions about the discourse conditions under which the DPRT can be used. In particular, fei should be infelicitous in contexts where S has evidence that A believes that p. This prediction is borne out as shown in (9).  (9)   Cx1: Martl and Alex chitchat. Martl tells Alex he doesn‘t have time to stick   around for dinner, since he’s going to the movies. Alex sets the dinner   table for 2 and Martl assumes the second plate is for him.     Bel (A, p)   <*,Cx1> I hob  fei koa  Zeit     zum Essn     I  have fei  NEG.DET   time    to    eat     “I really don’t have time to eat”    ‘[I believe you don‘t believe that] I don’t have time to eat.’   In Cx1, Alex knows that Martl can’t stay for dinner, since Martl explicitly told Alex that he has a movie date (Cx: disc). Even though Alex seemed to temporarily have forgotten about p, as evidenced by him setting the table for two, and thereby implicating Martl may stay for dinner (Cx: sit)., it is public knowledge (i.e. S and A should know) that Alex believes that p. In this circumstance, S (Martl) cannot use fei   124 in his response. Note that the DPRT doch in the same scenario is felicitous (10). A-oriented doch expresses that S believes that A believes and is committed to p.49   (10)   Martl: <✓, Cx1> I hob   doch koa        Zeit zum Essn     I have doch  NEG.DET  time   to   eat      “I don’t have time to eat”     ‘[I believe you believe] I don’t have time to eat.’  The next example further supports that S’s assumptions about A’s belief is an important factor for the felicitous use of doch. The example below uses two continuations for the utterance containing doch. One continuation makes reference to p already being known to A, i.e. A believes that p (11a), whereas the other continuation (11b) asks whether A didn’t know that p (i.e. it implies that S is not sure whether A knows p). As predicted, (11a) is felicitous, (11b) is not.  (11) Cx1:  Alex promises to do a chore for Dani before leaving on vacation, and   leaves the room. Roman, who witnessed this exchange, says to Dani:    Dofia  hod‘a doch koa   Zeit …   for.that  has.he doch NEG.DET time     “He doesn‘t have time for that...”   ‘[I believe you now that] He doesn‘t have time for that…‘  a.  <✓, Cx1 > …awa des woasst eh       “…but you know that anyways” b.   < *, Cx1> …host des ned gwusst?      “…didn‘t you know that?”                                                  49 I will discuss how accommodation plays a role in cases where A seemed to have forgotten about p, or never knew p in Chapter 5.   125 We further predict that doch is felicitous in situations when S wants to check information that A knows. The following example illustrates such a context. S wants to confirm A’s identity. It is clear to S that A, as a rational discourse participant, must know who he is. The use of doch here is felicitous (12a). As expected, fei is infelicitous (12b).  (12) Cx2: A villager who knows my dad from playing music runs into my    brother, whom he hasn’t met, at the village bakery. The family    resemblance is striking. He says:   a. <✓, Cx2> Du  bist  doch am Thoma sei Bua     you  are  doch DET Thoma his boy      “You must be a Thoma”      ‘[I believe you believe that] You are the Thoma family’s boy.’    b. < *,  Cx2>  Du bist fei am Thoma sei Bua  Consider the next example, which has as a premise from Cx: sit that S and A both know that Dani has lived in Munich for a while now. In this context, the most natural utterance contains doch (13a). The utterance without doch in this context is acceptable, but the consultant reported that he strongly prefers the utterance with doch. (13) Cx1:  I visit Munich. I want to go out in a particular neighborhood and ask   my friend Sylvia for advice where to go. We have a mutual friend Dani   who we both know has been living in the city for a while. She    responds:   a. <✓, Cx1> Da  Dani wohnt doch schon   a Zeit lang do….      DET  Dani  live doch already  DET time long there   b. < ?, Cx1> Da  Dani wohnt schon  a Zeit lang do….     DET  Dani  live already DET time long there        126   …dea  woass  bestimmt wos.   …he  knows  surely   something.    “Dani’s been living there for a while so he surely will have an idea.”    ‘[I believe you believe that] Dani’s been living here for a while now, he surely   will have an idea.’    If the context is changed slightly as in (14), doch is not acceptable in the discourse (14a). The altered context has a premise that I don’t know this person Dani that my friend Sylvia is talking about. I therefore cannot know where he lives. The version without doch is more acceptable in this context than in the context for (13b) above.  (14) Cx2:  I visit Munich. I want to go out in the particular neighborhood and ask   my friend Sylvia for advice where to go. She mentions a friend of hers   might know:     a. <*, Cx2 > Da  Dani wohnt doch scho  a Zeit lang do…. DET  Dani  live doch already DET time long there  b.  <✓, Cx2> Da Dani wohnt scho a  Zeit lang do…. DET  Dani  live  already DET time  long  there     …dea  woass  bestimmt wos.    … he  knows  surely   something.   “Dani’s been living there for a while, you know, so he surely will have an idea.”  ‘[I believe you believe that] Dani’s been living here for a wile now, he surely will have an idea.   Next I show that ƒei is infelicitous in self- talk scenarios, whereas doch is felicitous in these kinds of contexts. This again, is predicted under the proposal made here.  4.2.2 Soliloquy There is a special circumstance where one person instantiates both S and A, i.e., S talks to him/herself. This is known as a soliloquial context, and I refer to it as self-  127 talk.50 In these scenarios, where S = A, fei is predicted to be infelicitous. This is so, because a rational speaker knows about his or her own beliefs, and cannot tell himself or herself that I believe that you (=I) don’t believe p. The next example illustrates such a scenario, and shows that A-oriented fei is infelicitous.   (15) Cx1: Alex promises to do a big chore before leaving on vacation, and leaves   the room. Martl is muttering to himself (Martl = Addressee).    <*, Cx1 > Dofia  hod’a  fei  koa        Zeit         that.for  has.he   fei      NEG.DET  time         “He really doesn’t have time for this.”   ‘[I believe you don‘t believe] he doesn’t have time for this.’  Doch on the other hand is felicitous in self-talk contexts. This is expected, since a rational speaker knows about his or her own beliefs, and can easily "tell themselves" that I believe that you (=I) believe p.   (16) <✓, Cx1> Dofia    hod’a doch koa         Zeit  that.for  has.he  doch NEG.DET time   “He doesn’t have time for this.”   ‘[I believe you believe] he doesn’t have time for this.’  The S-oriented DPRTs ja is equally felicitous, with its own contribution due to its function. This will be discussed in detail in 4.3. For now, it suffices to point out that S-oriented DPRT ja is felicitous in self-talk.                                                    50 Soliloquy, and the discourse conditions for soliloquy has been studied by Hasegawa (2011) in detail. I here will continue to refer to it as 'self-talk' for exposition.     128 (17) <✓, Cx1 > Dofia  hod’a ja koa       Zeit     that.for  has.he ja  NEG.DET  time      “He doesn’t have time for this”     ‘[I believe] he doesn’t have time for this.’  The next example further supports the claims about doch. Again, fei is illicit.51  (18) Cx2:  I drive along and see a faintly familiar looking person crossing the    street. I do a double take, muttering to myself:   a. <✔,Cx2> Des  is  doch  da  Alex    that  is  doch  DET  Alex     “That's Alex.”    ‘[I believe you believe] that's Alex.’    b. <*, Cx2> Des is  fei da  Alex  that  is fei  DET  Alex   “That's Alex.”  ‘[I believe you don’t believe] that's Alex.’  I showed in this section that doch is licit, whereas fei is illicit in contexts where S talks to him- or herself (i.e., S = A). The same pattern emerged in contexts where it was established that A believes p. Next I show that fei is felicitous in contexts where A believes that ¬p.                                                    51 Note that proposals that assume doch to be inherently contrastive (e.g. Grosz 2010b) have a hard time accounting for the kind of data presented in (18). Here S is not correcting her own mistaken belief, or another’s person belief. She reminds herself about p, reaffirms to herself that she knows p. This use of doch, as well as other aspects of the functional range of doch will be discussed in detail in Chapter 5. For that reason I will delay further justification, but wanted to point out this problem with the assumption that doch lexically encodes contrast.     129 4.2.3 A believes that not p  In this section I show that fei is felicitous in contexts where A believes that ¬p, whereas doch is shown to be unacceptable. The following example illustrates this. Cx: sit provides indication to Regina (S) that Hanni (A) assumes she, Regina, doesn’t want coffee. Regina does want coffee, however, and can use fei but not doch in her response.  (19) Cx1: Hanni has a few friends at her house. She’s bringing coffee to a few   people, as well as some glasses of water. She puts a glass of water in   front of Regina, who says.    a. <✔,Cx1>  I  mog   fei   aa   gern      an   Kafä       I  like   fei   also  willing  DET  coffee      “I’d also really love a coffee.”     ‘[I believe you don’t believe] I also want coffee.’    b. <*, Cx1> I mog doch  aa  gern      an  Kafä  Regina has sufficient evidence from the situational context to believe that Hanni thinks she won’t drink coffee (since Hanni only serves her water). In this context, the use of fei is predicted to be felicitous (because it indicates that A believes ¬p) while doch is predicted to be infelicitous (because it indicates that A believes p).  The next example further illustrates this contrast between the two A-oriented DPRTs. In particular, S can makes assumptions about A’s epistemic state based on Cx: world. People are generally not expected to leave their zippers open when returning from the washroom, hence S can safely assume that A believes that ¬p (that the zipper is not open). Doch is correctly predicted to be infelicitous (20a), whereas fei is correctly predicted to be well-formed in this context (20b).    130 (20) Cx1: Hansi returns from the washroom. His zipper is down. Hanni says:   a. <*, Cx1> Dei Hos'ntiarl is doch auf    your pant.door is doch open     “Your fly is down.”    ‘[I believe you believe] your fly is down.’     b. <✔,Cx1>  Dei Hos'ntiarl is  fei auf    your  pant.door   is  fei   open     “Your fly is down.”     ‘[I believe you don’t believe] your fly is down.’  Next, I show that not only the discourse context and situational context, but also the interpersonal relationship between the discourse participants plays a significant role in determining the use of DPRTs.52 This can be viewed as a special case of world knowledge interacting with situational knowledge; the special cultural knowledge about assumptions we can make about the discourse participants’ belief about p. Witness the following data:  (21) a. Cx1: I’m going Christmas shopping with my cousin and am justifying the   purchase of a nice and cozy cashmere scarf for her mom, my aunt. I   say to her:    <✔, Cx1> Dei   Mama frierts doch immer  so…       your mom freezes doch always so…      … do  is des genau   des richtige.    …there  is DET  exactly DET right                                                     52 Interpersonal relationships could arguably be subsumed under situational as well as specific world knowledge. I wanted to point out the particular sensitivity of DPRTs to this aspect of the situational context, however. I will also address this in the discussion on accommodation in 5.8.   131   “Your mom is always so cold so this is just perfect.”   ‘[I believe you believe that] your mom is always cold, so this is just perfect.’  b. Cx2:  I’m going Christmas shopping with a friend and am justifying the    purchase of a nice and cozy cashmere scarf for my aunt. I say to her:     <*, Cx2>  Mei Tante  frierts doch   immer so     my  aunt   freezes doch   always so       … do  is des genau des richtige.    …there  is DET  exactly DET  right    “My aunt is always so cold so this is just perfect.”   ‘[I believe you believe that] my aunt is always so cold so this is just perfect.’  My cousin knows my aunt, her own mother, and therefore I can make the reasonable assumption that she knows about her mother always being cold. I can use doch, to show that I believe she knows this (21a). In contrast, I cannot use doch in Cx2  (21b) since someone who doesn’t know my aunt accompanies me. My interlocutor cannot reasonably know about my aunt always being cold.   4.2.4 A doesn’t believe that p In this section I show that fei can also occur in contexts where A gives no indication whether she believes p (or ¬p). That is, she has no public commitment to p or indicates that she doesn’t believe p. This is a more general condition than A believes that ¬p. Recall that S can make an assessment about A’s belief; she can believe A believes p, she can believe that A believes ¬p, but she can also be in a situation where she has no evidence at all about the epistemic states of A, or where she has indication that A does not believe p. This is different from S belief that A believes ¬p and can be reduced to having an explicit belief Bel (A,  ¬p), versus not believing that A believes   132 p, ¬Bel (A, p). I discussed the former above, whereas the latter, ¬Bel (A, p), is instantiated by out of the blue (OOB) contexts. OOB contexts don’t presuppose shared knowledge, in that it is not public knowledge that A is aware of p or believes p.  S can also be in a position where she has actual evidence that A has no belief about p, that is, A doesn’t believe p, but also doesn’t believe ¬p. This is instantiated when A has asked a question about p. The fact that A is asking a question is evidence for S that A does not know the answer, i.e. A does not believe p or ¬p.  This is summarized below.  (22) Two types of contexts for ¬ Bel (A, p)  (i) Out of the blue  (ii) S answers A’s question about p  Besides fei, doch also can occur in some contexts where A has no belief about p.53 However, in contexts that have as a premise that S has to assume that A doesn’t believes p, doch is predictably infelicitous. In this way, DPRTs explicitly restrict possible responses and discourse moves on the part of A.54  4.2.4.1 Out of the blue context When fei occurs in OOB utterances, it refers to salient propositions, which have not explicitly been uttered in the context. Doch in contrast to fei is infelicitous OOB,                                                 53 I will address the issue of accommodation, and how DPRTs often appear to force accommodation in Chapter 5.  54 Davis (2011) analyzes this in terms of a domain restriction over possible output contexts.   133 since it refers to a (positive) belief about p. Yet OOB contexts are those where it is public knowledge that A is unaware of p. The next two examples illustrate this.   (23) Cx1:  Hansi returns from the washroom. His zipper is down. Hanni says to   him:   a.  <✔, Cx1> Dei Hos'ntiarl  is  fei auf      your pant.door  is  fei   open      “Your fly is down.”      ‘[I believe you don’t believe] your fly is down.’   b. <*, Cx1>  Dei Hos'ntiarl  is doch auf     your pant.door is doch open      “Your fly is down.”     ‘[I believe you believe] your fly is down.’   (24) Cx2:  First thing said by an aunt to her niece whom she hasn’t seen in a year.     Griass  de  Maus,  wia  geht’s  da  denn?  Mei,…    Greet  you  mouse  how  goes.it  you  denn?  PRT…     “Hi darling, how are you? My, …    a. <✔, Cx2> … du   bist   fei   ganz    schee  g’waggs’n!     … you are    fei   whole  nice    grown      “…you really have grown quite a bit.”     ‘…[I believe you don’t believe] you have grown quite a bit.   b. <*, Cx2> …du bist  doch  ganz   schee  g’waggs’n!  (23a) and (24a) suggests that fei accesses more than the public commitment of A to ¬p, i.e. in cases where the context provides evidence that A believes ¬p.  Fei can be used as long as S has evidence that A doesn’t believe p. Fei crucially cannot be used OOB in contexts where A is, or should be aware of what is asserted in the utterance. Personal experience predicates such as ‘frieren’ exemplify this, since the person   134 undergoing the experience is the EPR, and the judge of their experience (cf. Stephenson 2007). In contrast to fei, doch is good with those kinds of predicates, predicted under the proposal made here. Since A is the expert on her feelings and bodily sensations, doch can be used to ‘remind’ A of her own state.55  (25) Cx1:  I say to my partner, who is sitting next to me shivering:     a. <*, Cx1 > Di     frierts  fei           you  freezes.it   fei      “You’re really cold”     ‘[I believe you don’t believe] you are cold.’  Consultant’s comment: ‘Des ko ma ned sog’n, des merkt derjenige ja sejba’ (you can’t say that, he notices that himself)   b. <✔, Cx1 > Di frierts      doch       you freezes.it  doch       “You are cold.”      ‘[I believe you believe that] you are cold.’  The consultant’s comment in (25a) is illuminating: it is fully consistent with the proposal made here, according to which fei expresses that A doesn’t believe p. In the context above, A necessarily must believe p, since A is the experiencer. It is illicit to tell A something that they necessarily know.56                                                     55 More on the reminding function of doch in Chapter 5.  56 Of course we can imagine contexts where a person would be unaware their own sensations. For example a little kid is playing in the lake until her lips turn blue, and she stand there shivering. A mom could possibly say ‘Di frierts fei’ under the assumption that the kid is not aware of her own cold body.    135 4.2.4.2 Answer to a question Another testing ground for the A-orientation of fei and doch are their use in answers to questions. Following Hamblin (1958), to know the meaning of a question is to know which propositions count as direct answers to that question. Asking a question implies that one does not know the answer to that question (rhetorical questions and examination questions aside). I begin with fei to illustrate this. With the use of the yes/no question in (26) S calls on A to reply to p or ¬p. To use either fei or doch in an answer to a yes/no question is correctly predicted to be infelicitous because fei is used to express S’s belief that A believes ¬p (26a), while doch is used to express S’s belief that A believes p (26b).    (26) Cx1:  Christa is about to go to the store and while getting ready asks whether   it is currently still snowing.   Christa: Schneibts   no?     snows.it  still     “Is it snowing still?”    Karl:   a. <*, Cx1>  Ja, s  schneibt fei.      yes  it snows  fei      “It’s snowing.”     ‘[I believe you don’t believe] It is snowing.’      b. <*, Cx1> Ja,   s schneibt doch     yes it  snows      doch      “It’s snowing.”     ‘[I believe you believe that] it’s snowing.’     c. <✔, Cx1> Ja,  s   schneibt     yes,  it    snows          “Yes, it is snowing.”   136 To answer a yes/no question with an utterance that contains ‘I believe you believe p/¬p’ violates the maxim of Relevance (Grice 1975) because A - by asking a question - clearly expresses that they don't know p.57  For completeness, note that fei is not ruled out as a response to yes/no questions as long as this response is not a direct answer as in (27) below. Here Hans’ question implies the proposition: You are not cold. This is so because by asking this question, Hans assumes that Johanna is warm enough, so a window can be opened. In this context, as predicted, fei is felicitous: Johanna responds that counter expectation she is cold.   (27) Cx:  Hans is asking whether he can open the window in the living room   Hans: Konn’e as Fensta aufmacha?    “Can I open the window?”    Johanna: Mi    friats fei    me   freezes fei      “I’m really cold.”   ‘[I believe that you don’t believe that] I am cold.’          This exemplifies the inferencing processes at play with the use of DPRTs (cf. Schmerse et al. 2013).  Now consider DPRTs used in answers to wh-interrogatives. With the use of a wh-interrogative, S calls on A to respond to something they don’t know, the variable expressed as the wh-word. In this context doch and fei are felicitous: Doch is used when of all the possible values which the variable represented by the wh-word can take, the value in the answer is one that A already knew (28a) while fei expresses that                                                 57 See Chapter 5 for more discussion on the relation between DPRTs and the Gricean maxims.   137 it is the one A didn’t know (28b).   (28) Cx2:  Hans has several guests in his house; they are all chatting with each   other. He has been asking who wants what to drink, but hasn’t gotten   conclusive answers. He asks:   Hans: Wea mog jetz oiss an Kafä?         “Who all wants coffee now?”    Johanna:    a. <✔, Cx2> D’Regina woidd  doch oan.     DET.Regina wanted  doch  one             “Regina wanted one.”         ‘[I believe you believe that] Regina wanted one.’    b. <✔, Cx2> D’Regina  woidd  fei   oan.             DET.Regina  wanted  fei  one              “Regina wanted one.”             ‘[I believe you don’t believe that] Regina wanted one.’    Summarizing this section, I showed that the DPRTs fei and doch are both used when S has a belief about A’s attitude towards p. The contexts used to test this claim are summarized in Table 14.  A belief Orientation DPRT Bel (A, p) A doch Bel (A,	¬p) A fei ¬Bel (A,p)/no belief about p A fei/doch Table 14: A-orientation and epistemicity  As a final support for the proposed contributions of the two DPRTs and their A-orientation take the minimal pair below. It shows that fei is felicitous in a context where the negative bias of an utterance is emphasized via focus accent on the   138 sentential negation (29a).  (29) Cx3: 3-year-old Marinus is trying to get into the locked living room,    which is closed off for Christmas preparations. His grandpa tries to   tell him that he can’t go in there so that Christkindl can come with   his angel helpers to bring presents and a Christmas tree. Grandpa   begins with:  a. <✔, Cx3>  Du konnst  fei   NED  do      neigeh             you  can      fei    NEG    there  in.go      “You really can’t go in there.”           ‘[I believe you don’t believe] You can’t go in there’  b. <*, Cx3>  Du konnst doch  NED  do     neigeh.            you can doch   NEG   there  in.go     ‘[I believe you believe] you can’t go in there’  Doch in contrast is infelicitous when negative bias is highlighted, as illustrated by (28b). It shows that doch is not compatible with a focus on the opposite polarity of the content of the utterance. This is predicted, since it goes against the core function of the DPRT doch, which highlights that A believes the content of the utterance.  4.3 S-orientation In this section, I discuss an S-oriented DPRT, namely MB ja. I present evidence that it serves to express S’s attitude toward the propositional content of an utterance. Hence, it can be classified as an S-oriented DPRT. As with A-orientation, I use contextual information to test the beliefs of S concerning p. In addition, I will use the context form Cx: form. Recall that I argued in 3.4 that form is the output of clause typing and intonation, and has a commitment associated. I also adopt an assumption from Rett (2013, 2014) that commitment serves as a proxy for belief. I therefore use.   139 Cx: form in the next section as an additional test frame to establish S-orientation. As with A-oriented DPRTs, a belief about p can come about via an overt linguistic antecedent in the discourse context Cx: disc, or via the situational context Cx: sit, instantiated by e.g. actions or situations the discourse participants are involved in.   (30) Test frames for S-orientation  (i) A belief irrelevant  (ii) S believes p:     a.  Bel (S, p) established by Cx: form   b. Bel (S, p) established by Cx: disc (immediately preceding discourse    or lexical content of matrix embedding verb)  4.3.1 A belief is irrelevant The first step in establishing that ja is S-oriented, is to show that it is not A-oriented. That is, with the use of ja S signals that s/he has an explicit attitude towards p whereas A’s epistemicity is irrelevant for the felicitous use of ja. Hence, we expect that S-oriented DPRTs can be used in conversations with strangers. Since ja is not signaling what S thinks about A’s attitude towards p, S need not know anything about A’s epistemicity. This prediction is borne out, as shown in (31a). The behavior of ja contrasts with that of A-oriented fei (31b), and doch (31c), which are both ruled out in this context.58                                                  58 Note, that a shared knowledge analysis for ja cannot capture this data adequately; ja in this example cannot express a presupposition on S’s side that A shares the propositional content expressed in the utterance.     140  (31) Cx1:  My mom tells a stranger next to her on the airplane about her regular  trips to Canada:   a. <✔ ,Cx1> I hob   ja    Enkel               drum…      I have ja  grandchildren  over.there…       …do  fliag’e  mindastens  oamoi  im   Jahr   …there  fly I   at.least   once   in.DET  year    “I have grandchildren over there, so I fly at least once a year.”   ‘[I believe that] I have grandchildren over there, so I fly at least once a    year.’    b. <*, Cx1>  I hob    fei  Enkel               drum…           I have  fei  grandchildren  over.there   c. <*, Cx1>  I hob  doch Enkel              drum…         I have  doch grandchildren  over.there  Recall from section 4.2.4.1 that both A-oriented DPRTs (fei and doch) are infelicitous in contexts that refer to personal experience. If ja is S-oriented only, it follows that it should b well-formed in such contexts because it does not express an assessment about A’s epistemic state. Observe the next example with a personal experience predicate.   (32) Cx1: I say to my partner, who is sitting next to me shivering:     Di frierts   ja…      you freezes.it ja         …ziag  da  liawa a  Joppn  oo.    …pull  you  rather  DET  jacket on      “You’re cold…you had better put a jacket on.”   ‘[I believe that] you’re cold… you better put a jacket on.’ If indeed A’s belief about p is irrelevant for the felicitous use of ja, we further predict   141 that it should be possible in contexts where A believes p, as established by the situational context (Cx: sit). This prediction is borne out as shown in (33).   (33) Cx1: Alexander is being accused by his classmates of stealing a special kind   of paint. His classmates interrogate him, he denies. At the end, one    suggests to look into his bag:    Alexander:  Vo  mir  aus,…       of me out…       …schaugts hoid  in mein  Schuiranzn  findts      ja   eh  nix.   …look        hoid  in   my school.bag  find.you   ja  eh  nothing     “If it’s for me, look into my bag, you're not gonna find anything anyways.”   ‘[do] look into my bag, [I believe[it was the case before that]] you’re not going   to find anything.’  In this context, A (the classmates) believes that p (they are going to find the paint) whereas S believes ¬p. This establishes that it is not S’s belief about A commitment that matter for the use of ja.59   The same effect, namely that A-belief is irrelevant for the felicitous use of ja is also observed when A’s commitment to ¬p is explicitly introduced by a linguistic antecedent (i.e., via Cx: disc). Consider (34). Here A (Martin) is committed to ¬p (‘I don’t want to watch this much longer’). Effendi (S) uses ja in his retort ‘you don’t have to (watch this much longer.)’.   (34) Cx 3:  Martin and Effendi are watching an Ox race. Sir Quickly and his ox   Ringo are part of the race. Martin is getting impatient with the antics   of Sir and his ox.  Martin:  Lang  schaug-e-ma  des  fei nimma  oo    long   look.I.me  that  fei  NEG.anymore  on                                                 59 Ja crucially, does not refer to shared CG knowledge between S and A, since in context (31), S and A have the opposite beliefs. More in Chapter 5.   142     “I’m really not gonna watch this much longer.”    ‘[I believe you don’t believe that] I’m not gonna look at this much     longer.’    Effendi:  Muasst      ja ned,  is    ja   imma    schnäi  vorbei       must.you  ja  not  it.is  ja  always fast       over     “You don’t have to, it’s always over quickly.”    ‘[I believe that] you don’t have to, [I believe that] it (the race) is     always over quickly.’    Furthermore, if the use of ja is not sensitive to A’s attitude towards p, then we predict that a statement with ja can be followed by continuation that indicates either that A believes p, or that A believes ¬p. This prediction is borne out, as shown in (35). (35a) shows a continuation of the preceding utterance which indicates that A knows p, whereas (35b) shows a continuation that indicates that A doesn’t know p. Both are acceptable in this context.   (35) Cx4: Alex promises to do a chore for me before leaving on vacation, and  leaves the room. Roman doubts his promise and says to me:    Dofia hod-a- ja  koa      Zeit   for.that has.he  ja  NEG.DET   time    “He doesn‘t have time for that”    ‘[I know] he doesn have time for that.’     a. <✔, Cx4> …Awa  des woasst eh  “But you know that anyways”    b. <✔, Cx4> …Host des ned gwusst? “Didn‘t you know that?”  In essence, no matter whether A believes p or believes ¬p, ja is felicitous. This is a clear indication that A belief does not play a role. I show in the next section that ja is sensitive to S belief about p.    143 4.3.2 S belief is relevant I show in this section that tests referring to S belief about p can be successfully employed in identifying those discourse contexts in which S-oriented ja is felicitous, as well as contexts in which it is not. Furthermore I show that ja is also sensitive to S’s belief by proxy of commitment. For this reason I take Cx: form into consideration as a further test frame that can indicate S belief. Cx: sit, Cx: disc, as well as Cx: form, can serve as an indicator of S belief about p, and are used to establish the necessary conditions for S-orientation  I show that ja is felicitous in contexts where S is committed to p (Bel (S, p)) , and that ja is not felicitous in contexts where S is committed to ¬p (Bel (S, ¬p)). This is true for contexts where S’s belief is established via Cx: disc (4.3.2.1) as well as contexts where S’s belief is established via Cx: sit (4.3.2.2)  4.3.2.1 S committed to p due to discourse context This section establishes that S commitment to p is relevant for the felicitous use of ja. ja occurs in declaratives with assertive force (Coniglio 2009, Kwon 2005, Thurmair 1989) hence they commit S to p. The minimal effect of asserting p is to add p to the public commitments of S; it publicly commits S to act as though she believed p (Condoravdi and Lauer 2011, Zeevat 2003, cf. Rett 2013).  However, ja is felicitous not only in matrix declarative assertions, as was illustrated in the examples in the previous section, it can also be found in some embedded clauses (cf. Coniglio 2009, Kwon 2005).60                                                  60 Zimmermann doesn’t consider SHG ja to be embeddable „[…] ja is generally impossible in complement clauses […], except under verba dicendi (often with subjunctive mood), even though there is no binding relation involved and even though a plausible interpretation is available in principle. […] ja is always evaluated with respect to the utterance context. Hence, it cannot be embedded, unless it   144 (36) Cx1:  Hanni and Gitta have been discussing over the last two days what to   give Katl for her upcoming Birthday. Hanni calls Gitta one morning   and says:    Mia     is  grod  eigfoin…    to.me  is   just   fall.in        … dass d  Katl  ja   scho     lang  Gebursdag ghabt  hod.   …that   DET  Katl  ja   already long   birthday      has      had    “I just remembered that Katl already had her birthday a while back.”   ‘I just remembered that [I believe that] Katl already had her birthday a while   back.’  The embedded clause in (36) inherits commitment from the embedding matrix verb (cf. Simons 2007). Hence, the subordinating clause Mia is grad eigfoin instantiates Cx: disc and determines whether S is committed to the proposition expressed in the subordinated clause. Felicity therefore depends on the lexical content of the embedding verb.  The proposal that ja expresses S belief in p predicts that it is infelicitous in contexts where S does not believe p, as presented in (37). (37a) shows ja embedded under a positive, attitude verb which report on the “private mental state of an individual” (Anand and Hacquard 2014). In embedded contexts like (37b), where the matrix attitude predicate is negated, ja is infelicitous.   (37) Cx:  Tina and Roman talk about Vroni’s English skills. Tina says:   a. I glab dass’  ja scho  eftas in Kanada  war   I believe  that.she  ja   already  often  in Canada was    ‘I think that [I believe that] she’s been to Canada several times so far’                                                                                                                                         forms part of a reported speech act under a verbum dicendi  […].” (Zimmermann 2007:20f.) The data in (37) and (38a) show that this is not quite correct for MB (cf. Coniglio 2009, Doering 2008).    145   b. *I glab     ned  dass’      ja   scho      eftas  in Kanada war     I believe NEG   that.she  ja   already  often in  Canada  was   Recall from Chapter 2 that DPRTs are known to be impossible under the scope of negation. The infelicity of (37b) could therefore stem from the fact that ja is embedded under negation. Consider the next example, however. The verb vagessen (‘forget’) presupposes the truth of the complement, and commits S to p, whether it is negated or not. ja is predictably acceptable in both cases, negated and not (38)-(39). This is evidence that it is the interaction of the lexical content of the embedding verb with negation, and not the presence of negation alone, which results in the (in)felicity of the S-oriented DPRT ja.   (38) Cx: Mom remembers that winter holidays start on Monday. She says to her   kids:   Am Monddog miasst’s ned in’d Schui...     “On Monday you don’t need to go to school…”    …I  hob  ganz  vagessn  dass  ja  da      scho  Ferien       san   …I  have  whole  forgot    that  ja  then   already   holidays    are    “…I totally forgot that holidays will have started by then!”   ‘…I totally forgot that [I believe that] holidays have started then already’  (39) Cx: The parents are talking about possible dates for booking a vacation.   They are trying to find a suitable date. Mom suggests the first week of   August for the getaway. Dad says:   Du   deafs ned  vagessn dass  ja do   übaroi Ferien     san…   you  may    NEG   forget    that  ja there  everywhere holidays  are…    “Don’t forget that everybody will have vacation time then…”   ‘You shouldn’t forget that [I believe that] it’s vacation time everywhere then...’   …do weads ganz schee zuageh  “…it’ll be pretty busy then.”        146 Ja cannot be used in the complement of verbs such as, leugnen (deny), hoffa (hope), frang (ask). It is inconsistent to hope, ask, deny, etc., but to believe p at the same time (cf. Hentschel 1986). This is consistent with the proposal that ja marks the host proposition as one which S explicitly believes to be true.   4.3.2.2 S committed to p due to form I present now another context where S is committed to p as a ‘fact’ (Zanuttini and Portner 2003), namely exclamations. Exclamations are a special context Cx: form. As discussed in 3.4, exclamations in MB are not always overtly marked with dedicated UoLs, or have a dedicated syntactic clause type, although some exclamatives are marked syntactically (cf. Brandner 2010, Rett 2011 for a detailed rationale about the distinction between exclamative and exclamation). They are characterized by a special exclamation contour represented below as √. This contour is realized as an extra high pitch accent, often on the first word, which is marked by CAPS.  (40) Cx1: 5 year old Elias is coming into the house all dirty after playing outside  in the mud on a rainy day. Grandma greets him at the door:    DU  schaugst  schee aus √     you  look         nice   out        “(My,) how you look!”        ‘You look nice.’  My proposal predicts that ja is felicitous in this context, since an exclamation commits S to p. This prediction is borne out as shown in the next example.   (41) Cx1: 5 year old Elias is coming into the house all dirty after playing outside   in the mud on a rainy day. Grandma greets him at the door:   147    DU schaugst  ja   schee  aus √           you   look        ja   nice    out     “(My,) How you look!”    ‘[I believe that] you look nice.’   The next example shows another exclamation context. The specific situation has no addressee. This underscores the claim that ja does not reference the epistemic state of A. Note that the A-oriented DPRTs fei and doch are not acceptable in this self-talk context. The example below is adapted from Kratzer and Matthewson (2009).  (42) Cx2: I am sitting in the coffee shop, working on my diss. I check the time,   and realize how late it is. I mutter to myself:  a. <✔, Cx2>  I muass  ja  in 5 Minuten  in da  Abteilung sei!     I must    ja  in 5 minutes in DET  department  be     “(Gosh), I gotta be in the department in 5 minutes!”    ‘[I believe that] I gotta be in the department in 5 minutes.’   b. <*, Cx2> I muass doch in 5 Minuten in da Abteilung sei!   c. <*, Cx2>   I muass fei in 5 Minuten in da Abteilung sei!    I have shown in this section that ja is felicitous in contexts where S believes p. Relevant contexts are root declaratives with assertive force, complements of verbs that express the epistemic state of the speaker, and exclamations, which do not require A. This is summarized below.  S belief Orientation DPRT Bel (S, p) S ja Bel (S,¬p) S -  Table 15: S-orientation and epistemicity   148 4.4 Predictions so far In this section I show some of the predictions of the proposed S- and A-orientation of the DPRTs discussed so far. I argued that DPRTs are sensitive to the epistemic states of the interlocutors. I also showed in Chapter 3 that the form of an utterance, i.e. its clause type and associated intonation contour, establish the particular epistemic stance of a discourse participant toward p. We would therefore expect these sources of epistemicity to interact. As I show in this section, the sensitivity of DPRTs to clause-type is the direct result of the compatibility between the epistemicity expressed with a DPRT, and the epistemicity expressed in Cx: form. Furthermore, for A-oriented DPRTs I show more contexts that show S’s assessment about A belief, and how A- oriented DPRTs interact with those.    A-oriented DPRTs:  I showed above that A belief is involved in the felicitous use of fei and doch. This sensitivity to A belief predicts that both DPRTs are felicitous in contexts that establish that A is committed to p. This is borne out in imperatives, where Cx: form establishes A commitment only. Consider the following examples.  (43) a. ✔ Gäh fei  in’d Uni!      go fei in.DET  Uni     b. ✔ Gäh doch  in’d Uni!   go    doch   in.DET  Uni  In contrast, we predict that A-oriented DPRTs cannot occur in contexts which do not require any belief on the part of A, but S belief only. This is borne out in exclamations, which do not admit doch or fei (44a-b).     149 (44) a.  *  Da  MARINUS gibt doch   da  Luzie  a    Bussl   √    DET Marinus  gives  doch  DET  Luzia DET  kiss    b. * Da  MARINUS gibt fei   da   Luzie  a     Bussl  √     DET Marinus   gives  fei    DET   Luzia   DET  kiss  Moving from Cx: form to another context, we can see that the proposal made here also makes correct predictions for answers to questions. I showed above that doch and fei are predictably infelicitous in answers to wh-interrogatives, if the person who asks the question can be assumed to genuinely not know the answer. This is illustrated in another example in (45).  (45) Cx3: At the main train station in Munich, Andreas, not a local, asks a    stranger for directions to Marienplatz:   Andreas: Entschuidgns,   wia   kumm  e’n       am         bestn  zum     Marienblotz?            excuse.me.2PL  how come   I.denn  on.DET   best    to. DET Marienplatz            “Sorry how do I best get to the Marienplatz?”   Stranger:    a. <✔, Cx3> Do miassn’s    d’S-Bahn           nehma,   oda    d’U3.    there must.you   DET.rapid.train   take,       or       DET.underground 3       “You gotta take the S-Bahn, or the U3.”   b. <*, Cx3>  Do     miassn’s   doch  d’S-Bahn           nehma  oda   d’U3.    there  must.you  doch  DET.rapid.train   take,     or      DET.underground     “You gotta take the S-Bahn, or the U3.”   ‘[I believe you know that] You gotta take the S-Bahn, or the U3.’  However, as illustrated next in (46), doch is felicitous if the same utterance appears in a different context. If S can reasonably assume that A knows, or is supposed to know about p, doch is predicted to be well-formed in the answer to a question. This could be the case in a context of a conversation among friends, who can be assumed to   150 know about each other’s general epistemic state regarding certain facts (such as being familiar with a city).   (46) Cx4:  Andreas asked his friend Dani for directions to a very popular and    well-known landmark in Munich. Dani knows that Andreas has   been using public transit in Munich for years.    Andreas:  Du   Dani…     you  Dani…      … wia   kumm e’n am bestn zum  Stachus?      … how come I.denn on.DET  best  to.DET   Stachus     “Hey Dani, how do I best get to Stachus?”  