Академический Документы
Профессиональный Документы
Культура Документы
W
DE
G
Topics in English Linguistics
33
Editors
Bernd Kortmann
Elizabeth Closs Traugott
Mouton de Gruyter
Berlin · New York
Cause - Condition - Concession - Contrast
Cognitive and Discourse Perspectives
Edited by
Elizabeth Couper-Kuhlen
Bernd Kortmann
W
DE Mouton de Gruyter
G Berlin' New York 2000
Mouton de Gruyter (formerly Mouton, The Hague)
is a Division of Walter de Gruyter GmbH & Co. KG, Berlin.
© Copyright 2000 by Walter de Gruyter GmbH & Co. KG, 0-10785 Berlin
All rights reserv~d, including those of translation into foreign languages. No part of this
book may be reproduced or transmitted in any form or by any means, electronic or
mechanical, including photocopy, recording, or any information storage and retrieval
system, without permission in writing from the publisher.
Printing: Werner Hildebrand, Berlin - Binding: Liideritz &. Bauer GmbH, Berlin.
Cover design: Christopher Schneider, Berlin.
Printed in Germany.
Contents
Introduction 1
Elizabeth Couper-Kuhlen and Bernd Kortmann
I. Cause
II. Condition
III. Contrast
IV. Concession
Index 469
Despite the division of our collected papers into sections, there are-
as might be expected-recurrent themes which cut across the C-relations.
One of the motifs in the cognitively oriented papers, for instance, is
Sweetser's (1990) seminal work on domains or levels of interpretation.
Noordmanlde Blijzer (this volume) adopt her distinction between content
and epistemic level and show how it will help account for differing
degrees of cognitive complexity in causal sentences. Other papers are
more critical of Sweetser's model: Lang (this volume) takes issue with
her claim that content, epistemic and speech-act levels of interpretation
can be determined independently of syntactic structure and, like Crevels
(this volume), fmds it necessary to supplement Sweetser's three domains
with a fourth, textual level. Given a fourth level, Crevels fmds the model
useful for investigating the formal means which typologically different
languages deploy in each domain. Pander Maat/Sanders (this volume), on
the other hand, reject Sweetser's model as a means of accounting for the
distribution and interpretation of specific causal markers in Dutch and
propose in its stead a notion of subjectivity.
Another recurrent motif in several of the chapters collected here is
Fauconnier's theory of Mental Spaces (1985, 1997). Dancygier/Sweetser
(this volume) show that important distinctions can be drawn based on the
ways in which causal and conditional conjunctions participate in the
configuration of mental spaces. Verhagen (this volume) uses a mental-
space model to account, e.g., for the relationship between concession and
(negated) causality, with which Konig/Siemund (this volume) are also
concerned. Yet AkatsukaiStrauss (this volume) are critical of a mental-
space account, because it neglects what they see as an inherent dimension
of counterfactual conditionality, namely that speakers are expressing a
stance towards the events in question as desirable or undesirable.
AkatsukaiStrauss advocate a subjective dimension in the analysis of
counterfactual conditionality and in this sense are in agreement with
Pander Maat/Sanders, who see subjectivity as responsible for the
behavior of specific causal connectives in Dutch discourse. It is perhaps
worth noting that both AkatsukaiStrauss and Pander Maat/Sanders take a
corpus-based approach to C-constructions and for this reason belong
more in the discourse than the cognitive tradition. Significantly, it is in
relying on real data rather than on introspection that these two studies
independently establish a need-above and beyond mental-space or
domain considerations-for taking the speaker's (or subject of conscious-
ness's) stance into account in the analysis of C-constructions, especially
for counterfactual conditionality and causality. In this sense they are
initiating the kind of dialogue between cognitive and discourse
4 Elizabeth Couper-Kuhlen and Bernd Kortmann
approaches to the C-relations which the editors have envisaged (cf. also
the chapters by Meyer and Konig/Siemund in this respect).
