Content uploaded by Elena Karagjosova
Author content
All content in this area was uploaded by Elena Karagjosova on Mar 26, 2020
Content may be subject to copyright.
German doch as a marker of given information
Elena Karagjosova
Computational linguistics, Saarland University
elka@coli.uni-sb.de
Abstract
This paper investigates two uses of the German discourse particle
doch,namelythecasesinwhichitoccursintheMiddlefiledeitherac-
cented or deaccented. We argue that none of these uses marks given
information or common knowledge per se. More closely, we suggest that
deaccented doch does not mark given information per se but as a result
of the association of the whole proposition in its scope with the topic
or Theme part of the sentence, and that accented doch restricts the al-
ternatives evoked by the default sentence focus. We also argue that the
common semantic denominator uniting the two uses of doch is the notion
of contrast which is differently manifested due to the prosodic difference:
evoking an alternative proposition in the case of accented doch and a con-
trast between the givenness of the proposition in the scope ofunaccented
doch and its assumed or evident nonsalience.1
1Introduction
It has been recognised that in natural language discourse, agentsuseredun-
dant information because they are aware of their resource bounds, namely their
limited attention and inferential capacity and their autonomy2(Walker 1993).
Providing information that is familiar to the hearer is thus a strategy that con-
tributes to successful communication given the resource bounds of human agents
and minimises the processing effort (ibid.). Especially in tutorial dialogue, pro-
viding information that is not new to the hearer may have various important
learning effects. For instance, using given information helps the dialogue par-
ticipants remember it (rehearsal effect), activates related concepts and supports
reasoning by making a known concept salient (Walker 1993). These effects are
of interest for the development of intelligent tutoring systems whichcanbeen-
abled to generate appropriately redundant information in order to boost the
performance of the student.
1This work is supported by the SFB 378 at Saarland University, Saarbr¨ucken. Thanks to
an anonymous reviewer for valuable suggestions.
2That is, the need to provide evidence supporting beliefs about mutual understanding and
acceptance.
1
However, these effects will not be fully achieved if the hearer does not recog-
nise the given information as such. For instance in the tutoring domain, marking
information as given can help a student recognise the intention of thetutornot
to provide new information but to activate knowledge that the student already
possesses and to elicit from the student related knowledge.
The work presented in the paper is ongoing research set within the frame-
work of the project DIALOG (Pinkal et al. 2001) aiming at the development of
a natural language dialogue interface for an automatic tutoring system in the
mathematical domain. In the project, a corpus of German tutorial dialogues
has been collected in a Wizard-of-Oz experiment environment (Benzm¨uller et al.
2003). Preliminary work on this corpus suggests that different linguistic means
will be used depending on the type of the given information, that is, whether it
repeats or paraphrases the antecedent3or whether it makes an inference from
the antecedent explicit. Walker 1993 provides a classification of Information-
ally Redundant Utterance (IRUs) based on the type of given information they
represent and on their communicative function. Walker does not address the
question of whether and how IRUs are marked as given. However, as already
pointed out, preliminary corpus investigations have indicated that there may be
correlations between types of IRUs and linguistic means for marking them.
This paper explores the use of the German discourse particle doch as a
linguistic means for marking such given information. It has been argued that
doch when used as modal particle indicates that the proposition in its scope is
assumed by the speaker to be common knowledge of the dialogue participants
(Karagjosova 2001) or that it marks the relation between the proposition and
the content as being old (Zeevat 2002). In the aforementioned corpus, given
information is marked by using discourse particles (henceforth DPs) like ja and
eben, as well as by using expressions like wie gesagt (as already said), past tense
etc. Although no occurrences of doch as a modal particle were found in the
corpus, we found several occurrences of its accentuated counterpart which in
some contexts seems to have a similar function of marking informationthathas
been previously considered by the dialogue participants. An investigation of the
difference between these uses of doch seems sensible not only with regard to the
teaching effects that markers of given information may have but also regarding
the still open question of the overall meaning of this lexeme.
The paper is organised as follows. In Section 2, we address the question of
what kind of given information can be associated with the two uses of do ch
we explore, namely doch as unaccented modal particle and doch as accented
adverb that occurs in the Middle Field of the German sentence. We will use
the taxonomy of assumed familiarity proposed in (Prince 1981) and used in
(Walker 1993) to determine this. Another question that is addressed in this
section concerns the type of IRUs the two uses of doch mark according to the
communicative function of the IRU as specified in (Walker 1993). It will come
out of the analysis that although both doch and doch can occur in utterances
3Antecedents are propositions that originally added the propositional content of the re-
dundant information to the discourse situation (Walker 1993).
