A unified DRT-based account of accented and unaccented middle field doch (In: Sprache & Datenverarbeitung)

Abstract

The paper presents a unified DRT-based account of the meaning and function of two apparently dissimilar uses of the German particle doch, namely accented and unaccented doch in the middle field. It is claimed that both uses express the discourse relation Correction, and that unaccented doch additionally indicates that the doch-host is assumed by the speaker to be given in the discourse context. It is further assumed that doch is weakly ambiguous between various relations of contrast, and its underspecified meaning is defined in the framework of UDRT (Reyle, Rossdeutscher and Kamp 2007). It is shown how in concrete discourse, a particular reading is selected from the underspecified meaning representation, depending on the information structure of the sentence, as well as on the syntactic and prosodic properties of the respective doch-use. This process is modelled in the framework of the most recent version of DRT (Kamp, van Genabith and Reyle to appear) and the version of DRT that takes into consideration the focus-background division of the sentence (Kamp 2004).

Full text

A unified DRT-based account of accented and unaccented
middle field doch1
Elena Karagjosova
The paper presents a unified DRT-based account of the meaning and function of two apparently
dissimilar uses of the German particle doch, namely accented and unaccented doch in the middle field.
It is claimed that both uses express the discourse relation Correction, and that unaccented doch
additionally indicates that the doch-host is assumed by the speaker to be given in the discourse context.
It is further assumed that doch is weakly ambiguous between various relations of contrast, and its
underspecified meaning is defined in the framework of UDRT (Reyle, Rossdeutscher and Kamp 2007).
It is shown how in concrete discourse, a particular reading is selected from the underspecified meaning
representation, depending on the information structure of the sentence, as well as on the syntactic and
prosodic properties of the respective doch-use. This process is modelled in the framework of the most
recent version of DRT (Kamp, van Genabith and Reyle to appear) and the version of DRT that takes
into consideration the focus-background division of the sentence (Kamp 2004).
1. Introduction
The German adversative connector doch (Engl. though, but) is notoriously ambiguous. It has
at least five syntactically and prosodically different uses that belong to different parts of
speech (conjunction, conjunct adverb, modal particle, response particle, sentence adverb) and
express various discourse relations, such as correction, semantic opposition and different
forms of concession. Two of its variants, accented and unaccented doch placed in the middle
filed, seem even to have diametrically opposed meanings, as noted first by Hentschel (1986)
and attested in (1):
(1) a. Er kommt DOCH.
'He is coming, although we believed he was not.'
b. Er KOMMT doch.
'You know he's coming.'
In Karagjosova (2009), I sketch an unitary DRT-based account of doch that takes into
consideration both its meaning and discourse effects. There, I focus mainly on two of the uses
of doch, namely the conjunction and the conjunct adverb. In the present paper, I extend this
analysis to the cases of the accented adverb and the unaccented modal particle doch. Both
uses of doch express correction, albeit of a different sort. Consider for instance (2), where at
some point A or B says Peter is coming to the cinema and later A learns that Peter is out of
town. B then draws the conclusion that if Peter is out of town, he is not going to the cinema.
The doch-utterance here does not correct An but the earlier utterance A0. The doch-variant that
1 This work is supported by the project Språk i Kontrast (Languages in Contrast) at the University of Oslo, NFR
158447/530. I would like to thank the audience of the ESSLLI 2008 workshop on formal and experimental
approaches to discourse particles and modal adverbs for discussion, as well as an anonymous reviewer for
helpful comments.