Dani:   Do  muasst  doch  d’S-Bahn nehma.    there  must.you  doch  DET.rapid.train take      “You gotta take the S-Bahn.”    ‘[I believe you know that] you gotta take the S-Bahn.’  Finally, consider a teaching context, where it is common knowledge that the interlocutor asking the question (i.e., the teacher) knows the answer to the question. In this example, modeled after Zimmermann (2008:ex 9b), doch is again predictably felicitous in the answer, whereas fei is not.   (47)  Cx5: The teacher asks Marinus the study questions from the day before:   Teacher:  Marinus, wos is jetz unsa Kreisstodt?         Marinus,  what is jetz our county.city      “Marinus, what’s our county seat?”   Marinus:    a. <✔, Cx5> Des is  doch  Miaschboch,  (oda)?     that is  doch  Miesbach  (or)     “That’s Miesbach, isn’t it?”   151    ‘[I believe you know that] That’s Miesbach.’   b. <*,Cx5> Des is fei Miaschboch, (oda)?    that  is  fei  Miesbach  (or)   S-oriented DPRTs:  I showed above that only S commitment, and no A commitment is involved in the felicitous use of ja. This predicts that the DPRT is infelicitous in contexts where no S commitment, but only A commitment is expressed. This is borne out in imperatives, where Cx: form establishes A commitment only.  (48) <*, Cx1 >  Gäh  ja  in’d  Uni!61     go   ja    in.DET  Uni  Furthermore, ja should be equally infelicitous in questions, since questions commit A to p. This is borne out in the next example. The clause type in the example is a declarative, yet the rising intonation / commits A to p, and results in question force. Recall that declaratives with \, resulting in assertion force, admit ja (49a).  (49) a. ✔ Da Marinus gibt   ja da Luzia  a   Bussl  \      DET Marinus gives ja DET Luzia DET kiss     “Marinus is giving Luzia a kiss.”     b. * Da Marinus  gibt ja da Luzia  a   Bussl / ?    DET Marinus gives ja DET Luzia   DET kiss   Questions expressed with wh- and polar interrogatives both also commit A to p, and                                                 61 Note that accent on ja renders this example felicitous. I turn to accented ja in Chapter 5, where I propose that it is O-oriented.    152 also do not allow for ja, equally expected.   (50) a. * Wea gibt ja da Luzie a  Bussl  / ?   who gives ja DET Luzia DET kiss  b.* Gibt da Marinus  da   Luzia  ja   a Bussl / ?   gives DET Marinus  DET   Luzia  ja   DET kiss  Recall from Chapter 3 that V1 clauses do not express S commitment to p, since they do not assert (Reis 1999). Therefore, when uttering a V1 utterance, S expresses no belief about p. Ja is predicted to be infelicitous in this context. This is borne out as shown in (51).  (51) *Gibt da Marinus ja   da Luzia   a  Bussl  \     gives  DET  Marinus  ja DET Luzia  DET  kiss           These interactions with the different clause types show that DPRTs show restrictions based on the compatibility between the belief expressed with the DPRT, and the belief (via commitment) expressed via the features of formal clause typing and the associated intonation. This suggests that the clause type dependency of DPRTs is mediated indirectly, and not based on a direct (syntactic or formal semantic) dependency.   4.5 O-orientation So far I have shown several ways of establishing whose belief is expressed in an utterance. I used a variety of contexts to do so. The individuals whose point of view was being tested were the discourse participants S and A. I now show that we need to   153 consider another point of view, as well. Since it does not directly target the discourse participant roles S and A, respectively, I will refer to it as O, for ‘other’. That is, O is not dedicated to targeting either S or A, rather O-oriented DPRTs display variable (chameleon-like) behaviour. Depending on the context, O can be instantiated by (the individual who is currently) S, or (the individual who is currently) A or some other individual who is currently neither S nor A. Thus, O can be understood as a variable epistemic reference point (ERP) or epistemic judge (in the sense of Stephenson 2007) of an utterance. I argue that the DPRTs jetz, and eh are associated with O.   The default ERP of a matrix declarative clause with assertive force is the utterer of the sentence, and generally coincides with the speaker S.62 In such contexts, O-oriented DPRTs will appear to be S-oriented. The ERP can be different from S, however (cf. Doherty 1987, McCready 2007, Stephenson 2006). For example, in interrogatives, the ERP is A while in utterances expressing a subjective judgment, the ERP is the judge.63   I show now that the ERP can be S, A, or O depending on the syntactic context, and the lexical items involved. Certain contexts can lead to ERP shift, where the evaluation context is different from the context of the proposition (Banfield 1982, Doherty 1987, Döring 2013).64                                                  62 There are contexts where the utterer does not coincide with the speaker. This is the case in indirect speech, for example, as well as the historical present (Banfield 1982, cf. Döring 2008, 2013), where the utterer of the proposition differs from the discourse participant or narrator of the story.   63 McCready (2007) suggests that S, A and O (judge in McCready’s 2007 terms) are seperate entities in a discourse context. He suggests that Kaplanian contexts are tuples of the following form: “Cx=a,i,t,l,j, a is agent of the context, i = interlocutor, t is time of utterance, l the location of utterance, and j is the judge.” I adopt this idea here, and show that O-oriented DPRTs support the need to separate the immediate discourse participants S (McCready’s agent), A (McCready’s interlocutor) and a separate other reference point O (McCready’s judge) (cf. Truckenbrodt’s 2006 contextual index).  64 Schlenker (2004) shows that indirect speech, as well as narratives in the historical present, are shifting contexts as well, where the speaker and the attitude holder, the EPR do not coincide.   154  We saw in previous sections that in embedding contexts, the embedded clause inherits the main clause’s EPR (i.e. S). In light of this discussion, consider now the following MB example. In (52) it is S, the person uttering the sentence, who thinks that Hans should be home (cf. Hacquard 2010). The modal in the example below has to have an epistemic interpretation. It expresses a necessity, given what the speaker knows at the time when the sentence is uttered. A paraphrase for (52) is, ‘given what I know now, it is necessary that Hans was home.’   (52) Da  Hans  miassad scho     lang  dahoam sei  the  Hans  should already long  at.home be   “Hans should long be home by now”    (modeled after Hacquard 2010: ex1)  Now observe the next example. In the subordinated clause in (53) the ERP is the subject of the embedding clause.  (53) D’ Hanni   moant…  DET Hanni  means   … dass da   Hans  scho     lang  dahoam sein miassad…   … that  DET Hans  already long   at.home be    should    ….awa I  glab    dass’a   heit    ned voa achte hoamkummt   … but I  belive  that’he   today NEG before  eight  home.comes    “Hanni thinks Hans should be long home by now, but I think he’s not gonna come home before eight.”  When an epistemic modal is embedded under an attitude verb, it does not necessarily inherit S as ERP: the modal is not interpreted relative to S. Instead, in this context the   155 ERP is the matrix subject (Hanni in 53) (Hacquart 2010:81, cf. Speas 2004, cf. Stephenson 2007).65  In (53) S and the subject are necessarily different individuals with different belief states. The propositional attitude verbs moana (‘be of an opinion’) and glam (‘believe’) each assign their subjects as the ERP of their respective complement clauses. In (53) this is a third person, Hanni for the former, whereas for the latter it is a first person pronoun, which indexically refers to S. It is in this sense that the ERP is the person from whose perspective a proposition is evaluated (cf. Waldie 2012). It can coincide with S, A, or some third person O, but it crucially depends on the subject, which can be S, A or a third person O.   I now introduce DPRTs that shift their reference point depending on the ERP. I begin with jetz and then turn to eh. To test for O-oriented DPRTs we have to explore contexts with variable ERPs. I use the following test frames for establishing the O- orientation for both these DPRTs. Note that the shifting contexts are the ones which establish the shifting behavior of O-oriented DPRTs, whereas (ii) and (iii) are contexts in which S does not need to have an attitude about p, and are therefore compatible with O-orientation.   (54) Test frames for O-oriented DPRTs  (i) shifting behavior: can occur in interrogatives, declaratives and imperatives  (ii) no S belief about p is expressed: instantiated by V1 clauses with \  (iii) can occur in Cx: disc which establishes S doesn’t know p                                                  65 Hacquard (2010) considers this an ‘attitude-holder’ oriented event, and distinguishes this from subject oriented events. Each of these events result in a different modal interpretation. This does not bear directly on the facts discussed here; the crucial point is that S is the ERP for contexts such as (52), whereas in contexts such as (53) the ERP is different from S.    156 4.5.1 Jetz By using the test frames introduced above, I show now that the DPRT jetz is O-oriented. It expresses that p is particularly salient at the time of utterance, in line with its core lexical meaning introduced in Chapter 3. The temporal content of the UoL jetz is not interpreted relative to the event, but relative to the speech act as fDPRT (see Chapter 6 for the role of the lexical content of the UoLs, which doesn’t itself indicate orientation). I proposed the following contribution for the DPRT jetz.  (55) jetz ≈ p is salient for O now  The first test frame used to probe for the O-orientation of jetz is its occurrence across a variety of syntactic forms, attesting to its shifting behavior. In this respect it differs from S- and A-oriented DPRTs. For O-oriented DPRTs, the EPR shifts depending on Cx: form. We see in the following examples that jetz can occur in declarative assertions interrogatives, and imperatives alike.   (56) Cx: Regina asks her two kids about who’s going to help her move some   things:    Wea  huift  ma  jetz  dann moang?   who  helps   me jetz  then tomorrow    “NOW who’s gonna help me tomorrow?” 66   ‘[It is salient now] who is going to help me tomorrow.’                                                   66 Note that this is not intended with a comma intonation, akin to “Now, who’s gonna help me tomorrow.”    157 Jetz in this wh-interrogative co-occurs with a temporal adverb moang ‘tomorrow’, again indicating that its temporal interpretation is not relative to the event.   In another example from my corpus we find an exclamation based on a wh-interrogative. Recall that exclamations express S commitment only (Rett 2011).  (57) Cx: Sepp,the driver of a beverage truck left his truck running and got out to   talk to the owner of the inn he is delivering to around the house. The   truck runs into the side of the house with a big crash. Sepp comes    running around the corner, yelling:    Scheiße! Scheiße, wia hod denn des  jetz     so  sei  kena?! Scheiße!...  shit  shit         how has denn this  jetz    so   be  can shit    … de Wand  de zoi’e dia natürlich, I Depp!...      DET  wall  DET pay.I you of.course I idiot   …Aja, des is  weil’e        ma   des einfach   oogwohnt    hob!      A       varreck!     DM   DET is  because.I   me    DET einfach   gotten.used.to   have. DM     die    “Shit! Shit, just how could this have happened … ‘Shit! Shit! [It is salient now that I am puzzled at] how this could have happened…’  …of course I’ll reimburse you for the cost of the wall, the idiot I am! Man, I just have gotten used to doing this! Goddamn!”        (Wer früher stirbt)  In this example, jetz occurs in an exclamation. The accident, the truck running into the wall, already happened, another indication that jetz cannot refer to the (time of the) accident itself.  Evidence that jetz is not S-oriented comes from the fact that – unlike S-oriented DPRTs - it can occur in A-oriented imperatives as well.   (58) Cx: Regina calls Lukas and requests that he help her move furniture    tomorrow:    Kumm jetz bitte  moang vorbei…    158   come  jetz please tomorrow over    … dass’e  den Schrank nausbringa  ko!   …that.I  DET closet out.bring  can   “Come over tomorrow please so that I can get the closet outside!”  ‘[It is salient now that you] please come over tomorrow so that I can get the closet  outside.’  Again, as in the previous examples, jetz co-occurs with a temporal adverb moang ‘tomorrow’, indicating that it does not have the function of a temporal deictic, but that of a DPRT.   Other contexts that are proposed as a suitable test frames for O-oriented DPRTs (and were shown to be illicit for S- and A-oriented DPRTs) are those in which no S belief about p is expressed. As established previously, I assume following Reis’ (1999) that this is the case in V1 clauses with falling intonation \. Observe the following data.  (59) Cx: Lukas is supposed to help Regina move some furniture in her house.   He disappears, and Regina, looking for him around the house finds out   from Marein, who is also helping to pack, that he left for home. Regina   says:    Geht  jetz  der  tatsächlich hoam \.    goes  jetzt  he   indeed       home     “Huh, he really went home.”    ‘[It is salient now that] he indeed went home.’   Note that since Lukas is already gone, jetz cannot have temporal reference relative to the event in this example. Instead its temporal reference is relative to the proposition; it functions to highlight that p is particularly salient at the moment of utterance.  A third test frame for O-orientations is where S does not believe or know p. These are negative embedding contexts with a 1st person matrix subject, which co-index O   159 with S. jetz is felicitous in such a context. If O is coindexed with A (via a second person pronoun) or a third person (via a third person pronoun), jetz is equally felicitous (60a-c).  (60) Cx: There is a discussion among several people about an upcoming public   holiday, and the question arises whether people have to go to work and   university. Roman tells Tina that he definitely will have to go to work.   He then adds:    a.  Awa I woass imma  no ned…    but I  know  always   still NEG     …ob  jetz    morgen    Uni  is      whether  jetzt  tomorrow Uni  is       “But I still don’t know whether University is now open tomorrow.”    b.  Aba du woasst imma   no ned…    but  you know  always  still NEG        … ob jetz   morgen     Uni   is    …if   jetz   tomorrow  Uni  is.     “But you still don’t know whether University is now open tomorrow.”      c.  Aba ea woass imma    no ned…    but  he  know  always still NEG           …ob   jetz  morgen    Uni is         …if jetz  tomorrow Uni is.           “But he still doesn’t know whether University is now open tomorrow.”          (when talking about another person participating in the conversation)    In sum, jetz can occur in contexts where S has no commitment with respect to p, instantiated by V1 + \. It also can across a variety of shifting contexts. These are instantiated by form types, such as declaratives, imperatives, exclamations, and wh-interrogatives. Another shifting context was shown to be in embedded clauses.     160 4.5.2 Eh The test-frames for O-oriented DPRTs also identify eh as O-oriented. Specifically, I propose that eh expresses that O was committed to p before the utterance time as summarized in (61). This is consistent with its lexical origin presented in Chapter 3.   (61) eh ≈ O believed p before the time of utterance  The first test frame used to test for the O-orientation of eh is its shifting orientation:  it can occur both in interrogative questions (62) and declaratives used as assertions (i.e., with \) (63). (62) Cx: Hans, a drummer, is wondering how to get his drum kit to the location   of his next gig tomorrow, since he doesn’t have a car. He asks his    colleague Gerhardt:   Hans: Wia bring’e jetzt moang des Schlogzeig do hi...     “How am I gonna get the drum kit there tomorrow…”     …Du Gerhardt,  fahrst du moang       eh   in’d      Stodt?     …you Gerhardt  drive  you  tomorrow   eh   in.DET  city      “Gerhardt, are you still going to the city tomorrow…    ‘Gerhardt, [was it true before me asking] are you going to the city     tomorrow?      …dad’st ma’s do mitnehma?     “… could you take it for me?”  Eh, as mentioned, is also felicitous in declarative assertions, as evidenced below.   (63) Cx: Hans, a drummer, is wondering how to get his drum kit to the location   of his next gig tomorrow, since he doesn’t have a car. His colleague   Gerhard responds:  Hans:  Wia bring’e jetzt moang des Schlogzeig do hi?    “How am I gonna get the drum kit there tomorrow?”   161  Gerhardt: I fahr moang  eh in’d Stodt…    I drive tomorrow eh in.DET city     …do  konn’e’s da mitnehma.    …there can.I.it  you with.take     “I am going to the city tomorrow still. I can take it for you.”    ‘[I believed before saying this that] I am going to the city tomorrow. I    can take it for you.  Unlike jetz, eh cannot occur in imperatives, as shown below.  (64) a. ✓ Bring jetz des Schlogzeig naus!    bring jetz DET drum.kit out     “Now bring the drum kit out!”     ‘[It is salient now that I want you to] bring the drum kit out.’    b. * Bring eh des  Schlogzeig naus!   bring eh DET drum.kit out  I assume that this is due to the lexical contribution of eh. That is, if O has to believe p before the utterance time, then it follows that p must hold before the utterance time. This is incompatible with the meaning of an imperative. With the use of an imperatives S asks A to add the content of the utterance to their discourse requirements. To mark this with eh as something that was a requirement for A before S uttered the request would in essence presuppose that A had some mind-reading skills, and could have known about the content of the request before it was uttered.   Unlike jetz, eh is also incompatible with V1 + \.  There were no instances in my corpus, and attempts to elicit such forms resulted in ill-formed utterances, such as below. (65) * Geht eh  der  tatsächlich hoam \.      goes  eh  he   indeed       home       162 Since by uttering V1 + \ a speaker presents p, rather than asserting it (Reis 1999), it would be odd to mark it as something that was valid, true or relevant to a before uttering it. The same is true for the form V1 + √. This form commits S to p, and results in an exclamation. Exclamatives, which express surprise, are inherently temporally restricted. The content p of the host utterance is marked as entering S belief at the time of utterance. If S were to believe p before the utterance time, p would not be surprising and hence would not trigger an exclamative. eh, due to the past temporal restriction encoded in its core lexical entry, however, is incompatible with the immediacy of exclamations and exclamatives.  Next, I turn to contexts where S does not believe p. These are negative embedding contexts with a 1st person matrix subject. In these contexts O is in effect coindexed with S. Recall that ja was infelicitous in several such contexts because as an S-oriented DPRT, it expresses that S believes p. This is incompatible with those kinds of matrix verbs which, when negated don’t commit S to their complement. eh, unlike ja is felicitous under such a verb, as shown in (66).  (66) Cx: Dani asks Tina about whether Alex should receive a special    invitation to the Christmas party. She responds:     a. I woass ned ob’a      ned eh  kummt      I know  NEG whether.he   NEG   eh  comes     “I don’t know whether he’s not already planning on coming.”    b. *  I woass  ned ob’a       ned    ja  kummt    We expect that with 2nd or 3rd person subject pronouns, eh is equally felicitous, since it is not tied to a specific speech act role. Rather it shifts its ERP with the pronoun. This prediction is borne out in the data below. eh can occur when the matrix subject is   163 co-indexed with A (67).   (67) Cx:  Tina tells Dani that Alex doesn’t need a special invitation to the    Christmas party. He’s going to show up anyways.      Du woasst  doch  dass’a    eh  kummt      you know  doch  that.he    eh comes     “You know that he’s gonna come anyways.”    ‘[You believe that] you know that [it was the case before uttering this that]    he is coming.’   