2
used as IRUs, only doch can be said to mark given information. Moreover, we
will argue that doch does not mark given information per se but as a result of
the association of the whole proposition in its scope with the topic or Theme
part of the sentence due to the fact that it is deaccented. On the other hand,
doch does not mark given information per se either but simply restricts the
alternatives evoked by the default sentence focus4(Section 3). An implication
of these claims is that the common semantic denominator uniting the two uses
of doch is the notion of contrast which is differently manifested due to the
different prosodic realisation: evoking an alternative proposition in the case of
doch and a contrast between the givenness of the proposition in the scope of
doch and its assumed or evident nonsalience.
2Dochandthegiven-newdistinction
In this section we explore the question of what kinds of given information can
be associated with the two uses of the German DP doch. It is often pointed
out that deaccented doch marks old information or shared knowledge of the
dialogue participants (Jacobs 1991; Karagjosova 2001; Zeevat 2002). On the
other hand, doch carries contrastive accent (Abraham 1991), and contrastive
accents can also sometimes mark given information5(Gundel 1998).
Prince 1981 proposes a taxonomy of assumed familiarity according to which
entities in a discourse model can be assumed by the speaker to have a different
familiarity status:
•new:canbebrand-new6(i.e. the hearer must create a new discourse entity)
or unused (i.e. the hearer is assumed to have a corresponding entity in his own
model and simply place it in the discourse model)
•(textually or situationally) evoked:representdiscourseentitiesthatwereonce
new or inferable or that are present in the discourse situation
•inferables7are entities the speaker assumes the hearer can infer via reasoning
from discourse entities already evoked
Walker applies this taxonomy to complex discourse entities like propositions
(Walker 1993). She bases her classification of IRUs mainly on contextual pa-
rameters like hearer old status (based on Princes taxonomy), salience status,
speaker parameters and prosodic realisation. Walker distinguishes three kinds
of IRUs having different communicative functions:
•Attitude IRUs whose function is to demonstrate the hearer’s attitude to an
utterance just contributed by a speaker to the discourse situation
•Attention IRUs whose function is to make a proposition salient (when it is not
salient)
4In the terminology of Vallduvi and Vilkuna 1998.
5The entities carrying contrastive accent are called then contrastive topics.
6These break down further in anchored and unanchored.
7Can be noncontaining or containing.
3
•Consequence IRUs which support inferential processes by making an inference
from their antecedent explicit or by supporting inferentialrelationstoother
salient propositions
According to this classification, doch-utterances function mainly as Atten-
tion IRUs and cannot be used as Attitude IRUs, whereas doch-utterances can
function as Attitude IRUs but not as Attention IRUs. Thus, an Attention IRU8
is presented in (1) below where the speaker reminds the hearer on the national-
ity of Ramesh in order to provide support for the acceptance of a good Indian
restaurant he suggests. We notice that doch cannot be used in this context, cf.
(1b.), whereas doch can, cf. (1c.).
(1) a. Listen to Ramesh. He is Indian.
b. H¨ore auf Ramesh. Er ist #doch Inder.
c. H¨ore auf Ramesh. Er ist doch Inder.
On the other hand, doch can be used in an Attitude IRU that demonstrates
understanding of what the previous speaker has just said, cf. (2), whereas doch
is not appropriate in this situation.
(2) A: Ich komme mit. (I am coming along.)
B: Du kommst doch mit. (You are coming after all.)
B’: Du kommst #doch mit.
This suggests that the proposition in the scope of doch can be evoked,
inferable or unused,butnotbrand-new, since Attention IRUs make an
already given proposition salient, whereas the proposition in the scope of doch
can be evoked, inferable or brand-new,butnotunused. This observation
is also supported by a corpus investigation we carried out using the Baufix
corpus.9These restrictions with respect to the assumed familiarity status of
the proposition in the scope of the two uses of the DP suggests that doch does
not mark the proposition in its scope as given, whereas doch seems to do so.
In the case of doch, the speaker makes a proposition salient which he believes
the hearer is already familiar with but may have forgotten, i.e. the proposition
is assumed not to be salient or active in the mind of the hearer. On the other
hand, doch seems to presuppose that the negation of the proposition in its
scope has been previously considered, i.e. that the negation of the proposition
is evoked or inferable. This however does not allow the proposition in the
scope of doch to qualify as given in the sense that the earlier consideration of
the opposite is revised and does not hold anymore: it is rather the result of a
revision of a previous belief of the speaker or hearer (or both) or the result of
8This IRU is of the subclass Deliberation IRU since it makes a proposition salient in order
to support deliberation about whether to accept or reject other currently salient propositions,
(Walker 1993).
9http://www.sfb360.uni-bielefeld.de/teilpro j/b1.html.Wedonotprovideexamplesdueto
space constraints.
4
eliminating one of two diametrical possibilities.10 In both cases the proposition
in the scope of doch cannot be assumed to be shared without it being evoked
in the discourse, i.e. it cannot be unused.