expresses this relation is accented, placed in the middle field of the German sentence and is
categorized as an adverb.2
(2) A0/B0: Peter kommt mit ins Kino.
'Peter is coming to the cinema.'
An: Peter ist verreist.
'Peter has left.'
Bn: Er kommt also DOCH nicht mit ins Kino.
'So he is not coming to the cinema, after all.'
A different kind of correction is found in (3), where the B-utterance corrects what the
speaker believes is a misconception of the hearer as regards the whereabouts of Peter: A
should know that Peter is out of town and should not have claimed that he is coming to the
cinema. In corrections like that, doch is unaccented, placed in the middle field of the German
sentence and categorized as a modal particle.
(3) A: Peter kommt mit ins Kino.
'Peter is coming with us to the cinema.'
B: Er ist doch verreist.
'He has left, as you should know.'
Based on an analysis of doch presented in Karagjosova (2008), I claim that both
accented and unaccented doch have a contrastive, more in particular corrective meaning, and
that unaccented doch additionally indicates that the doch-host is assumed by the speaker to be
given in the discourse context. I assume further that doch is weakly ambiguous between
various relations of contrast and define an underspecified semantics for doch in the
framework of UDRT (Reyle, Rossdeutscher and Kamp 2007) in terms of an UDRT
alternation, i.e. sequence of alternative DRSs. Each alternative DRS captures the different
information-structural units to which doch may be sensitive, and partly corresponds to the
interpretation that doch receives in the respective environment. I further show how in concrete
discourse, a particular reading is selected from the underspecified meaning representation,
depending on the information structure of the sentence, as well as on the syntactic and
prosodic properties of the respective doch-use. I model this process in the framework of the
most recent version of DRT (Kamp, van Genabith and Reyle to appear) and also employ the
2 The corrective meaning of this use of doch is maybe better illustrated by the following naturally occuring
dialogue example (from the Baufix corpus, http://www.sfb360.uni-bielefeld.de/transkript/, discussed in more
detail in Karagjosova 2006):
(i) A1: es geht nicht ('it does not work')
B1: du musst die Schraube drehen, [...] ('you must turn the screw')
A2: [...] hast recht, es geht DOCH ('you are right, it works after all')

version of DRT that takes into consideration the focus-background division of the sentence
(Kamp 2004).
The paper is structured as follows. In the next section 2, I introduce the background
for the analysis of the two uses of doch. The latter is presented in section 3. Section 4 offers
brief summary and conclusions.
2. Background
The meaning specification of doch that I assume here is originally proposed in Karagjosova
(2008) and is based on Sæbø 's (2003) analysis of German aber, where the semantics of this
connector is defined in terms of a contrast presupposition involving negation and topic
alternatives. In Karagjosova (2008), I apply Sæbø's analysis of aber on doch, since doch and
aber are partly synonymous.
The main idea in Sæbø (2003) is that semantic opposition is the basic contrast relation
expressed by aber from which other kinds of contrast such as various forms of concession can
be derived as a result of generating conversational implicatures based on Grice's Maxim of
Relevance. The main observation is that aber is sensitive to the information structure of the
sentence.3 More precisely, the contrast between two conjuncts C1 and C2 expressed by aber
can be seen as a semantic opposition between the contrastive topic of the aber-clause C2 and
an alternative to it that is provided by the first conjunct C1. For instance in (4), the contrastive
topic (CT) of C2, mittlere, is opposed to the CT of C1, kleine, and kleine is a contrastive
topic alternative of mittlere:
(4) [Für [kleine]T Betriebe hält sich der Schaden noch in Grenzen]C1; [für [mittlere]T aber wird
er allmählich ruinös]C2.
'For small companies, the harm is yet limited; for intermediate-size companies, however, it is
becoming ruinous.'
Based on this observation, Sæbø specifies the basic meaning of aber in terms of an
assertion and a presupposition in dynamic semantics in the following way: a sentence of the
form 'φ aber' updates the context σ to a context τ iff σ entails the negation of φ where the
contrastive topic of φ is substituted by some alternative α, and σ is updated by φ. Formally:
(5) σ[[φ aber]]τ iff σ |= ¬φ[T(φ)/α] for some alternative α and σ[[φ]]τ.
In other words, the presupposition requires that the context contains the negation of a sentence
which is just like the aber-sentence except for its contrastive topic. The contrastive topic of
the required sentence is a contextual alternative of the contrastive topic of the aber-sentence.
3 An analysis of aber based on similar observations is proposed in Umbach (2005).