A second notion to be found in both cognitive and discourse-oriented
papers collected here is polyphony (Verhagen), or multiple viewpoints
(Schwenter). For Verhagen, the interpretation of concessive and negated
causal sentences as well as of all sentences involving epistemic C-
relations calls for the construction of two nearly similar mental spaces in
Fauconnier's theory-or, equivalently, for the assumption of polyphony in
Ducrot's sense of the term (1984, 1996). Schwenter also mentions
Ducrot's theory of polyphony as one way of conceptualizing viewpoints
and linguistic structure. He opts, however, for Roulet's (1984) distinction
between one viewpoint (monological) and two viewpoints (dialogical),
both of which can be found with either one speaker or two. Schwenter
finds that dialogicality is not only conceptually present in adversative
constructions but also (to a lesser degree) in 'exhortative' and epistemic
causals (i.e., ones which accompany exhortative speech acts and
inferential conclusions). The notion of polyphony or multiple viewpoints
then appears to be relevant to both cognitive and discourse-based
understandings of the C-relations. And it is clearly implicated in the
interactional model of concession which Couper-KuhlenfThompson (this
volume) propose, where multiple viewpoints often go hand in hand with
multiple speakers-although, as Barth (this volume) shows, they need not
do so.
Related to the dimension of multiple speakers is of course the question
of whether a piece of discourse-specifically, a C-construction-is
spoken or written. This is a recurrent motif in several of the discourse-
oriented papers collected here. Aside from questions of overall frequency
of use (see, for instance, Altenberg 1984, 1986 or Ford 1993), medial
considerations are relevant for C-constructions in at least two other ways.
One of these is with respect to the order of clauses, which Auer's chapter
in this volume investigates with respect to wenn-clauses in German (cf.
English 'when'/'if -clauses). A second issue relating to C-relations in
spoken and written discourse is the question of whether mediality has an
influence on the choice of C-marker. Montolio (this volume) argues that,
as in English, complex conditional constructions in Spanish such as a
condici6n de que ('provided that') and a menos que ('unless') express a
more restricted conditional relation than si 'if and are reserved for formal
written registers. And Barth (this volume) shows that the set of
concessive markers as well as the relative frequency of parataxis as
opposed to hypotaxis is significantly different when spoken English is
compared to written English.
Introduction 5
References
Altenberg, Bengt
1984 Causal linking in spoken and written English. Studia Linguistica
38: 20-69.
Altenberg, Bengt
1986 Contrastive linking in spoken and written English. In: G. Tottie and
I. Baecklund (eds.), English in Speech and Writing. A Symposium,
13-40. Stockholm: Almqvist & Wiksell.
Ducrot, Oswald
1984 Le dire et Ie dit. Paris: Minuit.
Ducrot, Oswald
1996 Slovenian lectures/Conferences slovenes: Argumentative seman-
tics/Semantique argumentative. Ljubljana, ISH.
Fauconnier, Gilles
1985 Mental Spaces: Roles and strategies. Cambridge, MA: MIT Press.
Fauconnier, Gilles
1997 Cognitive Mappings for Language and Thought. Cambridge: Cam-
bridge University Press.
Ford, Cecilia E.
1993 Grammar in Interaction. Cambridge: Cambridge University Press.
Gohl, Christine and Susanne Gtinthner
1999 Grammatikalisierung von 'wei!' als Diskursmarker in der gespro-
chenen Sprache. Zeitschriftfur Sprachwissenschaft 18: 39-75.
Halliday, M. A. K.
1985 An Introduction to Functional Grammar. London: Edward Arnold.
Harris, Martin
1988 Concessive clauses in English and Romance. In: 1. Haiman and S.
A. Thompson (eds.), Clause Combining in Grammar and Dis-
course, 71-99. Amsterdam: Benjamins.
Hopper, Paul and Elizabeth C. Traugott
1993 Grammaticalization. Cambridge: Cambridge University Press.
8 Elizabeth Couper-Kuhlen and Bernd Kortmann
Konig, Ekkehard
1986 Conditionals, concessive conditionals and concessives: Areas of
contrast, overlap and neutralization. In: E. C. Traugott, A. ter
Meulen, 1. S. Reilly and C. Ferguson (eds.), On Conditionals, 229-
246. Cambridge: Cambridge University Press.
Konig, Ekkehard
1988 Concessive connectives and concessive sentences: Crosslinguistic
regularities and pragmatic principles. In: 1. A. Hawkins (ed.),
Explaining Language Universals, 145-166. New York: Basil
Blackwell.