O-oriented DPRTs sometimes appear to involve S attitude because they are compatible with S-orientation. However I argue this is an artifact of the contexts. Each sentence has to have an utterer. This utterer is usually coindexed with the speaker S, and considered to be the attitude holder. But as we saw above, the attitude holder of the content of an utterance can be different from the utterer of the sentence; independence evidence for this will not be discussed here, but comes for example from the interpretation of the historical present (Banfield 1982, Schlenker 2004). For this reason a tri-partition that separates the direct discourse participants into utterers, hearers, and attitude holders (EPRs), i.e. into S, A, and O was proposed.67   4.6 Conclusion In this Chapter I showed the empirical need to differentiate between DPRTs which are sensitive to A belief and S belief. I established two different orientations for the DPRTs fei, doch and ja, S-orientation and A-orientation. First I showed that ƒei and                                                 67 Raffaella Zanuttini suggests that O-orientation may not be about anyone’s belief at all. This aspect warrants further in depth research, and is in line with the observation that eh and jetz express primarily temporal content. I suggest in Chapter 7 that DPRTs are used to relate p to an aspect of the contextual index. S- and A-oriented particles relate p to S and A respectively, whereas O-oriented DPRTs relate to the temporal and local aspects of the contextual index.   164 doch are in complementary distribution across most of the contexts shown. Both are sensitive to A belief about the propositional content expressed in the utterance. A-orientation, refers to S’s assessment of A’s belief about p. Even if S believes that A has no belief about p, S still evaluates A’s epistemic state. So the belief that p, belief that ¬p, and absence of belief about p are classified as A- orientation.   For ja it was shown that A belief is irrelevant. It is S belief that proved to be the determining factor for the felicity of the DPRT.   I also established a third group of DPRTs, the O-oriented DPRTs jetz and eh. These DPRTs can occur in contexts where either S has no commitment to p (such as in V1 +\), or in contexts where S doesn’t believe p. O-oriented DPRTs shift their orientation, and can occur in a variety of clause types. This compatibility with clause types was shown to be not only dependent on the orientation of DPRT and epistemicity expressed in the form of the clause, but also on the lexical content of the DPRT. In this way, one of the often-cited properties of DPRTs, namely their sentence type restriction, falls out as a corollary of independent properties of clauses and the DPRTs, and are not directly built into them, as selectional restrictions for example. The other often noted property, that DPRTs are expressions of speaker attitude was shown to be more fine grained, and testable taking into consideration the epistemic states of S and A.   The findings of this section are summarized in the table below.        165 Epistemicity Orientation DPRT Bel (A, p) A doch Bel (A¬p) A fei ¬Bel (A,p)/no belief about p A fei Bel (A,p) and A=S A doch Bel (S,p) S ja Bel (S¬p) S ja ¬Bel (S,p)/ no belief about p O jetz/eh Bel (O, p) O jetz/eh Table 16: Participant epistemicity and DPRT orientation   166 Chapter 5: Deriving the functional range of DPRTs  5.1 Introduction I showed in Chapter 4 that DPRTs can be divided into three classes, based on whose epistemic state they express. Within their broad classification as A-, S- and O-oriented DPRTs, however, DPRTs can be further distinguished. That is, DPRTs supply a wide range of interpretational nuances to their host utterance. I refer to this as their functional range. The purpose of this Chapter is to show that the functional range of three DPRTs can be derived from the core contribution of the UoL and the contribution of different contexts (cf. Abraham 1991), as schematized in (1).  (1) fDPRT1=  UoL + Cx1   fDPRT2=  UoL + Cx2   fDPRT3=  UoL + Cx3    … etc.  For example the DPRT doch is reported to express a contrastive function (e.g. Grosz, 2010b), as well as a reminding function (e.g. Hentschel 1986, Gast 2008) in addition to a shared knowledge function (Bárány 2009, Helbig 1988). Similarly, ja is also reported to have a shared knowledge function (Kratzer 1999, Zimmermann 2011), and thus seems to have a functional overlap with doch. Unlike doch, however, ja has a surprise function as well (cf. Helbig 1988). Consider the following data, which show that ja and doch are equally acceptable in the context shown below. Both (2a) and (2b) evoke a sense that the host proposition is in the Common Ground (CG).     167 (2)  Cx1: A family with two little twins is at the desk of a car rental place. They   chose a compact car for rental. The clerk says:    a. <✓ Cx1> Sie ham ja Zwilling…    you have ja twins     …woin’s do ned a gressa’s Auto nehma?   …want.you there NEG DET bigger  car take.     “You have ja twins, don’t you want to take a bigger car?”     b. <✓ Cx1> Sie ham doch Zwilling…   you have doch  twins    …woin’s do ned  a  gressa’s  Auto  nehma?  …want.you there NEG DET bigger   car  take.    “You have doch twins, don’t you want to take a bigger car?”    Now observe that ja and doch, despite this functional overlap in Cx1 in (2) behave differently from each other in another context, Cx2 (3); ja here is acceptable, whereas doch is not.   (3) Cx2: I am sitting in the coffee shop, working on my dissertation. I check the   time, and realize what time it is. I mutter to myself:   a. <✓Cx2> I   muass   ja in  5 Minuten in    da  Abteilung sei!    I   must     ja  in  5  minutes in DET  department be            “(Gosh), I gotta ja be in the department in 5 minutes!”       (Modeled after Kratzer and Matthewson 2009: ex 10)  b. <* Cx2> I   muass   doch  in  5 Minuten in    da  Abteilung sei!    I   must     doch  in  5  minutes in DET  department be     “(Gosh), I gotta doch be in the department in 5 minutes!”    168 This contrast between the function of ja in (2), which overlaps with the function of doch in that context, and the function of ja in (3), which doesn’t overlap with the function of doch in that context, raises two questions. First of, how can we account for this functional range, i.e. the different types of interpretations of ja and doch in the examples (2) and (3)? I show here that functional range can be derived from the context the DPRTs occur in. Second, what is the difference between ja and doch, if sometimes they seem to express very similar notions as seems to be the case in (2)? I show in this chapter that both are compatible with similar contexts, and therefore appear to be similar to each other.  Both functions of ja, namely shared knowledge and surprise, are difficult to reconcile under the assumption that only one DPRT ja exists. We could, however, adopt a meaning maximalist approach, which posits a variety of DPRTs, each of which is lexically specified differently (e.g. Franck 1980, Helbig 1988, cf. Abraham 1991: Introduction, cf. Zimmermann 2011 for more references). Alternatively, we could adopt a meaning minimalist approach, as argued in Abraham (1991). He specifically argues “…it is implausible to assume that our memory will not make use of derivational processes of a general sort to relate the obvious meaning correspondences between the respective words, instead of listing them under different, unrelated entries. What this boils down to methodogically is strict observance of Ockam’s razor principle. […assume] no distinct lexematic listing unless it can be shown that no derivational reconstruction can be invoked.” (Abraham 1991: 208).  I side here with Abraham’s approach and propose a meaning minimalist way of accounting for the functional range of doch, ja and fei. In particular, I propose that each DPRT has a core DPRT function. The resulting variation in interpretation,   169 surfacing as functional range of a DPRT, is conditioned by context and the Cooperative Principle (Grice 1975). This will be shown with the three case studies in 5.5 for ja, 5.6 for doch, and 5.7 for fei.  I show now that the main function of DPRTs is to mark a divergence from the normal course of a conversation. The DPRTs I discuss here, as argued in Chapter 3, have in common that they mark discourse participant epistemicity. Epistemicity is not usually marked beyond what is associated with the speech act (SA); at the level of the SA, i.e. the output of a form and its associated illocutionary force, epistemicity is not overtly marked, but encoded in the given SA as commitments. These commitments are usually described by way of the constitutive rules associated with that SA (Searle 1969). Within a discourse, following these conditions then leads to an adherence to the normal course of a conversation (cf. Zeevat 2003). I show the constitutive rules for the SAs I discuss here in section 5.4, where I also introduce epistemicity matrices as a way to model epistemic states in the course of the grounding process. I introduced grounding in 3.3.1, but will discuss it in relation to the model introduced in this Chapter in section 5.3.   One way (of the many) to depart from the normal course of a conversation is to mark the divergence of an interlocutor’s epistemic states from the default, i.e. marking who is or is not committed to the content of the host utterance, and who does or does not believe the content of the host utterance at the time of utterance (or some other time). DPRTs are one tool to mark this divergence, and thus they modify speech acts (Zeevat 2003, cf. Rett 2013). DPRTs mark epistemic states, as I showed in detail in the previous Chapter. These beliefs indicated by DPRTs via epistemicity, can diverge from the belief state introduced by each speech act via commitments resulting in illocutionary force, which in turn is conventionally associated with their syntactic   170 form. This is how DPRTs indicate a change from the normal course of the conversation (cf. Zeevat 2003, cf. Zimmermann 2004, 2011).  Depending on the context, marking a divergence from default epistemicity also may lead to certain inference processes on the part of A. Therefore I will show in section 5.4 that certain inferences and discourse relations between clauses and context hold that are independent of DPRTs, before I turn to three DPRTs in more detail, and show in case studies how their functional range derives from inferences, how it changes in context, and the type of speech act they occur in.  5.2 Formalizing grounding In Chapter 3, specifically in 3.3.1, I adopt a discourse model that separates the CG further into discourse components that refer to the beliefs of S, as opposed to the beliefs of A (cf. Farkas and Bruce 2010, Malamud and Stephenson 2015, a.o.).  The discourse component tasked with storing public S beliefs will be referred to as speaker ground (GroundS). Public S beliefs are those beliefs that S has committed to publicly. The respective component for A belief is the addressee ground (GroundA) (Lam 2014, Thoma 2014, Heim et al. 2016, Wiltschko and Heim 2016). Both GroundS and GroundA can contain the same elements as the CG (introduced in 3.3.1). The CG here can be conceived of as the intersection of GroundS and GroundA. Note that each discourse participants’ Ground can also contain beliefs about the epistemicity of the interlocutor.   I implement the separation of the epistemic states of each discourse participant here by following Heim et al (2016) and Wiltschko and Heim (2016): S’s belief about p will be represented as Bel (S, p). A’s belief about p will be represented as Bel (A,   171 p).   Figure 9 below shows a multi-step process for grounding an assertion with the propositional content p. It is to be read beginning at the top left, representing GroundS, proceeding toward the right in the top row, then dropping one row lower, then moving from right to left, as indicated by the direction of the arrows.  The bottom row, the CG column, is the end result of this specific grounding process, which had as a premise a positive response from A.   The way this figure is read is as follows:  S presents an utterance with content p in the presentation phase (see 3.1.1). p then is on the conversational Table, as well as the fact that S believes p, i.e. Bel (S, p). By default, the fact that S has uttered something is in GroundA (i.e. latest move). In the acceptance phase, under the assumption that A makes a positive step toward integrating p into her beliefs, A’s acceptance of p is on the Table. From there S can add the fact that A believes p into her set of beliefs. Only at the end of this grounding process, p, the fact that S believes p, and the fact A believes p is CG between the interlocutors.    GroundS Table GroundA Presentation phase Bel p              à p, Bel (S, p)    à Bel (S, p) Acceptance phase Bel (A, p)      ß Bel (A, p)       ß Bel (p) Common Ground                                                           ⇓  accept p, Bel (S,p); Bel (A,p) Figure 9: Grounding an assertion  Not every SA commits the same discourse participants to the content of the utterance. In addition, the timing of when a belief holds may vary. I will therefore make use of a more fine-grained model of showing the belief states of each DPRT in a given SA via epistemicity matrices. The two phases of grounding, presentation and acceptance   172 phase, can be mapped onto different times. The former maps onto the time of utterance tU, whereas the latter maps onto a time after the utterance time. However, we also need to consider participant epistemicity before the time of utterance (t<U ),  since DPRTs are sensitive to it as well. Therefore, an epistemicity matrix for an assertion looks as in table 17.   t<U tU S A S A assertion Bel p  Bel p  Table 17: Epistemicity matrix for assertion  The content of the matrices I will use for the analysis is supplied by the constitutive rules, in particular, by the preparatory rules for the given SA. These will be introduced in the next section, as well as how to understand the normal course of a conversation and deviation from it.   5.3 The normal course of a conversation I showed in Chapter 3 that each clause type is linked to a participant’s discourse commitment and by proxy, to their belief (cf. Rett 2013). I did this by showing the particular structural and morphological characteristics that constitute a specific clause type.  As we saw, modifying Cx: form via intonation can lead to a change in the commitments (beliefs) of the discourse participant (S or A), and can therefore lead to a change in the associated illocutionary force. Hence a modification in form can lead to a modification of the speech act. I therefore consider speech acts to be derived from   173 the form (cf. Beyssade and Marandin 2006). While there is a default relation between the form of an utterance and its function (and by proxy, the commitments expressed), a change in form may correlate with a change in this function.   Rett (2013:20) argues that to add an attitude marker to an utterance with default function changes it and in particular, it changes or adds to the preparatory and sincerity conditions that are typically associated with the SA. Whereas Rett considers intonation and attitude markers like alas in her account, I propose that her insight also captures the contribution of DPRTs. Specifically, S and A’s epistemic states can be overtly marked by DPRTs, if they diverge from the normal course. It is precisely in that sense that DPRTs are markers of divergence from the normal course of a conversation (non-standard utterances in Zeevat’s (2003) sense). This is one of the great insights of Jacobs (1986), who analysed DPRTs as modifiers of illocutionary types. Whereas the analysis presented here differs from Jacobs’, I nevertheless agree with his assessment about the core purpose of DPRTs.   (4)  Wenn Verbstellung, Verbmodus, Intonation etc. einen bestimmten  Illokutionstyp X festlegen, so wird daraus durch Hinzunahme eines  Abtönungsmittels ein anderer Illokutionstyp X’, der […] in ihren  Anwendungsbedingungen eingeschränktere Version von X ist.  (Jacobs 1986:103)   When verb position, verb mood, intonation etc. determine a specific illocution  type X, it turns, via the addition of a ‘toner’ (DPRT- added by ST) to another  illocution type X’. This type is more restricted in its conditions of use than X  is. (translated by ST)  Crucially, Jacobs (1986) invokes use-conditions in his characterization of the function of DPRTs; the felicitous use of DPRTs is not determined by considerations of truth-conditions, but subject to specific contextual restrictions regarding the epistemic states of S and A. This is precisely the finding presented in the previous chapter,   174 where I showed that each DPRT is sensitive to a specific discourse participant’s epistemicity.  Searle (1969), following Austin (1962) identifies constitutive rules for SAs, which establish the normal course of an utterance/conversation. He specifies four types of constitutive rules for utterances (repeated from 3.4). Recall that I disregard the essential rules for the purpose of the current analysis.   (5) Constitutive rule for speech acts  (i) content rules   (ii) preparatory rules  (iii) sincerity rules   (iv) essential rules  The content rules refer to the locutionary act; the actual utterance, its propositional content, and its meaning, including the phonetic realization, syntactic form, and semantics of the lexical items involved in the utterance. The preparatory rules are most relevant for the analysis of DPRTs presented here. I will use these to indicate the epistemicity of the discourse participants in the normal course, and use them as a base for the content of the epistemicity matrices.  These serve as the base to indicate a deviation from the normal course, with DPRTs as indicators of this deviation (cf. Egg 2010). I will introduce the epistemicity matrix after presenting the constitutive rules.  The sincerity rules concern the situation in which a given speech act is uttered, i.e. they refer to the contextual preconditions for a given speech act.   DPRTs serve to amend the discourse commitments of the SA participants. Thus, DPRTs modify the SA by amending aspects of the preparatory rules associated with   175 them (Egg 2010; Rett 2013, 2014; Zeevat 2003).68 To understand how exactly DPRTs modify SAs by amending the discourse commitments and thereby beliefs of the interlocutors, and thus signaling a departure from the normal course, I introduce the conditions for SAs leading to the normal course next.  Assertion: A standard assertion is conventionally expressed with the clause type declarative (but see the discussion in Chapter 3 for clause type and illocution force mismatches). The constitutive rules (minus the essential rules) for uttering an assertion are as follows: In order for S to felicitously utter an assertion with content p, she has to believe p (i.e. Bel (S, p)). It is also not obvious to her that A believes p. S wants A to accept p, that is, S wants to add Bel (A, p) to her set of beliefs (Rett 2011, cf. Searle 1969).  (6)  Constitutive rules for assertions    content rule:  the content of an assertion is any proposition p    preparatory rules:  a. S has evidence that p is true b. it is not obvious that to both S and A that A believes p   sincerity rule: S believes p  Question: A standard question is typically expressed with the clause type interrogative. We have seen two types of interrogatives, polar interrogative and wh-interrogative. The constitutive rules for both are similar, but differ in some crucial aspects: the content rule and the sincerity rule. (7) presents the rules for polar                                                 68 Rett (2013, 2014) sees intonation as a morpheme, too, (a UoL in Wiltschko’s 2016 terminology) which also can serve as a speech act modifier.   176 questions, typically expressed with V1 clauses with final rise. (8) presents the rules for wh-questions, typically expressed with V2 wh-interrogatives.  (7)  Constitutive rules for polar interrogatives   content rule:  the content of a polar interrogative is a set of two propositions    {p, ¬p}  preparatory rules:   a. S does not know whether p is true or not  b. S believes that A knows whether p is true or not  c. It is not obvious to both S and A that A will provide an answer  without being asked   sincerity rule: S wants to know whether p or ¬p   (8) Constitutive rules for wh-interrogatives content rule:  the content of a wh-interrogative is a set of propositions with an     open variable represented by the wh-word69 preparatory rules:   a. S does not know the value for the open variable  b. S believes A knows the value for the open variable  c. It is not obvious to both S and A that A will provide an answer  without being asked sincerity rule: S wants to know the value for the variable represented by the wh-word   Request: A standard request is typically expressed with the clause type imperative. Its semantic content is a property, which S wants A to fulfill  (in Portner’s 2004 terms, imperatives represent actions A should take).  (9)  Constitutive rules for requests  content rule:  the content of a request is a property S wants A to have  preparatory rules:   a. S believes A has the ability to fulfill the request                                                  69 I will continue to refer to the content of any of these clause types as propositions, or p, for exposition purposes.     