Consider again (4). The presupposition can be verified, since in the negated aber-
sentence, we replace its contrastive topic mittlere, for the alternative, here the contrastive
topic of C1 kleine, and get that the harm for small companies is not ruinous. This is entailed
by C1, since C1 asserts that the harm is limited.
(6) σ |= ¬(für mittlere Betriebe wird der Schaden ruinös)[mittlere/kleine] iff
σ |= ¬ (für kleine Betriebe wird der Schaden ruinös)
The topic of the contrast presupposition is defined by Sæbø as „the portion of the
sentence for which the context provides a substitute.“ Contrastive topics are one such case.
Sæbø considers further cases which do not involve contrastive topics. He argues that there we
deal with an „implicit topic“ that in general is the complement of the apparent focus.4 A
simple example is (7), where the focus is nicht lang, and the „implicit topic“ is the
complement of the focus, namely lang. The presupposition is verified: replacing the
complement of the focus of the aber-clause, lang, for the alternative in the preceding clause,
steil, renders that the context should entail that the forest paths are steep, which is indeed so.
(7) Die Waldwege sind [steil]α, aber [nicht [lang]IT]F.
'The forest paths are steep but not long'.
σ |= ¬(¬ (die Waldwege sind lang))[lang/steil] iff
σ |= ¬(¬(die Waldwege sind steil))
In Karagjosova (2008), I show that the information structural units to which doch
pertains when trying to identify and verify the contrast presupposition, can be not only
contrastive topic or the negation of the focus, but also verum focus, the constituent in the
scope of the focussed negation, or the discourse topic. Further, when doch is interpreted as
correction, the „topic“ of the contrast presupposition, i.e. the part of the sentence for which
the context provides a substitute, coincides with the alternative, thus reducing the
presupposition to the requirement that the context contains a sentence with the reversed
polarity. For instance in (8), the focus is on doch, and the complement of the focus is nicht,
since doch asserts the sentence that hosts it. The alternative is the sentence negation nicht in
the first utterance. The alternative coinsides with the topic, and the presupposition can be
4In Sæbø (2003), the „implicit topic“ is reconstructed as a result of pragmatic reasoning that involves a process
of accommodation which in turn triggers implicatures which generate the concessive readings of the connector.
For instance in (7), the „implicit topic“ lang vs. nicht lang (or its equivalent kurz) is identified on the basis of
the following reasoning: coordination alternatives require a relevant parallel or Common Integrator (CI, Lang
1977) between them. A CI between steep and long is more plausible than between steep and not long or short
when it comes to forest paths: both steep and long paths are strenuous. Identifying the CI forest paths are
strenuos gives us also the concessive opposition reading of the sentence: the first conjunct supports the
proposition that the paths are strenuous, whereas the second runs against it.

verified, albeit not in the context of the immediately preceding utterance, but of the more
remote A1:
(8) A1: Es geht [nicht] α.
'It does not work.'
B1: Du musst die Schraube drehen.
'You have to turn the screw.'
A2: Hast recht, es geht DOCH.
'You are right, it works after all.'
σ |= ¬(¬ (es geht nicht)[nicht/nicht] iff
σ |= es geht nicht
The unaccented middle-field doch indicates intuitively that the proposition expressed
by the sentence belongs to the common knowledge of speaker and hearer. The correction
pertains therefore to the set of propositions that are assumed to be common knowledge. It is
triggered by a (manifested or hypothetical) deviant opinion on the part of the interlocutor, as
in (9), where A demonstrates lack of knowledge of the assumed common ground proposition
„Peter is out of town“: from the assertion that Peter is going to the cinema, speaker B can
infer on the background of general world knowledge and assumptions of cooperativity that A
does not know or is currently not aware of the fact that Peter is out of town since otherwise he
would not have asserted (9-A):5
(9) A: Peter kommt mit ins Kino.
'Peter is coming to the cinema.'
B: Er ist doch VERREIST.
'But he has left (as you should know).'
The fact that this doch marks the proposition expressed by the doch-host as given
information suggests that the „topic“ we are dealing with here can be identified with the entire
sentence, e.g. that Peter has left in (9). The alternative is identical with the „topic“, and its
negation is suggested by the context. Indeed, the contrast presupposition can be verified in the
context of utterance A: the sentence that Peter has not left can be reasonably assumed to
follow from the sentence that Peter is coming along to the cinema.
(10) σ |= ¬ (Peter ist verreist) [Peter ist verreist/Peter ist verreist]
σ |= ¬ (Peter ist verreist)
5 This variant of doch, namely doch as a modal particle (MP), can be felicitously used not only when the
interlocutor demonstates unawareness of a proposition that the speaker believes to belong to the common
knowledge, but also in cases when the speaker is perfectly aware that the proposition is actually not common
knowledge. This property of doch (as well as of other German MPs) is sometimes referred to as „the
perfidousness of MPs“, cf. Reiter (1980).