Kortmann, Bernd
1997 Adverbial Subordination. A typology and history of adverbial sub-
ordinators based on European languages. Berlin: Mouton de
Gruyter.
Lenk, Uta
1997 Marking Discourse Coherence. Functions of discourse markers in
spoken English. Tiibingen: Narr.
Roulet, Eddy
1984 Speech acts, discourse structure, and pragmatic connectives.
Journal ofPragmatics 8: 31-47.
Rudolph, Elizabeth
1996 Contrast: Adversative and concessive relations and their
expressions in English, German, Spanish, Portuguese on sentence
and text level. Berlin: de Gruyter.
Sweetser, Eve
1989 From Etymology to Pragmatics. Metaphorical and cultural aspects
ofsemantic structure. Cambridge: Cambridge University Press.
Thompson, Sandra A. and Robert E. Longacre
1985 Adverbial clauses. In: T. Shopen (ed.), Language Typology and
Syntactic Description, Vol. 2 Complex constructions, 171-234.
Cambridge: Cambridge University Press.
The relevance of causality-
ity seems to have its locus in the social sciences and humanities. Every-
day action, group cohesion on all levels, economic transactions of all
kinds, notions of obligation and responsibility, law as the formalisation
of the latter notions, and many other topics of social sciences and hu-
manities crucially depend on causality. We may thus take the above
examples as a first approximation to a concept of causality that is lo-
cated in a social rather than in a physical world. In other words, the uni-
verse of which causality is the cement, the world in which causality is
the most fundamental fact, is primarily a socially constructed universe, a
socially constituted world.
This article does not undertake to solve the problem of causality. Its
objective is much more modest. Based on observations made in the phi-
losophical, scientific, psychological, social scientific, and linguistic lit-
erature, I will defend a set of linguistically inspired theses concerning
the relationship between causality and relevance (cf Meyer 1983: 125-
126), putting it into the context of the present discussion and of an
emerging research programme.
I will begin with a (very) brief history of the concept of causality
(Section 1). I will then show how the history of the concept, the way it
has been discussed in the literature, and the problems it has raised, seem
to indicate a certain constraint on the notion of cause that is best de-
scribed by a notion of relevance (Section 2). Next I will ask whether this
notion of relevance is explicable in terms of the presently most influen-
tial pragmatic theory of relevance (Sperber and Wilson 1995 2) (Section
3), and discuss problems arising from such an attempt. Finally, I will
argue for a basically discursive explication of the notion of causality in
the framework of a coherence relations approach, establishing causality
as one of several primitive universal principles that govern the construc-
tion of coherence in discourse (Section 4).
A man, say, falls down a flight of stone steps as he begins the descent. The
steps are slippery and the man's mind is elsewhere. This is a sufficient ex-
planation of his fall. But of course not every preoccupied man falls down
every flight of slippery steps he descends. There is absolutely no question of
our formulating or envisaging exceptionless laws, ... , to cover all such
cases. (Strawson 1985: 131)
That is, all rational and enlightened people will assume that the
man's falling down the stairs is fully in accordance with some presumed
laws of nature and, in principle, fully explainable in such terms. And
yet, in ordinary language we do not expect such full "scientific" ac-
counts to be given, and indeed it is highly doubtful whether such full
causal accounts of singular events are useful, desirable or, for that mat-
ter, feasible.
For all practical purposes, we are content with much less than a list-
ing of all the conditions, regularities, probabilities, and laws of nature
that account for the event in question. A satisfactory explanation, in
fact, is not one that lists all this. A satisfactory explanation is concise
and concentrates on one or just a few causes. The problem that scientists
and philosophers have had with the notion of causality emerging from
this observation is that they cannot fmd a unique logical or mechanical
or otherwise physical characterisation of a state of affairs that happens
to be intuitively singled out as the "cause" of a certain event. Whatever
is named as "the cause" of something, does not seem to differ in any
significant way from other factors present in the situation in question.
There simply seems to be no logical or scientific basis for calling some-
thing a cause.