Furthermore, as it is often pointed out, doch expresses contrast (Jacobs
1991; Karagjosova 2001).11 Thus, doch is said to presuppose the negation of the
proposition in its scope and that this proposition should be part of thecontext
(Zeevat 1999), i.e. the proposition should be given in the sense of evoked.
However, our corpus study has shown that with respect to its presupposed
negation, the proposition in the scope of doch can be evoked, inferable,
brand-new or unused. This means that the presupposition triggered by doch
can be either satisfied by the context or accommodated.12 In other words, doch
seems to indicate that a belief revision has been carried out, by making salient
an earlier consideration of the negation of the proposition in its scope.
In the case of doch, the contrast is not at the propositional level as argued
in (Karagjosova 2001), which is why it is often felt to be very weak or nonexis-
tent. The observation that doch-utterances are used mainly as Attention IRUs
suggests that doch can be said to express a contrast between the assumption of
the speaker that the proposition in the scope of doch is given, i.e. known by
the hearer, and the evidence or assumption that this knowledge is not salient
anymore. In other words, doch seems to indicate that a proposition that is
believed by the speaker to be given is nonsalient.
3Theinformationstructureofthedoch-sentence
So far we have argued that both uses of doch we consider mark contrast, and
that doch marks given information, whereas doch doesn’t. In this section
we address the question of whether marking given information is an inherent
property of doch as opposed to doch and other uses of this DP like response
particle and adversative conjunction.
Prince 1981 points out that the given-new distinction concerns the tailoring
of the utterance by a speaker to meet the particular assumed needs of the in-
tended receiver and that information-packaging reflects the sender’s hypothesis
about the receiver’s assumptions, beliefs and strategies.
In the previous section, we argued that doch indicates that the proposition
in its scope is the result of a revision/correction where the opposite was pre-
viously believed/considered, and that the proposition in the scope of doch is
assumed to be given (unused, evoked) but not salient.
If a proposition is the result of a revision, then it must be considered new for
10See also (Stede and Schmitz 2000) who talk about the property of doch to mark revision.
doch can be also used to mark a correction of the previous speaker, e.g. ’A: Maria ist verreist.
B: Nein, sie ist doch nicht verreist.’, where the doch-utterance suggests e.g. ’Yes, she wanted
to leave, but she changed her mind’.
11Some authors (Helbig 1988; Zeevat 2002) argue that doch does not mark contrast in all
contexts in which it can be used.
12This is compatible with Umbach’s analysis of but (Umbach 2001).
5
the discourse.13 Consequently, it can be seen to belong to the sentence Rheme.
And if a proposition is given, then it must belong to the topic or Theme ofthe
sentence.
By virtue of carrying contrastive accent, doch evokes an alternative which
is an alternative of the whole proposition, namely its opposite. If the sentence in
the scope of doch has a second, noncontrastive pitch accent, then the function
of doch seems to be moreover to restrict the set of alternatives evoked bythe
entity carrying the second pitch accent. Consider (3).
(3) A: Wer hat die Party organisiert? (Who has organised the party?)
B: Es war Bill. (It was Bill.)
B’: Es war doch Bill.
B”: Es war doch Bill.
The topic of the three B utterances is x:x hat die Party organisiert.How-
ever, B” answers a question different from the overt question, namely not Who
has organised the party? but Has Bill organised the party?. This is what Um-
bach 2001 calls quaestio: an a posteriori reconstructed question addressed by
the utterance.14 This suggests that doch in (3B”) serves next to evoking an
alternative proposition also to narrow down the set of alternatives evoked by the
default sentence focus.15 . In other words, we can conclude that the proposition
in the scope of doch will be considered to belong to the Rheme, and when a
constituent besides doch carries a pitch accent, it will be interpreted as Rheme-
Focus (in the terminology of Steedman 2000), unless the doch-utterance is used
as an IRU.
In the case of doch, due to its deaccentuation the whole proposition in its
scope is associated with the topic or Theme of the sentence. With doch,the
quaestio is the same as the overt question. Since it does not carry a contrastive
accent, no alternative proposition is evoked. In case the doch-sentence involves
a constituent carrying contrastive stress, it is to be interpreted as contrastive
topic.
The above considerations suggest that the overall meaning of the discourse
particle doch as used in the Middle field can be specified in terms of contrast
of different type, depending on the prosodic realisation of the DP, and that
the apparent additional meaning component of the modal particle doch related
to the shared knowledge of the dialogue participants can be seen as an effect
of its prosodic realisation. This view brings us closer to solving the taskof
finding a common denominator of all four uses of the DP doch (modal particle,
adverb, conjunction and response particle) most of which do not involve a shared
knowledge assumption but indicate contrast of some kind.