Based on the observation that the doch-variants differ in the information-structural
units to which they are sensitive, I argue in Karagjosova (2008) that the semantics of doch is
best captured by enumerating the different ways in which the contrast presupposition it
triggers may be instantiated in concrete discourse. I formulate the semantics of doch as an
UDRT alternation, i.e. a disjunction of alternative DRSs, which is a technique used in UDRT
(Reyle, Rossdeutscher and Kamp 2007) for specifying the meaning of ambiguous lexical
items. The representation in (11) is intended to capture this „meaning potential“ of doch:
(11)
In (11), π and π' are discourse referents for representing clauses, as in SDRT (Asher and
Lascarides 2003), π is the clause hosting doch, and Fc is the complement of the focus F. The
representation is intended to express that doch triggers the presupposition that there is a
sentence π' in the discourse context such that π' is the negation of the result of replacing the
respective information structural unit of π by its corresponding alternative. The special
disjunction sign between the DRSs is an operator used for representing lexical ambiguity
(Reyle, Rossdeutscher and Kamp 2007), and underlined discourse referents are anaphoric
referents that have to be bound to an antecedent in the context or accommodated. The two
cases which I focus on in the next section, namely (8) and (9), are captured by the third and
the last alternative DRSs respectively.6 The first DRS deals with the case where doch is
6 The presupposition of the latter, namely the modal particle doch, resembles the presupposition that proper
names introduce: by using the name the speaker presupposes familiarity with it, cf. Kamp, van Genabith and
Reyle (to appear). Similarly, by using doch, the speaker presupposes familiarity with π but also that it has been
suggested that ¬π. Admittedly, the meaning specification of this use of doch misses an important intuition,
namely the fact that in cases like (3), the doch-utterance conveyes that the speaker believes that the other
interlocutor is not aware of the proposition that is assumed to belong to the common knowledge. In Karagjosova
(2004), I provide a formalization of the meaning of the MP doch which captures this intuition, employing
notions from belief revision models such as active and inactive beliefs (Wassermann 2000). The question of how
this analysis is compatible with the one presented here and whether the unawareness component can be ascribed
to contextual enrichment must be pursued in future work.
The reason for not having just ¬π in both alternative DRSs is the idea that the meaning representation should
reflect the contextual conditions under which the doch variants are used, more specifically the information-
structural unit to which the respective doch-variant pertains.