Logical analyses of the notion of cause have operated with necessary
and sufficient conditions. The two intuitively most plausible and (pre-
sumably for that reason) most influential defmitions have been Mackie's
(1965) "INUS condition" and the "counterfactual conditional", origi-
nally suggested by Hume and reintroduced into the modern debate by
The relevance ofcausality 13
The problems with both these defmitions are manifold (Sosa 1975: 3-
4), but it is not my purpose to discuss them here. There is just one point
that I want to make: Mackie refers to a set of conditions that is unneces-
sary, but sufficient for the effect. The factor singled out as "the" cause is
not the whole set of conditions, but only an insufficient but necessary
part of this set. The problem that this definition raises is that it is not
very useful for an application of the concept of causality in the analysis
of everyday communication. The definition covers a number of condi-
tions that we would not call causes in a given situation. That is, Mackie
describes a necessary but not sufficient condition for calling an event a
cause. He describes a semantic precondition for a proposition to serve as
a causal explanation. Real events always have a large number of neces-
sary and sufficient conditions. Arguing that one circumstance was the
cause of a given event means foregrounding one insufficient and neces-
sary condition and stating it, and allocating all the other conditions that
jointly were unnecessary but sufficient to bring about the effect to the
background. Thus, the very problem of Mackie's definition is an indica-
tion of the basically pragmatic character of any reasonable notion of
causality that would be applicable in the analysis of discourse.
It is a situation that logical positivists and context-free semanticists
cannot easily cope with, but it is not unfamiliar to pragmatically ori-
ented linguists: reality is always understated and underdetermined by
the propositional content of utterances. The meaning of utterances al-
ways goes beyond what was said. There is always a set of meanings that
were meant but left unsaid. And there is always a background of knowl-
edge that is taken for granted, that is not problematised, that is simply
not at issue. If we believe Searle (1983: 141), this background contains
infinitely many propositions. And indeed, it is precisely this characteris-
tic of the set of sufficient conditions for a certain event that has led early
modem philosophers (such as Hume) to confine the notion of causality
to "lawful" instances and has driven 19th and 20th century scientists and
philosophers to despair.
Let me try to illustrate this with Strawson's example (in fact, any exam-
ple of a causal relation would do). What is the set of conditions that is
jointly sufficient (though unnecessary) to bring about "the man's" fall
(for simplicity's sake let's call him Jones)? The two conditions men-
The relevance ofcausality 15
tioned by Strawson are in the foreground, of course: the steps were slip-
pery, and Jones was absent-minded. Apart from these, I have already
mentioned one further condition: the fact that Jones did try to descend
the stairs at the very time when the first two conditions were already
given. It seems that these three conditions, each of them independent of
the two others, were jointly sufficient for Jones' fall. But, as Strawson
(1985: 131) pointed out, "not every preoccupied man falls down every
flight of slippery steps he descends." So there must have been further
conditions that were necessary to make the set of conditions sufficient
for Jones' fall to occur: maybe if Jones had been wearing different shoes
(say, with non-slip rubber soles), he would not have fallen. Or, if there
had been a railing for him to hold on to, he would not have fallen. Or, if
his ankles had not been so stiff that day, he would not have fallen. Or,
perhaps, if he hadn't been drinking, in addition to his absentmindedness.
Or ...
On the other hand, we might also begin to question the most obvious
and plausible causes that were given: granted that the steps were slip-
pery, were they slippery enough for a man to slip? Granted that Jones
was absent-minded that day, was he absent-minded enough not to notice
the slippery state of the steps? That is, even the most readily accepted
explanations for an event are not undebatable. There is always a possi-
bility that the most plausible explanation is erroneous. Suppose that
Jones is absent-minded most of the time and the steps are always slip-
pery and he walks down these steps several times a day without slip-
ping. It is obvious that in this case we would still accept the explana-
tions given by Strawson to a certain extent, although we would have
evidence for the fact that the explanation is not sufficient. Even though
we would be willing to grant that there must have been other "causes"
for Jones' fall, the explanations "slippery steps" and "Jones' absent-
mindedness" are just too good to be discarded.
What this discussion is supposed to show is: as scientists and phi-
losophers found out long ago, there is no such thing as "the" cause or
"set of causes" of an event. Causality emerges as a discursive phenome-
non. Note that all the "causes" mentioned above could be adduced in
some kind of dispute over the "causes" of Jones' fall. But it depends
precisely on the kind of dispute as to which of the causes will be ad-
duced by whom.