177  b. It is not obvious to both S and A that A will take the requested   action without the request  sincerity rule: S wants A to fulfill the request  Exclamation:   Unlike other speech acts, exclamations do not have a ‘typical’ clause type associated with them;70 they can be expressed with V2 structures, V1 structures, bare DPs, or bare exclamative words such as mei (‘my’). The exclamative speech act expressed in sentence exclamations, however, has the following associated constitutive rules, as established by Rett (2011).  (10) Constitutive rules for sentence exclamations   content rule:  the content of an exclamation is the proposition p denoted    by that sentence preparatory rule: S has direct evidence that p is true. sincerity rule: S is surprised by p  	Adhering to these constitutive rules presents the normal course of a conversation, however, following the constitutive rules alone does not lead to coherent discourse. If S deviates from this normal course, A is led to believe that there is a reason for this deviation, due to the cooperative principle (Grice 1975). This leads discourse participants to use independent pragmatic inference processes and deductive reasoning to establish coherence. This will be presented next, before turning to the three case studies.                                                 70 There is crucial distinction between exclamations and exclamatives. MB, like German has a dedicated exclamative in the form of a wh-clause, with the verb in its base position as in (i).   (i)  Wos’a  eam  oiss  vasprocha  hod! what.he  him  all  promised  has “Oh the things he promised him!”  See however Brandner (2010) for a different approach, and references therein for support of the distinction made here (cf. Rett 2011).   178 5.4 DPRTs, inferences and discourse coherence  Utterances, whether they contain DPRTs or not, don’t occur in a context-less vacuum. In order to assign the appropriate meaning to an utterance, it is crucial to take into consideration its context of utterance. Consider the following example, which illustrates a typical case of pragmatic inferencing, showing how discourse relations, albeit not overtly expressed, are necessarily part of the interpretation of an utterance in context.   (11) Cx:  Martl asks his friends about an upcoming party.     Martl:  Wea kummt’n olla zum Alex seina Feia?    “Who all is coming to Alex’ party?”    Tina: D’Sonja konn ned kemma.    DET.Sonja  can NEG come     De muass se um  ihre Zwilling kimman.    she has self about  her twins   care      “Sonja can’t come- because she has to care for her twins.”    ‘Sonja can’t come- she has to care for her twins.”  Observe the causal relation connecting the two utterances. The fact that she has to care for her twins is interpreted here as the reason why Sonja cannot come to Alex’ party despite the fact that this causal relation is not overtly expressed (via a conjunction for example).  Pragmatic inferencing is responsible for the fact that the meaning of an utterance is enriched by additional information that is not directly encoded in the linguistic signal. Inferencing is one way in which discourse coherence is established. Coherence, according to Kehler (2011:19) “is defined in terms of the underlying semantic relationships that characterize and structure the transitions between utterances.” It   179 determines implicit relationships between clauses such as a causal or a temporal connection, by connecting the ‘what is said’ to the ‘what is meant’ as was shown in (11). Formally represented linguistic meaning, that is, meaning represented by truth-conditional, compositional semantics, must be combined with conceptual knowledge as part of broader human cognition (Irmer 2009, Kehler 2011). Crucially, discourse coherence can be guided and facilitated by the use of lexical items, but it is not exclusively dependent on it (Kehler 2011). Irmer (2009:38 emphasis ST) argues that "coherence is established by virtue of rhetorical relations which mark a relationship between chunks of texts. Rhetorical relations can be expressed by cohesive means such as discourse connectors (e.g. "and", "but"), but in many cases they are not explicitly marked".   Kehler (2011) advocates for a classification of coherence relations based on three types of primary, basic connections among ideas, taking inspiration from David Hume’s (1749/1955) Inquiry concerning human understanding. These three relations are (i) cause/effect, (ii) resemblance, and (iii) contiguity.  Discourse cohesion, based on these three basic relations, obtains whether these relations are overtly marked or not. Crucially, coherence obtains whether overt cohesive markers (such as connectives, anaphors, etc.) are present or not. Coherence in fact, does not obtain due to the use of these cohesive markers, but rather, coherence has to be established as a prerequisite for the felicitous use of cohesive markers (Kehler 2011). Thus, cohesion is based on contextual, extra-linguistic knowledge, that is, a matter of cognitive skills. And this is precisely where pragmatic inferences come into play.    One framework, under which ‘what is said’, and ‘what is meant’ has been captured, is Grice’ s (1975:26f.) proposal of the maxims of conversation and the   180 overarching Cooperative Principle.  (12)  Cooperative Principle: Contribute what is required by the accepted purpose of          the conversation.    The Gricean model thus presupposes that two discourse participants are rational and cooperative, and make maximally relevant contributions to the CG.  The DPRTs under consideration here can only be adequately captured when considering the interplay between pragmatic inferences based on establishing discourse coherence, the cooperative principle, and the contribution of the individual DPRT. That is, context Cx plays a crucial role in establishing the function f of DPRTs. This is again captured by the blueprint used throughout this dissertation, and repeated in (13).    (13)   fDPRT= UoL +Cx  Context Cx here includes the form of the utterance, and the associated discourse relations and inferences that hold, regardless of the contribution of the UoL itself (that is, the DPRT).   I now turn to the DPRTs, and show that it is not that the DPRTs themselves vary in their core function, but that the contribution of the context, including the inferences associated with it, derives the fine-grained differences in function; contextual variation creates the functional range of DPRTs.     181 5.5 The functional range of ja Even in its use as a DPRT, ja is multifunctional. The following functions have been identified for SG ja: (i) referring to shared knowledge, (ii) expressing surprise, (iii) expressing emphasis, (iv) indicating a causal relation between an antecedent utterance and the host utterance for ja (Helbig 1988). This is summarized below:  (14) DPRT functions of ja DPRT  Function DPRTja f1: shared knowledge f2: surprise f3: emphasis f4: reason  Recall from 3.2 that a proposition (i.e. the content of an utterance) does not enter the CG by virtue of being uttered. Instead it needs to be grounded in a two-step process, exemplified with an assertion in Figure 10 repeated from 5.3.    GroundS Table GroundA Presentation phase Bel p        à p, Bel (S, p)   à Bel (S, p) Acceptance phase Bel (A, p)      ß Bel (A, p)      ß Bel (p) Common Ground                                                           ⇓  accept p, Bel (S,p); Bel (A,p) Figure 10: Grounding an assertion  For an assertion, this means that p is grounded, and becomes CG only when A does not disagree, or when A shows signs of agreement. Alternatively, A might disagree with S, and as a consequence A will not add Bel (p) to GroundA. As a result, S might also remove Bel (p) from GroundS, in particular if A presents convincing evidence against Bel (p). In that case p is not CG, and may also not be in GroundS anymore.   182 I propose that ja indicates that S will not remove p from GroundS. As illustrated in (15), when uttering [ja (p)], S places p, and Bel (S, p) on the conversational table. With the use of ja S indicates that she does not consider changing her mind with regard to p now or during the course of the conversation.   (15) ja(p) ≈ Bel (S,p) (tU)  and  ∀ t >tU :  Bel (S,p) (t)  This proposal strengthens and elaborates on the previous proposal made in 3.2.3. ja, in essence, marks that the acceptance phase of the grounding process is skipped; S believes p, whether A accepts p or not. This is how the overarching ‘backgrounding’ effect of ja can be understood: if S indicates that she believes, and will continue to believe p, whether A accepts p as their belief or not. p is not up for discussion, according to S. This is illustrated below:   GroundS Table GroundA Presentation phase Bel p   à p, Bel (S,p)   à Bel (S,p) Acceptance phase ⇓   Common Ground Bel (S,p) Figure 11: Grounding with ja  In this way, the core contribution of the DPRT ja is something like I firmly believe, as informally proposed in 3.2.3. This is of course consistent with its use as an affirmative response particle, where it serves to value the polarity of the proposition positively. Given the constitutive rules associated with positive assertions, S will be assumed to believe p.    In the next sections I show that there are several contexts that are compatible with the core contribution of ja (15). Depending on the particular context, however the   183 contribution of ja has a different effect. This allows us to see the functional range of ja.  5.5.1  Shared knowledge According to several analyses, with the use of ja S refers to knowledge that is mutually shared between the speech act participants. Some accounts argue for this explicitly, such as Kaufmann (2004) for example, who claims that ja contributes a presupposition that relates to the mutual knowledge shared between S and A, i.e. to CG (also e.g. Kaufmann and Kaufmann 2012, Kratzer 1999, 2004, Thurmair 1989, Weydt, 2006). Others take the CG interpretation to be more of an implicit contribution of the DPRT, such as Zimmermann (2011), according to whom ja "indicates that the speaker takes the hearer to be aware of [p]”.  I argue that this CG effect is just that; an effect. I claim that it is inferential, and can be deduced from the core meaning of the DPRT in combination with its context of use. This proposal is consistent with evidence discussed in Kratzer and Matthewson (2009), who show that SG ja is often used in contexts which report obvious, readily observable facts, or in contexts that entail facts that have already become CG. In other words, ja is used in contexts in which it is appropriate for S to say that she is not going to remove p from GroundS without awaiting A’s acceptance. This can be seen as the often-cited uncontroversiality interpretation of ja referred to in Lindner (1991) and used in Grosz (2010b), and by others. Something that is uncontroversial then can be assumed to be CG.  Crucially, the context can contribute a shared knowledge inference, i.e. add that the proposition is CG at the time of utterance, independently from the contribution of ja.   184 This is schematized in the epistemicity matrix for shared knowledge in Figure 13 below; this matrix represents the speech act participants’ belief regarding p at the time of utterance tU and at a time just before the utterance t<U. Shared knowledge presupposes that A believes p and S believes p at the time of utterance tU. It implies that if A believes p at tU, A also believed p just before tU, i.e. at t<U. Shared knowledge diverges from a normal assertion in that here A does not believe p at the time of utterance or before. In the epistemicity matricx presented in table 18, empty cells under A indicate that S may not know anything about A’s epistemic state regarding p. Empty cells under S indicate that S doesn not have a commitment to p.    t<U tU S A S A assertion	 Bel (p)	 	 Bel (p)	 	shared knowledge Bel (p) Bel (p) Bel (p) Bel (p) surprise   Bel (p)  Table 18: Epistemicity matrix for shared knowledge  Witness the following context, which instantiates this. The discourse situation Cx: disc establishes that the family has twins, i.e. that it is CG that p.  (16) Cx1: A family with two little twins is at the desk of a car rental place. They   chose a compact car for rental. The clerk says:     a. Sie ham ja Zwilling…   you have ja twins     …woin’s do ned a gressa’s Auto nehma? …want.you there NEG DET bigger  car take     185 “Since you have twins, don’t you want to take a bigger car?” ‘[I firmly believe that] you have twins, don’t you want to take to a bigger car?’   b. Sie ham Zwilling…  you have twins   …woin’s do ned a gressa’s Auto nehma? …want.you there NEG DET bigger  car take.  “You have twins. Don’t you want to take a bigger car?”  The immediate discourse context leads to the possible assumption that the speech act participants both share as relevant CG knowledge that this family has twins. This assumption is independent of the use of ja. This situation is compatible with the use of ja in the clerk’s utterance as in (16a). However, the same situation also allows for an utterance without ja, as in (16b) where p (Sie ham Zwilling) is still taken as shared between S and A.   Consider the next example, which also shows that shared knowledge is not a necessary condition for the use of ja. The host utterance is uttered in a context where S infers from what she is told that A doesn’t believe p.  (17) Cx: I tell my mom that I took out a 50,000 Euro student loan. She says:    Du host ja an Schlog.   you have ja DET hit    “You’re totally crazy!”    ‘[I firmly believe that] you’re crazy!’   It can be safely assumed that the addressee doesn’t share the speaker’s sentiment that she is crazy. Again, A in this context, according to S, doesn’t believe p; the addition of ja expresses that S, however, firmly believes p and will not remove it from her GroundS.   186  This establishes that shared knowledge is not a sufficient condition for the use of ja. Moreover, shared knowledge is also not a necessary condition for the felicitous use of ja. In particular it can be used if S is surprised about p and A is not (yet) aware of p. This is discussed in the next subsection.   5.5.2  Surprise Another common context for the use of ja is in utterances expressing surprise. However, surprise is neither a necessary nor a sufficient condition for the felicitous use of ja. Hence, surprise cannot be directly encoded in the lexical entry of this DPRT (cf. Kratzer, 2004). Rather, the element of surprise falls out from the discourse context and the form of the utterance. In what follows, I discuss the use of ja in exclamations. Surprise arises when S gets to believe p at the time of utterance and when p is a new (relatively unlikely) belief. In other words, S did not believe p prior to the time of utterance (<tU ). As discussed in section 4.3, A’s belief is irrelevant for the felicitous use of ja. This yields the epistemicity matrix for surprise summarized in Table 19.    t<U tU S A S A surprise   Bel p  assertion Bel p  Bel p  Table 19: Epistemicity matrix for surprise  In this section I present data that have been widely used for various analyses of SG ja that associate a surprise component with this DPRT. What has been overlooked, however, is that each of the examples used to show the surprise effect instantiate an   187 exclamation context. Consider for instance, the widely cited example from Lindner (1991), rendered here the way she does.   (18) Mother is looking out of the window    M: Da  ist  ja ein Zeppelin!   there  is  ja DET zeppelin   “There is a Zeppelin!”    Child comes running to the window:  Oh, ist der groß!          “Wow, that is big!”        (Lindner 1991: ex 12. p171)  Note that an extra high pitch accent on Zeppelin is needed for the well-formedness of this example. Whereas the clause type of the utterance is a declarative, the extra high pitch accent √ triggers exclamation force (see section 3.4). In this example the mother, by looking out the window, has direct visual evidence for the existence of the Zeppelin (blimp), which is a necessary condition for the exclamation speech act (5.3). Blimps in the sky can be taken to be a rather unusual sight. Note that the surprise context is not a sufficient condition for the use of ja. Instead, the sentence can also be felicitously used as a surprise utterance without ja as shown in (19).   (19) Cx1 : Mother is looking out of the window  M:  Do  is a ZEPelin √!   there  is DET zeppelin   “There is a Zeppelin!”   Since the surprise reading comes about with and without ja, ja is merely compatible   188 with the context; it does not in itself contribute a surprise meaning.  Kratzer (2004) proposes that the content of the host utterance for ja is verifiable on the spot. Recall that the preparatory rules for exclamations include that S has to have direct evidence that p is true. I now show that utterances that match Kratzer’s (2004) condition are exclamations; therefore the meaning component, ‘verifiable on the spot’ is contributed by the preparatory rules for uttering an exclamation, which state that S has direct evidence that p is true. This derives the surprise reading of ja.  Another well-known example from the literature provides further support for the claim that surprise is contributed by context, not by ja.  (20) Spencer is walking up the stairs in front of Webster.   Webster:  Du hast ja ‘n Loch im Ärmel.    you have ja DET hole in.DET sleeve     “There is a hole in your sleeve.”     Kratzer (2004:126) cites this example from Lindner (1991) to argue for the idea that the ‘facts’ expressed in utterances with ja either have to be shared knowledge or have to be verifiable on the spot. She doesn’t, however, provide the complete discourse context given in the original example in Linder (1991). This missing context (given in (21), provides crucial clues about the epistemic state of A (Spencer in this example). It shows that A, Spencer, did NOT know about the content of the proposition expressed by the utterance. This is an additional piece of evidence that ja does not reference mutually shared knowledge, that is, shared knowledge is not a necessary (or sufficient) condition for the use of ja. This is consistent with the analysis developed here, according to which the use of ja is independent of the epistemic state of A.     189 (21) S is climbing the stairs in front of W  W: Du hast ja  ' n Loch im Ärmel!    you have ja DET hole in.DET sleeve     “You've got a hole in your sleeve, you know. “71    S:  Wo?      “where? “(looks at sleeve)     W: Da!     “There!” (Points at sleeve)     A: Ach, dann krempel ich ihn eben rauf.     “Oh well, I'll just roll it up then.”           (Lindner 1991 ex:13)   Spencer, the addressee, clearly does not know that he has a hole in his sleeve, as evidenced by the discourse (he is asking where, then responds with ach “Alas/oh well”, which in turn indicates that p is new)72. Again, the surprise effect comes from the host utterance, not from ja; it is merely compatible with the context, but does not contribute a surprise component. This is supported by the fact that Lindner’s original example is marked as an exclamation by means of an exclamation mark “!” (which is missing in Kratzer’s  rendition of the example).   Ja is felicitous in this context, since S has direct evidence that p, and p is verifiable on the spot, in Kratzer’s (2004) terms. Therefore S does not need A’s acceptance. This is consistent with the present analysis according to which ja indicates that S’s belief of p is not contingent on the acceptance of p by A.                                                    71 This is Lindner’s translation.     190 5.5.3  Emphasis  Another context of use in which ja is appropriate is if S is certain about p, but believes that A thinks otherwise.  If ja is used in this context, the effect is emphasis.  Recall that the assertion of [ja (p)] includes that S believes p, puts it on the Table, and won’t remove it from GroundS. This signals to A that the ja modified utterance is not up for discussion, no matter what A may believe. The addition of ja, which signals to A that S won’t remove Bel (p) from GroundS thus leads to the effect of emphasis on p. This is illustrated below:  inference/effect t<U tU S A S A emphasis Bel (p) and Bel (¬Bel (A, p)  Bel (p) and Bel (¬Bel (A, p)   assertion	 Bel (p)	 	 Bel (p)	 	Table 20: Epistemicity matrix for emphasis  The example below illustrates the use of emphasis with ja. The context provides information about the epistemic states of the interlocutors (kids who are accusing Alexander of having stolen a special paint). Alexander, using ja in the response shows that he won’t change his mind about p, i.e. that they won’t find anything.  (22) Cx: Alexander is being accused by his classmates of stealing a special kind   of paint. His classmates interrogate him, he denies. At the end, one of   them suggests to look into his bag. Alexander responds:    Von mir aus…   “If it's for me…     …schaugts hoid in mein Schuiranzn…   …look        hoid in my school.bag     … es  findts ja eh nix.   191   …you.2PL   find ja eh nothing   “Go ahead, look into my bag‚ you're not gonna find anything anyways.”  ‘Go ahead, look into my bag [I firmly believe that [it was the case before uttering  this that]] you're not gonna find anything.’   5.5.4  Reason In this section I show a fourth context of use compatible with the meaning of ja, which I call the reason context (cf. Helbig 1988). This particular interpretation of ja has, to my knowledge, not been discussed in the literature, but is a prevalent use of the DPRT. Reason interpretation for ja arises when two conditions are met: (i) S utters a host utterance with content pi, indicating Bel (S,pi); (ii) S uttered an utterance with content pj immediately preceding, indicating Bel (S, pj).  I showed in 5.4 that discourse coherence is established via certain inferences. Certain discourse relations, such as a causal relation leading to reason interpretation, can hold between clauses in a discourse independent of the use of DPRTs (Kehler 2011). Example (23) is repeated from above, and shows such a causal relation, where pi is interpreted as the reason for pj.   (23) pj: D’Sonja konn ned kemma.  DET.Sonja can NEG come.       pi  De muass’se um ihre Zwilling kümman  DET  has.self for her twins  care   “Sonja can’t come, since she has to care for her twins.”  ‘Sonja can’t come. She has to take care of her twins.’  The addition of ja leads to the following effect: since ja marks that S will not remove p (here pj) from GroundS, pi is ‘backgrounded’ in the discourse context; it is not up for discussion, according to S. By backgrounding pi, ja in effect foregrounds pj,   192 facilitating the reason interpretation. Crucially, ja itself doesn’t specifically encode reason function. This is formalized in the epistemicity matrix for reason below.    t<U tU S A S A reason Bel (pj) and Bel (pi)  Bel (pi)  assertion	 Bel (p)	 	 Bel (p)	 	Table 21: Epistemicity matrix for reason  Witness the following example to illustrate this, modified from Kratzer (1999).   (24) pi: D’Sonja konn ned kemma.     DET.Sonja can NEG come        pj  De muass’se ja um ihre Zwilling kümman.    she has.self ja for her twins  care    “Sonja can’t come. It’s because she has to care for her twins.”   ‘Sonja can’t come. [I firmly believe] that she has to care for her twins.’  The next example also illustrates a causal relation between two propositions, which holds with or without ja.   (25) p: D’Regina kummt heid zum Essn vorbei.     “Regina is coming to eat (with us) today.”     pj: Do mach’e a  Gmias.    there  make I  DET  vegetables    “So I’ll make veggies.”     a. pi:  Sie isst ja koa  Fleisch       she eats ja  NEG.DET meat    193    “It’s because she doesn’t eat meat.’  ‘[I firmly believe that] Regina doesn’t eat meat.’  b.   pi’: Sie ISST koa  Fleisch    she eats NEG.DET  meat.    (modeled after Thurmair 2013)   Again, a reason interpretation obtains with or without ja, as evidenced by (25a -b). Note however the obligatory special pitch accent on the finite verb, which leads to a special emphasis on the propositional content of the utterance (verum focus, Höhle 1992). Without this verum accent the utterance would not be acceptable in this context.  I will discuss this need for either the DPRT or the special accent in Chapter 6.   The utterance with ja facilitates the ‘reason’ inference in the following way; with S marking that she won’t remove pi from GroundS, a rational, cooperative interlocutor A, will attempt to make S’s contribution maximally relevant to the discourse, and relate it to pj. By marking pi as ‘not up for discussion’ with ja, S conveys to A that the fact that Regina doesn’t eat meat is not something she wishes to discuss, i.e. it is backgrounded. A is thereby discouraged to challenge pi. Since pi is not up for challenge, the inference about the coherence relation between the two clauses (here reason) is facilitated (cf. Kehler 2011).   5.5.5  Summary In this section, I argued that the functional range of the DPRT ja is an artifact of the core function of ja and the contribution of the context in which it is used.  (26) fDPRT,ja = ja + Cx   194  In particular, I introduced epistemicity matrices for the contexts that are compatible with the core function of DPRT ja, as summarized below  (27) ja (p) = Bel (S,p) (tu)  and  ∀ t >tU : Bel (S,p) (t)    t<U tU S A S A shared knowledge Bel p Bel p Bel p Bel p surprise   Bel p  emphasis Bel p and Bel(¬Bel (A,p)  Bel p and Bel(¬Bel (A,p)  reason Bel pj and Bel pi  Bel pi  assertion	 Bel p	 	 Bel p	 	Table 22:The functions of ja  What this epistemicity matrix reveals is that what all of these contexts have in common is that S believes p at the time of utterance. And I argue that this is the only core meaning associated with ja, as shown in Chapter 4. The epistemicity matrix of each host context for ja varies, however, and also contributes a meaning; this is how the functional range of ja can be understood.    5.6 The functional range of doch Like ja, the DPRT doch is multi-functional. Each of its functions depends on the context of use. The functional range of doch includes (i) expressing shared knowledge (Grosz 2010, 2014) (ii) reminding A of p, (iii) backchecking, and (iv) indicating a contrast (cf. Gast 2008, Hentschel 1986). This is summarized below:     195 (28) DPRT functions of doch DPRT Function DPRTdoch f1: shared knowledge f2: reminding f3: backchecking f4: contrast  Based on its diachronic origin as an A-oriented deictic marker (in Hentschel 1986 calls it ‘emphatic affirmative marker’, cf. Karagjosova and Zeevat 2007), I suggested in Chapter 3 that doch can be informally paraphrased with I believe that you believe p. This amounts to saying that doch marks that S believes p is in GroundA at the time of utterance, as indicated in (29).  (29) doch (p) ≈  Bel (A,p) (tU)   With the use of doch, S marks that A knows p (or at least could reasonably know that p) at the time of utterance. Hence A need not accept p and therefore the acceptance phase can be skipped, just like it was the case with ja. It is in this sense that doch marks the utterance as deviating from the normal course of the conversation. Figure 12 illustrates the grounding process with doch.   Ground S Table Ground A Presentation phase Bel p Bel (A, p) à p, Bel (S,p) p, Bel (A,p)  à Bel (S,p) Bel (S,(Bel (A,p)) Acceptance phase    Common Ground Bel (S,p) Figure 12: Grounding with doch  Several of the functions of doch have been widely discussed in the literature, first and   196 foremost the contrastive function (e.g. Hentschel 1986, Thurmair 1989, Abraham 1991, Lindner 1991, Bárány 2009, Grosz 2010b, Egg and Zimmermann 2011). The uses presented in (28) have not, however, been unified in previous analyses. In what follows I show that the functional range of doch can be derived from the core contribution of doch in interaction with its context of use.   5.6.1  Contrast A common use of doch is in contrastive contexts. As mentioned in 3.5, contrast is commonly argued to be the core function of doch, and assumed to be part of its lexical entry (Hentschel 1986, Thurmair 1989, Abraham 1991, Lindner 1991, Bárány 2009, Egg and Zimmermann 2011, Grosz 2010b). Here I argue that the contrastive function can be derived as a pragmatic inference based on the context of use of doch. Hence we do not need to postulate it as an intrinsic part of the meaning of the DPRT itself (Krifka 2013). In essence, I argue that doch is compatible with a contrastive context, but does not itself encode contrast. Rather contrast comes about because with the use of doch, S says that A knows p. So if uttered in a context where A states that s/he doesn’t know p, this derives a contrast. The relevant epistemicity matrix for the contrastive interpretation is given below.   t<U tU S A S A contrast Bel¬p  Bel p Bel ¬p Bel p assertion Bel p  Bel p  Table 23: Epistemicity matrix for contrast  Grosz (2010:3) presents the following example to support the contrastive hypothesis.    197   (30) Mary: Schau mal! Diese Blumen  sind so hässlich.  look PRT these flowers are so ugly   “Have a look! These flowers are so ugly.”  Bill:  Was hast  du  denn?  Diese  Blumen sind doch schön!  what have you PRT   these   flowers are  doch  beautiful  “What's your problem? These flowers are [DOCH] beautiful!”           ⇒ [p The flowers are beautiful] is used to correct [q the flowers are ugly].           (Grosz 2010: ex 3)   Crucially, the contrastive interpretation in (30) arises even without the DPRT as shown based on the MB examples in (31).73   (31) Maria:  Schaug amoi! De Bleame do san vielleicht  greislig  look  PRT DET flowers there are vielleicht 74 ugly   “Have a look! These flowers are so ugly.”  Willi:   a. Wos host’n? De  san SO schee!  what have.you PRT. they are  so beautiful   “What's your problem? These flowers are so beautiful!”  b.  Wos host’n?  De san doch schee!  what have.you. PRT.  they  are doch beautiful   “What's your problem? These flowers are beautiful!”   Hence we have to conclude that contrast is not a sufficient condition for the use of doch; it is, however, a context compatible with the function of doch. Contrast is the                                                 73 A retort without doch would need the particle so, and an emphatic accent on so, in order to render an acceptable discourse respronse in this context, or, alternatively, an extra high pitch accent on schee.  74 Vielleicht here is used as a DPRT, strengthening the exclamation. The original Grosz example with the intensifier ‘so’ strikes me and other speakers of Bavarian as slightly odd. This may be due to the fact that it is an example from a differnt variety.     198 result of pragmatic inference based on the antonymy of ‘ugly’ and ‘beautiful’.   The crucial point of this example is that the contrastive interpretation is not dependent on doch but derives from its context of use. While the retort without doch is possible, the utterance with doch is preferable. This is expected, under the assumption that in this specific context, S wants to convince A that indeed, ‘p’. S does so by marking Bel (p) as part of GroundA. This means S bypasses the acceptance phase of grounding, in essence preventing A from not accepting p. The contrasting interpretation in the example above then arises due to the need of the interlocutors to establish discourse coherence, and to interpret each contribution as maximally relevant in the given discourse.   The next example further supports the idea that contrast can be inferred from context, which is compatible with doch.   (32) Cx: Kathrin and Hansi are cousins. Kathrin knows where Hansi has been   for vacation ever since they were little. They chat about where they   each might want to go next.    Hansi: I war no nia am Mea    I was still never at.DET ocean     “I have never been to the ocean.”   Kathrin: a. Du warst doch letzt's Jahr in Kroatien      you were doch last year in Croatia   “But you were in Croatia last year.”  ‘[I believe you believe that] you were in Croatia last year.’    b. Du warst letzt's Jahr in KROATIEN      you were last year in Croatia     “(But) you were in Croatia last year.”    Here the discourse situation introduces a relevant contrasting proposition ¬p (I war no nia am Mea’). The function of doch supports the contrastive effect, but does not   199 induce it. This is because it can be reasonably assumed that as a rational discourse participant, Hansi knows where he was last year, i.e. at the ocean in Croatia. It is also possible, however, that he might have forgotten, or simply had another “mental image" of the ocean than the one evoked by the Mediterranean around Croatia. Kathrin’s use of doch foregrounds the fact that p is in GroundA at the time of utterance. Yet given Hansi's previous assertion that he has never been at the ocean, the use of doch here can be construed as contrastive. All doch does is to show that S (Kathrin) thinks A (Hansi) knows that he was in Croatia. Contrast falls out from inferencing, but need not be explicitly encoded in the particle.  Fully in line with the reasoning pursued here, Krifka (2013), too, argues that the contrastive interpretation associated with the use for doch is due to its context of use.75   We have now seen that contrastiveness is not a sufficient condition for the use of doch. The data discussed in the following subsections, establish that contrastiveness is also not a necessary condition for the use of doch. Hence we can conclude that contrastiveness is not directly encoded in the lexical entry of doch.                                                   75 Krifka argues that DPRT doch derives from a propositional anaphor doch that denotes a proposition p, presupposes an alternative proposition p′ and asserts that ¬[p ∧ p′] are part of CG. He argues that it is the implicit p′ which doch expresses an adversative attitude to, but that this arises via the context. He presents the following response particle use, which he judges as a“clearly non-adversative use[s]”, e.g. after negated questions:  (44) A:  Hat  er  keinen  Keks   gestohlen? has  he  NEG.DET  cookie  stolen?  “Didn’t he steal the cookie?”  B: Doch.       “Yes (he did)”    200 5.6.2  Reminding Another context of use compatible with the core function of doch is the reminding context.   The following example is from Hentschel (1986) who refers to this us as “erinnerndes doch” 'reminding doch' (cf. Gast 2008).76  (33) Da    war    doch   neulich der schwere Unfall  in unserer Straße.    there was    doch   recently DET heavy accident in our street    Und  da  hat  sich jetzt  ergeben,  dass…   and  there  has  refl.  now  turned.out  that….  “(As you might remember), there was this bad accident on our street recently. And so it turns out that….” ‘[I believe you know that] there was this bad accident in our street recently. And so it turns out that…’    Reminding doch, according to Hentschel (1986:134) refers to a ‘presupposed mental state of the addressee’ (“angenommener gedanklicher Zustand des anderen”). This yields the following epistemicity matrix for the reminding function of doch. Note that there are two contexts to consider: S can either assume that A does not believe p (¬Bel (A,p) at t<U) , or else S may not know anything about A’s epistemic state regarding p.       t<U tU S A S A reminding (1) Bel p ¬Bel p Bel p Bel p reminding (2) Bel p  Bel p Bel p Table 24: Epistemicity matrix for reminding                                                 76 This example is presented as illustrating a reminding use, yet it gives no contextual information or background, which would allow us to track the actual discourse participants’ epistemic states. In other words, the contribution of doch here is merely asserted, but not actually shown, as is the case with many other DPRT examples in the literature.   201 I now show that doch is compatible with contexts which instantiate this type of epistemicity matrix. Consider the example in (34) which provides an elaborated context for the utterance (in 33). A suitable discourse context for (33) will minimally have to include that S and A are familiar with each other, e.g. they may be neighbors. If familiar with each other, and under the premise that as long as A can reasonably know p, S can skip the acceptance phase by using doch.77    (34) Cx: One neighbor meets the other at the garden fence. The     conversation turns to the eternally debated intersection in the    village.     Do war doch letzt's Jahr der schwere Unfoi…  there was doch last year the heavy accident      …und do ham's   bei da Gemeinde na gsogt…   …and there have.they at DET city.hall then said     …dass do a Ampe   heasoi.   …that there a traffic.light  here.shall   “As you know there was this bad accident last year, so the folks at the city hall           said that they would put a traffic light up.” ‘[You know that] There was this bad accident last year, and then the folks at the city hall said that they would put a traffic light up.’    (33) and (34) would be felicitous as an opening statement in a conversation between two neighbors. In this instance, even if A might in fact not know p, S can safely assume that A can reasonably know p, since noteworthy news like a big accident down the street can be assumed to be mutually known to neighbors. Both examples crucially are not a felicitous way to open a conversation between strangers, and could not be used in a context where I chat with the person next to me sitting on the bus,                                                 77 I will show in 5.8 that the kind of ‘accommodation’ which is induced by DPRTs is different from presupposition accommodation in the sense of von Fintel (2008).    202 whom I just met.   This is illustrated in the example below where two strangers talk to each other on an airplane. Recall that in this context S-oriented ja is felicitous. In this context, doch is infelicitous because A cannot reasonably be assumed to believe p. Hence the acceptance phase cannot be skipped.  (35) Cx1:  My mom tells a stranger next to her on the airplane about her regular   trips to Canada:  <*, Cx1>  I hob doch Enkel  drum…  I have doch grandchildren over.there      …do  fliag e mindastens oamoi im Jahr.    …there fly    I at.least  once in.DET year       “As you know I have grandkids over there, so I fly over at least once a year.”   ‘[I believe you believe that] I have grandkids over there, so I fly over at least   once a year.’  The reminding effect for doch in all of the scenarios presented above falls out as follows: the normal course of grounding an assertion with content p involves a two step process. Crucially the preparatory conditions of an assertion include that A does not know p at the time of utterance. In a context where S has no explicit evidence whether p is in GroundA at the time of utterance, but S can safely assume that A knows p, or reasonably can know p, doch serves as a reminder of that knowledge about p. It is in that sense the DPRT marks a departure from the normal course of the conversation, and signals a non-standard assertion in the sense of Zeevat (2003).   Witness two other reminding contexts with doch below. The first context in (36) makes it clear that both S and A share knowledge of p. Here doch is felicitous. As   203 shown in (36) if the interlocutors do not share knowledge of p, doch is not.   (36) Cx1: Franz hears music on the radio, which is played on the intercom in the   supermarket where he is shopping with his dad. His brother Sebastian   is on a field trip at a radio station. When the station identifier is played,   Franz says to his dad:    < ✓, Cx1 > Do  is  doch  heid  da  Sebastian.    there  is  doch  today  DET  Sebastian   “SeBASTIAN is THERE today!  ‘[I believe you know that] Sebastian is there today.’    (Wer früher stirbt…)  (37) Cx2: Franz hears music on the radio, which is played on the intercom in the   supermarket where he is shopping with his dad. His dad is not very   involved with the boys, and usually doesn’t know what they are up to   in school. When the station identifier is played, Franz says to his dad:    < *,Cx2 > Do  is  doch  heid   da  Sebastian.     there  is  doch  today DET  Sebastian     “SeBASTIAN is THERE today!     ‘[I believe you know that] Sebastian is there today.’   The proposal according to which doch marks that p is in GroundA can derive the reminding function. In contrast, if contrastiveness were built into the lexical entry of doch, then the reminding function cannot be explained. Contrast in the next example (38) is introduced by the accent on the negative particle NED and on the verb IS (introducing a polarity/verum focus effect, Höhle 1992).  (38) Cx: There’s an Ox race, and Sir Quickly’s Ox Ringo is nervous. Sir has to   go to get a cassette tape with Beatles music, since this is the only way   Ringo will run the race. Sir asks Martin to take care of Ringo while he   is gone, and to calm him down, since there is a lot of commotion. A   Bavarian band is playing marches in the background.     Sir:  An tambourine man kennst du ned, gä? Des kannt’st eam vorsinga.    204    “You don’t know ‘Tambourine man’, do you? You could sing that for    him.”    Martin: Jetz schlaich de, an Defiliermarsch pfeif’e eam vor!           “Get lost now, I’ll whistle the Defiliermarsch for him!’ 78    Sir:  NED an     Defiliermarsch…           NEG  DET    defilier.march           …drum   IS a  doch so nervös!    …PRT.REASON  is he  doch  so nervous              “Not the defiliermarch, that’s why he’s so nervous!’  Consider the next dialogue as final support for the hypothesis that doch does not make reference to a contextually salient, contrasting proposition ¬p. In the discourse below Sebastian asserts that he does not want to die (p). Sepp in his response with doch does not contradict Sebastian, but affirms p. That is “not dying” is activated with both Sepp and Sebastian, and Sepp is in agreement with Sebastian about this. Here all doch does again, is to mark that p is in GroundA.  (39) Cx: Sebastian, a 11-year-old boy is obsessed with the afterlife. He is    worried that he will end up in hell, and not in heaven.     Sepp:  Wos is’n los mid dia? Wos schaust’n a so?     “What’s going on with you? Why the sad face?”  Sebastian:  I  deaf     auf ga    koan      Foe  steam.      I  be. Allowed     on  INTENSIFIER NEG.DET case  die    “I can’t die under any circumstance.”    Sepp: A      geh du  Esel! Du muast  doch  aa  ned steam!      PRT   go   you donkey you must   doch  aa79  NEG  die                                                      78 This is a very typical, iconic Bavarian march. 79 aa (from auch, ‘also’) is a DPRT here. It has an additive core function.    205    “C’mon, you donkey! You don’t have to die!”    ‘[You believe that] you don’t have to die.’   5.6.3  Backchecking Related to the reminding function of doch is its backchecking function. Backchecking refers to a context where S believes p, but wants to confirm that p is indeed true. This is summarized in the epistemicity matrix in Table 25.   inference/effect t<U tU S A S A backchecking Bel p Bel p Bel p and p true? Bel p Table 25: Epistemicity matrix for backchecking  I show here that backchecking is derived from the discourse situation, and is not directly encoded in the lexical entry of doch. Consider the example in (40) where S checks with A whether p is true.   (40) Cx1: Someone who knows my dad runs into my brother whom she hasn't   met. The family resemblance is striking. She says:     Du bist doch am Thoma sei  Bua, oda?  you are doch DET Thoma his  boy CONF   “You are the Thomas' boy, right?”     ‘[You believe that] you are the Thomas' boy [confirm that this is true].’  The backchecking effect in this example arises due to the context of use. In particular, the sentence final particle oda is used to request confirmation that p from A. (Wiltschko and Heim 2016 for a detailed analysis of confirmationals such as oda and geu). The object of confirmation can vary, however, from confirmation of the truth to   206 the proposition to confirmation of A’s belief about the proposition. Note that in the same context as above, the utterance below is equally felicitous without doch indicating that backchecking is not a sufficient condition for the use of doch.    (41) Cx1: Someone who knows my dad runs into my brother whom she hasn't   met. The family resemblance is striking. She says:   Du bist am Thoma sei  Bua oda?  you are DET Thoma his  boy or   “You are the Thomas' boy, right?”    ‘You are the Thomas' boy [confirm that this is true].’  Doch in (40) above adds the additional information that, according to S, A knows p. Therefore backchecking obtains without doch, but the DPRT, due to its contribution is highly compatible with this context. The next example illustrates the same effect.   (42)  Cx2: I’m walking with my dad through the home village. I see a young man   coming out of the bakery and ask my dad.   a. Des is am Lechna sei  Bua oda?  that is DET Lechna his  boy or    “This is the Lechner’s boy, right?”    ‘[confirm that it is true that] this is the Lechner’s boy.’  b. Des is doch am Lechna sei Bua oda?    that is doch DET Lechna his boy or    “This is the Lechner’s boy, right?”    ‘[Confirm that you believe that it is true that] this is the Lechner’s boy.’  c. Des is doch am Lechna sei Bua.   that is doch  DET Lechna  his  boy    “This is the Lechner’s boy”    ‘[I believe you believe that] this is the Lechner’s boy.’  (42a) shows that, as above, backchecking obtains due to the particle oda, and is   207 independent of the function of doch. (42b) shows that doch as expected is compatible with this context. (42c) shows that doch can be used in the context without the particle oda but here the interpretation isn’t with the backchecking function.   In sum, backchecking is a contribution of the context, not a direct contribution of doch. The function of doch is merely compatible with it.   5.6.4  Shared knowledge Propositions marked with doch are often described as marking shared knowledge between S and A. Accordingly, doch would mark p as being in the CG. For example Thurmair (1989), Zeevat and Karagjosova (2007) and Grosz (2010b) argue that doch marks its host utterance as ‘familiar, old, given, uncontroversial or shared'. I show now that the core function of doch proposed in (29) is compatible with contexts that require shared knowledge.  The epistemicity matrix for an utterance that expresses a proposition that is shared knowledge is shown below.    t<U tU S A S A shared knowledge Bel (p) Bel (p) Bel (p) Bel (p) assertion Bel (p)  Bel (p)  Table 26: Epistemicity matrix for shared knowledge  Recall that p does not become part of the CG simply by virtue of uttering an assertion: A needs to accept p either by means of explicit agreement (e.g. ‘yes/I agree’) or implicitly by not contradicting S. For a regular assertion, S does not believe that p is in A’s ground at the time of utterance.    208  As argued above, due to the core function of doch, S essentially proposes that the acceptance phase is skipped. This derives the shared knowledge effect: with uttering [doch (p)], S indicates that she assumes p to be in GroundA already. Since by virtue of asserting p, S also indicates that p is in GroundS, this means that p must be in the CG.  The example below, repeated from Chapter 4 shows that if it is obvious that p cannot be in GroundA, doch is predictably infelicitous.   (43) Cx1: I’m going Christmas shopping with my cousin and am justifying the   purchase of a nice and cozy cashmere scarf for her mom, my aunt. I   say to her:   a. <✓, Cx1> Dei    Mama friads doch immer so…    your  mom freezes doch always so       …do     is des  genau   des richtige.   …there is DET  exactly DET right  “Your mom is always so cold so this is just perfect.” ‘[I believe you believe], your mom is always cold, so this is just perfect.’   b. <*, Cx1> Dei   Mama  friads  immer   so…    your  mom freezes  always so…       …do    is des  genau   des   richtige.    …there is DET  exactly DET right (44) Cx2: I’m going Christmas shopping with a friend and am justifying the    purchase of a nice and cozy cashmere scarf for my aunt. I say to her:    a.  <*, Cx2> Mei Tante friads doch imma so…     my aunt freezes doch always so      …do  is des genau des richtige.    …there is DET exactly DET right  “My aunt is always so cold so this is just perfect.” ‘My aunt is always so cold so [I believe you believe this] this is just  perfect.’   209  b. < ✓, Cx2 > Mei Tante  friads  imma so…    … do is des genau des richtige.  Consider another example, which provides further evidence that doch appeals to a belief in GroundA, which is in turn compatible with a context in which knowledge is shared. In those cases where it already is shared knowledge that p, doch can underscore this, as illustrated in 5.5.1. In contrast, where p is not shared, but S marks p as known to A with doch, this actually can be challenged.80 The following example was observed as a natural discourse.   (45) Cx: Sonja who lives in far away Canada is home visiting. Her mom left the   house, saying she had to go to the doctor. That particular doctor had an   office in the village of Valley, as well as in another village. Sonja asks   Hans about where her mom went.    Sonja:  Is’d Mama noch Valley zum Dokta g’fahrn?         “Did Mom go to the doctor in Valley?”    Hans:  De ham doch scho lang zuag’macht.     they have doch already long closed  “As you know, they already closed a long time ago.” ‘[I believe that you believe that] they closed a long time ago.’     Sonja:  Wia soi I des wissn? Papa- I wohn nimma do!     “How am I supposed to know that? Dad- I’m not living here any     more!”                                                   80 The fact that the assertion with doch, in particular the idea that p is known to A can be challenged reveals a deeper fact about doch if not about DPRTs in general; only at-issue content can be challenged, but if the contribution of doch here can be challenged, as this example suggests, then the DPRT seems to be ‘at-issue’, counter common analysis (e.g. Kratzer 1999, Gutzmann 2008, 2013). Following Rett (2013), I consider it to contribute to illocutionary content, which encodes epistemic stance. It seems illocutionary content can be challenged.   210    Hans:  Ja <nods>,  oiso. Dann. De ham zuagmacht.     yes          well then they have closed   “Yes. Well. In that case. They closed. “        (Conversation observed Dec 8, 2015)  This naturally occurring exchange is illuminating insofar as Hans’ response marks p as known to A (Sonja). Sonja challenges this assumption about p being (reasonably) known to her explicitly in her response. Since she is living abroad, she corrects Hans by expressing that it is NOT reasonable for her to know this.  Hans, realizing, corrects himself.   5.6.5  Summary I argued in this section that the functional range of the DPRT doch can be derived from the core contribution of the DPRT and the context it occurs in.   (46) fDPRT,doch = doch + Cx  Shared knowledge, backchecking, reminding, and in particular, contrast, was argued to arise via a process of pragmatic inferencing. In cases where the discourse situation introduces a relevant contrasting proposition ¬p, the use of doch has a contrastive effect. Where the context introduces shared knowledge, doch has a shared knowledge effect, etc. This effect can also come about without the DPRT. I summarize here the epistemicity matrices introduced for each context doch was discussed in.   (47) doch(p) = Bel (A,p) (tU)    211  t<U tU S A S A contrast Bel (¬p) Bel (p) Bel (¬p) Bel (p) reminding Bel (p) ¬Bel (p) OR -- Bel (p) Bel (p) backchecking Bel (p) Bel (p) Bel(p) and p true? Bel (p) shared knowledge Bel (p) Bel (p) Bel (p) Bel (p) Table 27: Functions of doch  What remains constant across all contexts within which doch is felicitous is the fact that A believes p at the time of utterance.  This is precisely the contribution of doch as argued in Chapter 4. Since the rest of the epistemicity matrix varies, in each context the apparent function of doch also varies. This derives the functional range of doch.  5.7 The functional range of fei The DPRT fei has not received much attention within a formal framework (except Schlieben-Lange 1979, Thoma 2009). From the data and reports about the general meaning of the DPRT, two main functions of fei emerge, newness (Schlieben-Lange 1979) and emphasis (Merkle 2005). These functions are illustrated below.   (48) DPRT functions of fei DPRT  Function fei f1: newness f2: emphasis  In what follows I show that these functions can be derived from the core meaning of fei and the context; the nuanced interpretation arises through variations in context fei is compatible with.     212  In Chapter 3 I proposed the following contribution of the DPRT fei.  (49) fei ≈ I believe you don’t believe p  This amounts to saying that with the use of fei, S indicates that she believes that p is not in GroundA at the time of utterance as in (50).   (50) fei(p )≈ ¬Bel (A,p) (tU)  This proposal captures the two conditions of use for fei I presented in Chapter 4. Fei can be used in contexts when (i) A is committed to ¬p, as well as when (ii) A has no public commitment to p. It follows that with the use of fei S signals that the acceptance phase of the grounding process can be bypassed again.81 This is illustrated below.    Ground S Table Ground A Presentation phase Bel (p) ¬Bel (A, p) à p, Bel (S,p) p, ¬Bel (A,p)  à Bel (S,p) Bel (S,(¬Bel (A,p)) Acceptance phase    Common Ground Bel (S,p) Figure 13: Grounding with fei  I show now how the newness effect for fei, as well as the emphasis effect reported for the DPRT can be derived from (50) and the contribution of context.                                                  81 Martina Wiltschko points out that the insight that the acceptance phase is bypassed should be observable in the responses to utterances with these DPRTs. The investigation of possible responses to utterances with DPRTs in general opens up another avenue of research,and potential evidence toward the contribution of DPRTs.    213 5.7.1  Newness Schlieben-Lange (1979) considers the contribution of the DPRT fei as marking the propositional content of the host utterance as new. Before turning to the role of fei let us look at a context in which the content of an utterance can be considered new. Newness can in principle arise in two contexts. The first is when A has no belief about p at the time of utterance (and the time preceding it). This corresponds to the preparatory conditions under which a regular assertion with content p is uttered. This includes situations in which it is not obvious that A believes p. That is, S utters a sentence when she believes its content is new to A. Hence, by hypothesis, an assertion introduces new information. But since this is part of the normal condition for assertions, it need not be marked by DPRT. Instead, what is relevant for the present purpose is a different type of context where p is considered new. If S has reason to believe that A doesn’t believe p at the time of utterance and the time preceding it, then p is new. This is the special content that deviates from the normal conditions for assertions and hence is marked by the DPRT fei. The relevant epistemicity matrix is given below and compared with the one for normal assertions.  inference/effect t<U tU S A S A newness Bel (p)  ¬ Bel (p) Bel (p) ¬ Bel (p) assertion Bel (p)  Bel (p)  Table 28: Epistemicity matrix for newness  I have shown in Chapter 4 that fei is compatible with a context where S has reason to believe that A doesn’t believe p. In such contexts, an assertion without fei is not acceptable, as shown in (51).   214 (51) Cx: Dani and Alex are driving in the car together. Dani is the driver. They   approach a red stoplight, but Dani shows no sign of slowing the car or   engaging the brakes. Alex says:   a.  <✓, Cx> Es is fei  rot.     it is fei  red       “Hey it’s red.”     ‘[You don’t believe that] it’s red.’    b.  <✓, Cx> Es  is ROT  √.82  it is red   “It’s red!”   c. <*, Cx>  Es is rot.     it is red  The newness-effect associated with fei can be derived from its core function and one of the conditions for uttering assertions with newness effect. What fei adds in (51) is that according to S, A’s belief diverges from the normal conditions for assertions. This normal condition includes that for all S knows, A doesn’t believe p. However, fei expresses that S explicitly believes that A does not believe p.  That is, the contexts for uttering [fei (p)] felicitously differs from those in which p alone can be uttered felicitously, as evidenced by (51c).   A seeming counterexample to this is presented in the context below, repeated from Chapter 4.  Both utterances, one with and one without fei, are felicitous in this example.                                                   82 Extra high pitch on the predicate/verb has an effect that has been reported to be similar to verum focus (cf Höhle 1992); it evokes alternatives to the whole proposition, along the lines of [it is red, it is not red]. This leads to the effect of emphaisi on the whole utterance. This accent is necessary in this context to render the utterance acceptable. More on this in Chapter 6.   215  (52) Cx1: Hansi returns from the washroom. His zipper is down. Hanni says to   him:  a.  <✓, Cx1> Dei Hos'ntiarl  is  fei auf     your pant.door is fei  open      “Your fly is down.”     ‘[I believe you don’t believe] your fly is down.’   b. <✓, Cx1>   Dei Hos'ntiarl is auf     your pant.door is doch open          “Your fly is down.”     ‘[I believe you believe] your fly is down.’  The difference between an assertion without fei and one modified by fei is that in the latter, S explicitly expresses her assessment of A’s epistemic state as ¬Bel (A, p). It can be safely assumed that it is situational knowledge that people do not walk around with their zippers open. Since Hansi has his zipper open, in this context Hansi (A) shows evidence to Hanni (S) that he doesn’t believe p (that his zipper is open). fei is sensitive to exactly this aspect. This assessment about A’s epistemic state as ¬Bel (A, p) is missing in (52b). Here S simply states that p, and for all S knows, A may or may not know p. The context in (52) is compatible with both utterances, but a more fine-grained context that was presented in (51) showed that fei is sensitive to a difference in epistemicity; S has evidence that A doesn’t believe p. In particular, fei is not felicitous in regular assertions, uttered in a context compatible with a plain assertion. This is shown next:  (53) Cx2: It’s late evening, and Hanni comes down the stairs into the living    room where Hans is sitting. She utters:     a. <✓, Cx2> I  ko  ned  schlafa. I can  NEG sleep    216 “I can’t sleep.”        b. <*, Cx2> I ko  fei  ned  schlafa.     I can  fei NEG sleep      “I can’t sleep.”     ‘[I believe you don’t believe that] I can’t sleep.’  This data shows that whereas fei is compatible with a newness context, i.e. where S has clear evidence that ¬Bel (A, p), whereas it is not compatible with a context which does not give S evidence that ¬Bel (A,p); this is the context for a regular assertion.  5.7.2  Emphasis Closely connected to the newness effect is the emphasis effect of fei reported in Merkle (2005), and anecdotally observed by several of my consultants. I show now that newness and emphasis are essentially derived the same way, but in different SAs. Newness is related to assertions, whereas emphasis is related to requests. The epistemicity matrix for emphasis is repeated from 5.5.3.  inference/effect t<U tU S A S A emphasis Bel p and Bel (¬Bel (A, p)  Bel p and Bel (¬Bel (A, p)   Table 29: Epistemicity matrix for emphasis  The emphasis effect can be derived from the constitutive rules associated with uttering certain speech acts. Recall that an assertion, by virtue of including the preparatory condition that p, is not obviously known to A, and that S wants A to add Bel (p) to GroundA. This in and of itself has a newness effect; assertions are uttered when S thinks p is new to A. This was shown to be independent of the DPRT.    217  The same is the case for requests. The constitutive rules for requests include the condition that it is not obvious to both S and A that A will take the requested action without the request expressed with the utterance. In effect, uttering a request is instructing A to do the requested action in a context where there is no evidence that A would do p without being told. That is, the request is new to A; this is how newness and emphasis are related.  The felicitous context for the use of a request with fei adds the modification that according to S there IS evidence, or a reasonable assumption that A wouldn’t do p without being told. This is illustrated in the next example, which is based on the premise that kids have a tendency to not always behave like perfect angels. This is of course something that their moms are particularly aware of.   (54) Cx:  When her children are saying goodbye to spend the weekend with    Grandma, mom tells them:   Seid-s  fei   schee brav     bei  da   Oma!  be.2pl  fei   niece well.behaved   at    DET   Oma  When a mom reminds her kids to behave, she (S) does this under the presumption that the kids (A) didn’t already have this as an instruction in their GroundA. That is, Bel S (¬Bel (A,p)). This is how the emphatic effect of adding fei to the utterance in (54) can be understood.   The next example serves to further illustrate this effect. To utter a request fei (p) in a context where S has can reasonably assume that A would do p is odd.   (55) Cx: Karl always closes the windows when leaving the house. His wife   Christa, who is away for a few days checks in with him over the    phone. She reminds him:    218   a.  <*, Cx> Mach fei as Fensta   zua  bevoa’st  gehst,  gä    make fei DET  window closed before.2sg go.you  CONF.        “Do close the window before you go, eh!”    ‘[Confirm that [you don’t already intend to]] close the window.’   b.  <✓, Cx> Mach as Fensta   zua  bevoast gehst, gä     make DET window closed before.2sg go.you CONF.           “Close the window before you go, eh!”    ‘[Confirm that [you intend to]] close the window.’  5.7.2  Summary We have seen in the previous section that the functional range of fei, expressing both newness and emphasis derive from the same contextual conditions. Both functions are an artifact of the contexts fei occurs in, in combination with the core contribution of the DPRT itself.   (56) fDPRT,fei = fei + Cx  The epistemicity matrixes for both newness and emphasis are the same and are summarized below.   (57) fei(p) ≈ ¬Bel (A,p) (tU) inference/effect t<U tU S A S A newness Bel (p) ¬ Bel (p) Bel (p) ¬ Bel (p) emphasis Bel (p) ¬ Bel (p) Bel (p) ¬ Bel (p) Table 30: Functions of fei  The perceived difference between newness and emphasis arises due to the different   219 speech acts fei occurs in. Crucially, both effects are not directly encoded in the lexical entry for the DPRT. I showed that newness (assertion) and emphasis (request) obtain without the DPRT, and that fei is therefore not necessary to for these two functions.     5.8 DPRTs, presuppositions, and accommodation We have seen contexts in the sections above which showed that DPRTs are compatible with shared knowledge. The question arises how DPRTs are related to accommodation then. Consider again the two DPRTs ja and doch in an example repeated from above:  (58) Cx1: A family with two little twins is at the desk of a car rental place. They   chose a compact car for rental. The clerk says:   a. Sie ham ja Zwilling…   you have ja twins     …woin’s do ned a gressa’s Auto nehma? …want.you there NEG DET bigger  car take  “Since you have twins, don’t you want to take a bigger car?” ‘[I firmly believe that] you have twins, don’t you want to take to a bigger car?’   b. Sie ham doch Zwilling…   you have doch twins     …woin’s do ned a gressa’s Auto nehma? …want.you there NEG DET bigger  car take  “Since you have twins, don’t you want to take a bigger car?” ‘[I believe that you believe] you have twins, don’t you want to take to a bigger car?’      220  c. Sie ham Zwilling…   you have twins     …woin’s do ned a gressa’s Auto nehma? …want.you there NEG DET bigger  car take  “Since you have twins, don’t you want to take a bigger car?” ‘You have twins, don’t you want to take to a bigger car?’   It seems that DPRTs like ja and doch instruct A to accommodate p, particularly in circumstances where it is not clear that A believes p.   However, as I argued above, the DPRTs are compatible with this shared knowledge context, but they do not induce it. Therefore the type of ‘accommodation’ happening with DPRTs is different from what is generally referred to as accommodation. Consider the next example, which forces accommodation of p.  (59) I have to bring my three kids to daycare   If I uttered this sentence to a complete stranger, they could accommodate that I have three children (and that they are going to daycare). This, as was shown, is different with DPRTs. An addition of doch in (60), for example, does not allow A to accommodate that they already know the fact that my mom has three grandchildren in Canada; accommodation of previous knowledge of this is simply impossible for A and the use of doch is not possible.   (60) Cx1:  My mom tells a stranger next to her on the airplane about her regular   trips to Canada:   <*, Cx1>  I hob  doch Enkel              drum…     I have  doch grandchildren  over.there   221       …do  fliag’e  mindastens  oamoi  im   Jahr   …there  fly I   at.least   once   in.DET  year    “I have grandchildren over there, so I fly at least once a year.”   ‘[I believe that you believe that] I have grandchildren over there, so I fly at least   once a year.’    I presented a discourse condition in which S and A have to be familiar enough to share background assumptions for the felicitous use if doch. I called this what can reasonably be known (by A). Thurmair (1989) for example notes an increased use of DPRTs in informal and familiar oral language settings over formal settings; the proposal presented here, namely that DPRTs are used by a given speaker to express a certain epistemic state (either hers, or her interlocutors’), provides an explanation for the rise of DPRT use in familiar settings without introducing a presuppositional meaning component per se. Note that DPRTs are not always ill-formed in a formal setting either; whereas a speaker is not in a position to make assumptions about a stranger’s epistemic state out of the blue, she is, with evidence (either via linguistic antecedent or via actions which indicate an epistemic state) in a position to make an assumption about A’s epistemicity. That is, the discourse context (provided by Cx: form, Cx: disc, Cx: sit and Cx: world) places S in a position to make a reasonable assumption about A’s epistemic state.  The reported ‘friendliness’ effect of DPRTs (Thurmair 1989) is achieved in that DPRTs indicate to an addressee that the speaker takes her perspective and mental state into account, which is why the speaker’s contribution using a DPRT is often conceived as amiable (Weydt 2006).    222 5.9 Conclusion In this Chapter I addressed the question regarding the multi-functionality of DPRTs within their DPRT function, introduced in Chapter 2. I showed that DPRTs vary in their interpretations depending on the context. Context includes the immediate discourse, the situation in which the discourse occurs, as well as the world-knowledge shared by all discourse participants.  (61) fDPRT  + CxDisc,Sit,World   à functional range  I conducted three case studies deriving the functional range of three MB DPRTs: ja, doch and fei. I showed the variety of functions they fulfill is determined by their core function as DPRT, and the contribution of the discourse context they occur in. In particular, epistemicity matrices were used. These show a descriptive model of the epistemic states of the speech act participants in a specific context, and are independent of the DPRTs. The matrices show that an individual DPRT, in particular, a meaning component expressed with the DPRT may be compatible with certain contexts, but that the DPRT does not have to encode fine-grained contextual information itself. Furthermore, the variation in interpretation arises from the variation in the context the DPRT appears in. This compatibility was indicated by bolding the respective meaning components that are compatible. I repeat the matrices below.      223 (62) ja (p) = Bel (S,p) (tu)  and  ∀ t >tU : Bel (S,p) (t)   inference/effect t<U tU S A S A shared knowledge Bel (p) Bel (p) Bel (p) Bel (p) surprise   Bel (p)  emphasis Bel(p) and Bel(¬Bel (A, p)  Bel (p) and Bel(¬Bel(A,p)  reason Bel (pj) and Bel (pi)  Bel (pi)  assertion	 Bel (p)	 	 Bel (p)	 	Table 31:The functions of ja  (63) doch(p) = Bel (A,p) (tU)  inference/effect t<U tU S A S A contrast Bel(¬p)  Bel (p) Bel (¬p) Bel (p) reminding Bel (p) ¬Bel (p) OR -- Bel (p) Bel (p) backchecking Bel (p) Bel (p) Bel (p) and p true? Bel (p) shared knowledge Bel (p) Bel (p) Bel (p) Bel (p) Table 32: Functions of doch  (64) fei(p)≈ ¬Bel (A,p) (tU) inference/effect t<U tU S A S A newness Bel (p) ¬ Bel (p) Bel (p) ¬ Bel (p) emphasis Bel (p) ¬ Bel (p) Bel (p) ¬ Bel (p) Table 33: Functions of fei  The discussion in this chapter addresses one of the core properties of DPRTs introduced in 2.2.1, showing how the multi-functionality derives from contextual factors. I also briefly addressed whether DPRTs can be considered presuppositional in the sense of Stalnaker (1978, 2002). First of, as already shown in the previous chapter, not all DPRTs make reference to shared beliefs in the CG, and in that way   224 cannot be considered presuppositional. Any presupposition arising form an utterance is independent of the DPRT. In cases where DPRTs make reference to A belief, I showed that they do not allow A to accommodate S’s belief about A.        225 Chapter 6: The syntax of discourse particles  6.1 Introduction Thus far, I have established that the DPRTs I investigated can be divided into three basic classes: A-oriented, S-oriented and O-oriented. I showed for a subset of these particles that their functional range can be derived from their core meaning in interaction with their context of use.  The purpose of this chapter is to show that orientations too can be derived, however in different ways. In particular, I argue that orientations are syntactically conditioned.   I start by ruling out a hypothesis according to which these orientations are lexically encoded. I refer to this hypothesis as the lexical hypothesis and discuss it in section 6.2. I show that the lexical hypothesis faces a variety of problems. I contrast this approach with the syntactic hypothesis according to which the orientations are syntactically conditioned (section 6.3). These two competing analysis are summarized below.  (1)  Two hypotheses to account for DPRT orientations    a.  Hypothesis 1 -lexical hypothesis:      difference in function is due to difference in UoLs    UoL1 : f1  UoL2 : f2   UoL3 : f3         226   b.  Hypothesis 2 -syntactic hypothesis:      difference in function is due to the syntactic context    f1= UoL +CxSYN1  f2= UoL +CxSYN2  f3= UoL +CxSYN3   Given the relevance of DPRTs to speech acts, and due to their sensitivity to the epistemic states of the speech act participants, I frame this discussion within current theories of the syntax of speech acts and speech act structure. I show that the Universal Spine Hypothesis, introduced in Chapter 1, serves as an ideal model to capture the findings. I adopt existing proposals to extend the Universal Spine, and show that DPRT functions can be modeled with an Extended Universal Spine.   I show several pieces of evidence in support of the syntactic hypothesis, in 6.4, 6.5 and 6.6. These come from the obligatoriness of DPRTs in certain contexts, the co-occurrence patterns of DPRTs, scopal interactions with other particles (confirmationals), speech act adverbs and speaker oriented adverbs, as well as accent on DPRTs. In 6.7 I hypothesize about the peculiar syntactic integration of DPRTs, and 6.8 summarizes  6.2 Lexical hypothesis How shall we account for the DPRT orientations established in Chapter 4? According to the lexical hypothesis, S-, A-, and O-orientations are lexically encoded in each DPRT, i.e., they form an intrinsic part of their lexical entries as in (2).      227 (2)  Hypothesis 1 -lexical hypothesis:    difference in function is due to difference in UoLs  UoL1 : f1 UoL2 : f2  UoL3 : f3   This type of lexical approach seems to be widely adopted by the majority of DPRT researchers. Most take DPRTs to be lexically specified to encode their contribution, as different from other functions (e.g. Helbig 1988, Thurmair 1989). Some take their semantic specification to be encoded (e.g. Grosz 2010), others take clause-type compatibility to be a syntactic feature specification (Bayer and Obenauer 2011). In this section, I review three problems that arise under the assumption that DPRTs are (semantically or syntactically) inherently lexically specified in this way.   6.2.1 Problem #1: Clause type restrictions are not feature specification As we have seen, not every DPRT is compatible with every clause type (e.g. Bayer 2008, Bayer and Obenauer 2011, Coniglio and Zegrean 2012). Clause type restrictions are often analyzed as being lexically encoded clause type features on the DPRT.  For example, Bayer and Obenauer (2011) assume that the SG DPRT denn does not only encode S concern via a dedicated feature [+conc], but that it is also associated with an uninterpretable feature that restricts it to interrogatives ([uQForce]). The lexical entry they assume for denn is given in (3).  (3)   denn [+conc, uQForce]  Following this type of analysis, one could specify the varying orientations of the DPRTs under investigation here as well. For example, we could posit that ja is   228 lexically specified for S attitude whereas fei and doch are lexically specified for A attitude.83  In addition, a specific force feature, such as [u xForce], akin to (3), could be made responsible for the clause type restriction we observe with these particles. Hence, we could associated ja, fei, and doch with the lexical entries in (4), respectively.  (4)   ja [+S, u xForce]  fei [+A, u xForce] doch  [+A, u xForce]  However, the postulation of a force feature raises two questions. First, it presupposes that the illocutionary force of a sentence is syntactically encoded (Rizzi 1997). I do not adopt such an approach to clause typing and illocutionary force assignment here. This is because, as we have seen in Chapter 3, clause type and illocutionary force are independent of each other. Hence the syntactic encoding of illocutionary force cannot be as simple as positing a feature. It would fail to recognize that the form of a clause (its clause-type) is in part independent of its function, i.e. its illocutionary force (cf. Allan 2006, Portner 2004, Zanuttini and Portner 2003). Secondly, each of the DPRTs we have explored was shown to be compatible with a variety of clause types, and with a variety of illocutionary forces (cf. Chapter 4). To see this, consider again doch. It can be used with declaratives with assertive force, as well with imperatives. It can, however, not occur with declaratives with question force (triggered by rising intonation).                                                     83 I keep the hypothetical lexical specifications very basic for exposition purposes   229 (5)   a.  ✓ Des is  doch  auf  \ Declarative with assertion force  DET is doch  open  b.  * Des is doch  auf  / Declarative with question force  DET  is  doch open  c.  ✓ Mach  des doch  auf  \ Imperative with directive force  make  DET doch  open   This pattern suggests that doch is sensitive to illocutionary force rather than clause-type: it is restricted to directives and assertions. If doch were lexically specified, it would have to have to include features for both directive and assertive force in its lexical entry ([u assertive, u directive]. Observe, however, that doch cannot occur in directives that are based on V1 clauses (6b), or on directives based on V2 clauses (6d).   (6)   a. Machst       des  auf  \  make.you  DET  open   “(You) open this!”    b. *  Machst  des doch  auf  \  make.you DET doch open    c.  Du machst       des  (gfälligst)  auf  \  you make.you  DET  (kindly)    open   “(You) open this!”    d. * Du machst       des  doch  (gfälligst) auf  \  you make.you DET doch   (kindly) open    “(You) open this!    This shows that sensitivity to illocutionary force is not the right way to characterize the clause type restriction of doch (Thurmair 1993); hence a feature specification for   230 illocutionary force can also not be on the right track (cf. Struckmeier 2014).   6.2.2 Problem #2: DPRTs are multifunctional As we have seen, one of the properties of DPRTs that has been widely observed is their pervasive in terms of their categorial status. As in other German varieties, most MB DPRTs have homonyms with other functions, such as discourse markers (ja, eh) scalar particles (bloß), affirmative particles (ja, doch), conjunctions (aber), adverbs (dann, jetz, scho) and adjectives (ruhig) (e.g. Abraham 1991, Thurmair 1989, Weydt 1969). The notable exceptions are the DPRTs fei and hoid (halt), which exclusively have DPRT function (cf. Kwon 2005 for halt). The various functions of the DPRTs discussed here were introduced in Chapter 3, and are repeated here for convenience; they range from response particle, to conjunction, adverb and discourse marker.   DPRT  Response particle  Conjunction Adverb Discourse marker ja ✓ ✗ ✗ ✓ doch ✓ ✗ ✗ ✓ fei ✗ ✗ ✗ ✗ eh ✓ ✓ (rare) ✗ ✗ jetz ✗ ✗ ✓ ✓ Table 34: DPRT multi-functionality  The DPRT function of these UoLs is assumed to diachronically derive from these other functions (e.g. Abraham 1995, 2000; Kwon 2005, Thurmair 1989), via the process of grammaticalization (Lehmann 1995, Traugott 1995, cf. Abraham 1991, 1995) (see Chapter 7 for a discussion of grammaticalization). DPRTs are thus considered diachronically related to their other counterparts. Yet synchronically they are assumed to be lexically specified in different ways. This amounts to saying that there are different lexical entries for these UOLs, depending on their functions.    231  The pervasiveness of the multi-functionality of DPRTs begs the question however, as to whether multiple lexical entries are really the most economical way to model it. To admit between two and four different lexical entries for every item that also serves as a DPRT would force us to assume a model of the lexicon, which routinely admits multiple lexical entries. Particles like ja and doch would need minimally three lexical entries, each defined for their specific function.  (7)   ja1 : DPRT ja2 : response particle ja3 : discourse marker  Furthermore, the pervasiveness of multi-functionality also tells us that we need a model that predicts it, rather than one that can simply deal with it by brute force.  Consider in this context the following quote from Leiss (2005: 233) (via Wiltschko 2014): “Before research into grammaticalization was established, morphemes with identical form were preferably classified as homonyms rather than as motivated polysemy. What was striking about this method is that the postulation of homonymy did not have to be justified whereas postulation of polysemy was not easily accepted. Polysemy not only had to appear plausible, but it had to be proven. In contrast, postulation of homonymy was acceptable even if it was implausible and counter-intuitive. Thus in older works on word formation, one can regularly find claims according to which propositions and form-identical verbal prefixes are homonyms. Such claims were never regarded as unscientific, to the contrary. They were – and still are – considered as an indication of methodical precaution. Many are not aware of this biased burden of proof. It   232 can be made explicit if we turn the burden of proof around in a thought experiment. Nobody seems to consider this possibility. The reversal of the burden of proof would mean that from now on postulation of homonymy will have to be proven, whereas postulation of polysemy would be considered as an indication of methodical precaution. Current methodology is different: researchers dealing with grammaticalization consider it their task to prove the motivation for polysemy. In contrast, there is no research agenda, which considers it necessary to prove and explain postulated homophony. But the cross-linguistic frequency of homophony requires an explanation. Why is the rich potential for symbolization not utilized? Why do the same forms of inflectional and derivational morphemes recur in different functions? When specific questions are not asked it is an indication that something is axiomatically excluded. What is the axiom, which would conflict with this reversal in the burden of proof the most? It is the axiom of the arbitrariness of linguistic signs.”  (Leiss 2005: 233, translation by Martina Wiltschko)   Recall that I argued in Chapter 5 that the functional range of individual discourse particles is derivable from context; the functional range of a single particle is the multi-functionality problem (polysemy in Leiss’ terms), in a different guise. Considered on a macro-level, multi-functionality can also be derived (cf. Wiltschko 2014). Assuming Wiltschko’s (2014) Universal Spine Hypothesis, which I will introduce in 6.3, allows us to derive multi-functionality as opposed to postulating homonymy; I will therefore adopt this model, and show in 6.4 how the multi-functionality of DPRTs and their S-, A-, and O-orientations are yet another   233 instantiation of the general idea that the function of a particular form derives from the UoL and the context it appears in (f = UoL +Cx).  6.2.3 Problem #3: Orientation correlates with context A third problem with the lexical hypothesis has to do with the way attitude is expressed.  Expression of S’s epistemic state is one of the most common descriptive functions cited for DPRTs (Weydt, 1969 a.o.). Speakers use them to "comment on the status of the information carried by their hosts" (Kaufmann and Kaufmann 2012:209). I showed in Chapter 4 that besides expressing S’s epistemicity with regard to the host utterance, DPRTs can also be used to convey S’s assessment about the beliefs of A (cf. Zimmermann 2011), as well as express some other participants’ belief. The expression of attitude is not unique to DPRTs, however. For example, attitude is also encoded in S(peaker)-oriented adverbs such as ehrlich (honestly), natürlich (‘naturally’), and epistemic adverbs such as angeblich (reportedly), anscheinend (‘seemingly’) (Ernst 2009, Frey and Pittner 1998, cf. Krifka 2013, Ross, 1970). The speaker attitude encoded in these adverbs is often assumed to be part of their lexical entry.  If DPRTs encode attitude lexically, as do S-oriented adverbs, we would predict that S-attitude is always interpreted as part of the lexical entry, independent of its distribution. If on the other hand S-attitude is conditioned by a particular syntactic context, we expect that S-attitude is dependent on distribution. I show here that DPRTs differ from S-oriented adverbs in that S-adverbs still have their S-oriented meaning when they occur in SpecCP, while DPRTs like fei and ja either cannot occur in SpecCP at all, or they lose their DPRT interpretation in this context. This is   234 illustrated with the examples below, which show that the S-adverb wahrscheinlich ‘probably’ can occur in SpecC (8b), whereas fei and ja cannot. Mono-functi