sensitive to the contrastive topics of the two connected clauses, as in (4); the second captures
cases like (7), where the IS units involved are the foci, and the fourth captures cases where
doch is sensitive to the negated background (NB) of its host (cf. Karagjosova 2008 for more
detail on these uses of doch).
The meaning specification of doch that I propose in (11) has certain drawbacks that I
discuss in Karagjosova (2009). Among its merits is that it reflects the main property of
discourse connectors, namely their contextual sensitivity, as well as, at least partly, the
interpretation of the connector doch in the respective context: For instance, when doch
pertains to the contrastive topic of the doch-host, its interpretation is semantic opposition, and
when it pertains to the complement of the focus of the doch-host, it is a form of concession.
However, there are additional contextual parameters that determine the interpretation of doch
in a particular context that are not captured by (11). For instance, in cases when the
presupposition is reduced, the interpretation of doch may be either correction or denial of
expectation, depending on whether doch is positioned in the middle field (correction) or the
initial field (denial of expectation). In Karagjosova (2009), I show how these additional
parameters come into play in a DRT-based account of the way in which a particular
interpretation of doch emerges from its underspecified meaning under a particular contextual
setting. The idea is roughly that the DRS construction is informed by the focus annotated
syntactic tree of the sentence hosting the connector. The semantic representations are built by
means of DRT-construction rules (Kamp and Reyle 1993). DRT-construction rules are rules
that are applied to the syntactic structure of the sentence. By the application of such rules the
discourse representation of the sentence is obtained. The construction rules select the reading
that corresponds to the syntactic and prosodic properties of the doch-variant that is used in the
concrete discourse, as well as to the focus-background structure of the discourse. The selected
doch-reading is a presupposition that next has to be either bound to an antecedent in the
context, or accommodated, in which case the content of the presupposition is added to the
context on the background of which the sentence is interpreted.
I next show how the process described above looks like in discourses containing the
two doch-variants under investigation.
3. The analysis
3.1 Accented middle-field doch
The first example is with accented doch in the middle field:
(12) Peter lügt DOCH.

'Peter is lying after all.'
The focus-annotated syntactic tree of this sentence is presented in (13). Here I assume
that doch is a sentence adverb, although it may be seen as attaching to the VP. The structure
in (13) is motivated by semantic considerations, since, semantically, doch modifies the entire
sentence. For the assignment of focus to the constituents in the syntactic structure I assume a
system like the one proposed in Riester (2005), where semantic-syntactic constraints are
defined by means of which syntactic constituents are marked as being part of the focus or the
background of the sentence (the sign ∼ is Rooth's focus interpretation operator, and C is a
variable that focus interpretation resolves or accommodates to a set of contextual alternatives,
see Rooth 1992).
(13)
In this context, focus is on doch itself, and this is a case where doch pertains to the
complement of the focus of the conjunct that hosts it. To choose the correct reading for doch
in this context from the ones specified in (11), a construction rule CR.dochADV can be
formulated stating that doch π introduces a presupposed speech act discourse referent π' and
the condition π': ¬ π[Fc(π)/Fc(π)].7 The application of construction rules is triggered by a
particular syntactic configuration, called triggering configuration. The triggering
configuration for CR.dochADV would be the structure in (14):
(14)
In the most recent version of DRT (Kamp, van Genabith and Reyle to appear), the first
step of the DRS construction is a preliminary sentence representation in which the
7 Strictly speaking, we deal here with the host including doch, which is syntactically integrated into the host
sentence and carries the focus, i.e. π corresponds to the top node S in the syntactic structure.

presuppositions of the sentence are explicitly represented. The second stage of the DRS
construction is the justification of the sentence presuppositions. In the DRT version that takes
the focus-background division of the sentence into consideration (Kamp 2004), focus
structure is represented as a triple <K0, K1, K2> consisting of a restrictor (a condition that
restricts the possible values of the focus variable), a focus frame (corresponding to the
background) and the focus constituent. I leave out the restrictor for simplicity in what
follows. In this framework, the representation of (12) will be the one in (15).8 As already
pointed out, the presupposition reduces here to the requirement that the context contains the
negation of π.
(15)
In (15), the left part between the angled brackets represents the focus frame, the right part the
focus constituent. The focus variables are set in boldface to indicate that they were obtained
by abstracting the focus marked constituents from the representation of the sentence thus
rendering the focus frame. In the focus part of the nonpresuppositional DRS, focus on doch is
represented as focus on the boolean value of the sentence. The latter is represented by the
variable B, and its value here is ASSERT, since doch asserts the sentence that hosts it without
having truth-functional effects on it.
The second stage of the DRS construction is the justification of the sentence
presuppositions. The presupposition of this use of doch may be either bound or
accommodated. I next consider the two possibilities in turn.
First, suppose the speaker or the interlocutor have suggested earlier that Peter was not
lying, i.e. Peter lügt nicht is already represented in the context. The left hand DRS in (16)
represents this sentence.9
8 I ignore here for simplicity the presupposition triggered by the proper name.
9 I assume for simplicity that only the negation is focussed, i.e. Peter lügt [nicht]F.