13Unless the utterance is an Attitude IRU.
14However, according to Umbach, the quaestio should entail theovertquestion,whichdoes
not seem to be the case in this example.
15According to the terminology of Vallduvi and Vilkuna 1998
6
4Conclusionsandfuturework
In this paper, we investigated two uses of the German discourse particle do ch,
namely the cases in which it occurs in the Middle filed either accented or deac-
cented. We argued that none of them marks given information or common
knowledge per se. We suggested that deaccented doch does not mark given
information per se but as a result of the association of the whole proposition in
its scope with the topic or Theme part of the sentence due to the fact that it is
deaccented. We argued that accented doch restricts the alternatives evoked by
the default sentence focus. We also argued that the common semantic denom-
inator uniting the two uses of doch is the notion of contrast which is differ-
ently manifested due to the different prosodic realisation: evoking an alternative
proposition in the case of accented doch and a contrast between the givenness
of the proposition in the scope of unaccented doch and its assumed or evident
nonsalience. This is compatible with the intuitions about the other two uses
of doch, namely as response particle and conjunction, which also express con-
trast and do not involve a common knowledge assumption by the speaker. The
approach is also compatible with analyses of related contrastive Russian modal
particles (McCoy 2001) and of the conjunction but (Umbach 2001), which also
consider the information structure of the sentence.
More work is needed to support the claims made in this paper. It is also
to be explored how the doch-proposition, which we argued to belong to the
Theme of the sentence, fits into the overall thematic structure of the discourse.
Nevertheless, the analysis of DPs in terms of their relation to the information
structure of the sentence seems to be a fruitful approach with respect to un-
derstanding the nature and specifying the (overall) meaning of DPs which tend
to have different interpretations depending on the context of their use, such as
auch and ja which also have accented and deaccented variants.
The future work on the more general issues addressed in the paperconcerns
the investigation of further means for marking given information andtheways
of formalising and implementing them in an intelligent tutorial system.
References
Abraham, W. (1991). Discourse particles in German: How does their illocutive
force come about? In W. Abraham (Ed.), Discourse particles in German,pp.
203–252. Amsterdam/Philadelphia: John Benjamins.
Benzm¨uller, C., A. Fiedler, M. Gabsdil, H. Horacek, I. Kruijff-Korbayova, M. Pinkal,
J. Siekmann, D. Tsovaltzi, B. Q. Vo, and M. Wolska (2003). A wizard of oz
experiment for tutorial dialogues in mathematics. In AIED.
Gundel, J. (1998). On different kinds of focus. In P. Bosch and R. van der Sandt
(Eds.), Focus i n N a t u ra l Lang u a g e P roce ssing.CUP.
Helbig, G. (1988). Lexiko n deutsch er Parti keln.Leipzig:VerlagEnzyklop¨adie.
Jacobs, J. (1991). On the semantics of modal particles. In W. Abraham (Ed.),
Discourse particles,pp.141–162.JohnBenjamins.
7
Karagjosova, E. (2001). Towards a comprehensive meaning of German doch. In
Proceedings of the ESSLLI 2001 Student Session,Helsinki.
McCoy, S. (2001). Colloquial Russian Particles -TO, ZHE, and VED’ as Set-
Generating (Kontrastive) Markers: A Unifying Analysis. Ph. D. thesis, Boston
University.
Pinkal, M., J. Siekmann, and C. Benzm¨uller (2001). Pro jektantrag Teilprojekt MI3
—DIALOG:TutoriellerDialogmiteinemmathematischenAssistenzsystem. In
For t s e tzungs a n t ra g S F B 378 — Re s s ource n a d a ptive k og n i t v e Prozesse.Univer-
sit¨at des Saarlandes.
Prince, E. (1981). Toward a taxonomy of given-new information. In P. Cole (Ed.),
Radical Pra gmati cs.NewYork.
Stede, M. and B. Schmitz (2000). Discourse particles and discourse functions. Ma-
chine Translation.
Steedman, M. (2000). Information structure and the syntax-phonology interface. A
revised version is to appear in Linguistic Inquiry, 31.4.
Umbach, C. (2001). Contrast and contrastive topic. In Proceedings of the ESSLLI
2001 Workshop on Information Structure, Discourse Structure and Di scou rse
Semantics,Helsinki.
Val lduvi, E. a n d M. Vilk u n a (199 8 ). On rhem e and ko ntras t . Syntax and Seman-
tics (29), 79–108.
Wal ker, M. (199 3 ) . Informational redundancy and resource bounds in dialogue.Ph.
D. thesis, University of Pennsylvania.
Zeevat, H. (1999). Explaining presupposition triggers. Submitted to ESSLI’99 vol-
ume.
Zeevat, H. (2002). Particels: presupposition triggers or context markers. forthcom-
ing.
8