(16)
Here, both π' and π'' have the semantics ¬lügt(p). Processing the presupposition introduced by
doch will result in the representation in (17), where the discourse referent π' is resolved to the
antecedent π''. Then the non-presuppositional DRS is merged to the context.10
(17)
In the second case, the context does not entail ¬π, e.g. the context is empty, which is
represented in (18).
(18)
10 A problem here is that when merging the non-presuppositional DRSs to the context, it will contain two
contradicting conditions Peter lügt nicht and Peter lügt which render the DRS inconsistent. This is however an
issue that must be dealt with by a theory of corrections. Existing DRT-based theories of correction such as Maier
and van der Sandt (2003) and Asher and Lascarides (2003) suggest that the corrected material is deleted from the
context. This is however unintuitive since deleting corrected material from the context makes it as if it was never
introduced into the discourse. The revision happens at the level of the beliefs of the interlocutors, rather than at
the level of the context. A more adequate treatment of correction requires therefore a representation of the belief
states of the interlocutors and the changes that the belief states experience. This issue is however too complex to
be further pursued here.

Here, the presupposition is accommodated: the clause Peter lügt nicht is added to the context
part, and processing the presupposition introduced by doch results in the representation in
(19). Then the non-presuppositional DRS is merged to the context.11
(19)
3.2 Unaccented middle-field doch
The second example involves the unaccented modal particle doch as in (20).
(20) Peter LÜGT doch.
'Peter is lying, as you should know.'
The focus-annotated syntactic tree of the sentence is presented in (21).12
(21)
11 Here, the same problem as in (17) occurs, possibly also a violation of van der Sandt’s (1992) consistency
constraint for accommodating presupposed material.
12 Here it could be also argued that doch attaches to S since it semantically modifies its host. However, then it
would be difficult to distinguish this use of doch from the unaccented conjunction, i.e. both variants would have
the same triggering configuration. This issue requires more careful consideration and must be postponed for
future research.

A DRS construction rule for this doch-variant that matches the relevant part of the tree
can be formulated as follows: introduce a presupposed speech act discourse referent π' and
two conditions on it, (i) π': ¬ π [π/π] and (ii) BelS GIVEN(π).
The discourse representation of the sentence is constructed in the usual manner. The
preliminary sentence representation in (22) contains the presupposition introduced by this
variant of doch which is reduced to the condition that π' is the negation of π.13
(22)
Again, there are two cases that need to be considered when it comes to the justification
of the sentence presuppositions. In the first case, Peter lügt nicht is already represented in the
context, e.g. the interlocutor has suggested the opposite of what the speaker believes is given,
as in (3), represented as the context-part in (23).14
13 I ignore here for simplicity other presuppositions such that the one triggered by the proper name.
14 Actually, this is a simplification, since ¬π is usually not directly asserted but can be inferred from the context,
as in (9).

(23)
Then processing the presupposition introduced by doch will result in the
representation in (24), where the discourse referent π' is resolved to the antecedent π''.15
(24)
In the second case, the context does not entail ¬π, e.g. the context is empty:
(25)
15 Again, merging the non-presuppositional DRS to the context will raise the inconsistency problem that I
pointed out in the preceding section.

In this case, the presupposition is accommodated: the representation of the clause Peter lügt
nicht is added to the context part, and processing the presupposition introduced by doch will
result in the representation in (26).16
(26)
3.3. Adding discourse relations
In Karagjosova (2009), I argue that the representations of discourses with doch would be
more adequate if they contained explicitly relations like Concession(π2,π') or Correction(π,π'),
possibly introduced by means of additional conditions specified in the construction rules for
the respective doch-variants. I also discuss how this approach is different from a standard
SDRT account. I show that on my account, the semantics of the connector contributes to
specifying the relation as well as to finding the correct argument for it. In the cases of doch
considered here, doch not only introduces a discourse relation, but also identifies (after
binding the presupposition) or introduces (after accommodation) its second argument. For
instance in (18), repeated below as (27) with explicitly indicated discourse relation, the
second argument of the correction relation is introduced into the context only after
accommodating the presupposition of doch. The resulting representation is (28).
(27)
16 Also here the inconsistency problem remains.

(28)
3. Summary and conclusions
I presented a unified DRT-based account of the meaning and function of two uses of the
German particle doch, namely accented and unaccented doch in the middle field. The analysis
is an extension of a previous analysis of two other uses of doch (suggested in Karagjosova
2009) and shows how existing DRT formalisms can be used for modelling discourse
connectors like doch. It also hints at the limits of these formalisms and how their inventory
must be extended in order to be able to capture the meaning and discourse function of these
expressions in a full-fledged formal framework.
References:
Asher, N. and Lascarides, A. 2003, Logics of Conversation, Cambridge University Press.
Hentschel, E. 1986, Funktion und Geschichte deutscher Partikeln, Niemeyer, Tübingen.
Kamp, H. 2004, Information structure in a dynamic theory of meaning, in ‚Proceedings of the
Linguistic Society of Korea’.
Kamp, H. and Reyle, U. 1993, From Discourse to Logic, Kluwer Academic Publishers,
Dordrecht/Boston/London.
Kamp, H., van Genabith, J. and Reyle, U. to appear, Discourse representation theory, in D.
Gabbay, ed., ‚Handbook of philosophical logic’.
Karagjosova, E. 2004, The meaning and function of German modal particles, Ph.D. Thesis,
Saarabrücken Dissertations in Computational Linguistics and Language Technology.
Karagjosova, E. 2006, Correction and acceptance by contrastive focus, in D. Schlangen and
R. Fernandez, eds., ‚Proceedings of brandial 2006’, University of Potsdam, 26—33.
Karagjosova, E. 2008, Contrast and underspecification, in A. Grønn, ed., ‚Proceedings of
SuB12’, 287—302.

Karagjosova, E. 2009, Towards a DRT-based account of adversative connectors, in A. Riester
and E. Onea, eds., ‚Focus at the syntax-semantics interface’, SinSpec 03, Working papers
of the SFB 732 „Incremental specification in context“, Stuttgart University, 35—52.
Lang, E. 1977, Semantik der koordinativen Verknüpfungen, Akademie-Verlag, Berlin.
Maier, E. and van der Sandt, R. 2003, Denial and correction in layered DRT, in I. Kruijff-
Korbayova and C. Kosny, eds, ‚Proceedings of DiaBruck’, Saarland Universtity.
Reiter, N. 1980, Die Perfidie des deutschen ja, Deutsche Sprache (4), 342—355.
Reyle, U., Rossdeutscher, A. and Kamp, H. 2007, Ups and downs in the theory of temporal
reference, Linguistics and Philosophy (30), 565—635.
Riester, A. 2005, Context and focus projection. A compositional, intonation-based account of
focus interpretation, in E. Maier, C. Bary and J. Huitink, eds, ‚Proceedings of SuB9’,
299—313.
Rooth, M. 1992, A theory of focus interpretation, Natural Language Semantics (1), 75—116.
Sæbø, K. J. 2003, Presupposition and contrast, in Weisgerber, ed., ‚Proceedings of SuB 7’,
257—271.
Umbach, C. 2005, Contrast and information structure: A focus-based analysis of but,
Linguistics 43(1), 207—232.
van der Sandt, R. 1992, Presupposition projection as anaphora resolution, Journal of
Semantics (9), 333—377.
Wassermann, R. 2000, Resource-bounded belief revision, Ph.D. thesis, Amsterdam
University.