Suppose Jones' wife had always been of the opinion that that stair-
case is dangerous and had always been warning Jones to use it (suppos-
ing there was an alternative available). In this case, the cause of Jones'
16 Paul Georg Meyer
fall that would be accepted by his wife would be the fact that Jones de-
scended the stairs in the first place. ("Did you have to use that danger-
ous flight of stairs? I told you one day you'd break your neck on it".) Or
that Jones descended the stairs while they were slippery ("Why couldn't
you wait until the steps had dried?") (Supposing, e.g., that they were
slippery because Jones' wife had just scrubbed them).3 For his wife, in
that case, the slippery state of the stairs would be taken for granted, just
like her husband's absentmindedness, which she has learnt to put up
with in long years of marriage.
Now take the case of an insurance company. Depending, of course,
on what kind of risk had been the object of the insurance, all kinds of
questions might be raised. Who was responsible for the slippery state of
the steps? Why did the flight of stairs have no railing? Was Mr. Jones'
absent-mindedness due to alcohol or drugs? What was the purpose of
his descent down the stairs? Was it on his way to work, for example?
Was Mr. Jones' stiflhess in the ankles due to an accident or some ill-
ness? Note that the insurance company's questions open up a much
wider background by considering what might provisionally be called
second- or third-order causes. The reason for this is that an insurance
company will tend to follow up causal chains until somebody is found
who can be blamed for an accident. Thus, for an insurance company or a
lawyer, the cause of Mr. Jones' fall might be the fact that the owner of
the house failed to replace a broken railing or to provide some kind of
slip protection measures for the dangerous stone steps, or the fact that
Mr. Jones' doctor prescribed some drug for him that affected his capac-
ity to keep his balance to a sufficient extent to cause, in connection with
the slippery state of the steps and Jones' absentmindedness, his fall
down the stairs.
That is, when considering the set of conditions jointly sufficient to
bring about an event, it is not sufficient to look at "first-order" causes
only. In any case, it is a question of focus and perspective what is re-
garded as "first-order cause". Take Mr. Jones' absentmindedness as an
example. Strawson said it "caused" Mr. Jones' fall. But looking more
closely at the actual event, we may be certain that innumerable other
"causes" intervened between Mr. Jones' absentmindedness and his fall.
Absentmindedness cannot immediately "cause" a fall. Before the fall
there must have been a slip. And it is difficult to ascertain at which point
in the causal chain leading to Mr. Jones' fall absentmindedness is sup-
posed to come in. Did absentmindedness cause him to slip, and the slip
cause him to fall? Or did he slip first, and did absentmindedness prevent
The relevance ofcausality 17
him from keeping his balance after the slip? Or did he slip and lose his
balance inevitably, and did absentmindedness prevent him from holding
on to something, which might have prevented his fall? Or was absent-
mindedness a limiting factor of all of Mr. Jones' movements that day,
and did it have disastrous effects all through the causal chain leading to
the accident?
Even if this was a real case, I doubt that all this detail could be ascer-
tained. It should be clear, however, that a distinction between first-order
and second- or third-order causes is impossible and has no systematic
status. What we regard as a cause is seldom the "immediate" cause (in
the sense that no intervening event between cause and effect can be
found) because our observations are seldom so fine-grained that "imme-
diate" causes would be noticed. Let us suppose that theoretically we
come to the conclusion that some basic muscular movement of Mr.
Jones immediately prior to his fall must have been the "immediate"
cause. Presumably neither Mr. Jones nor a by-standing observer would
have been able to observe and identify this "immediate" cause. Pre-
sumably this "immediate" cause was the inevitable result of some prior
events, conditions, and maybe of Mr. Jones' decisions, and these will
most likely be singled out as "the cause". We tend to look at events and
causal chains not from a mechanistic perspective, but from a perspective
that allows us to see and represent points of potential intervention (cf
Von Wright 1973). We look for conditions that could have been differ-
ent. And usually conditions can be different if people behave differently
or change conditions deliberately. Mr. Jones might not have used that
dangerous staircase at all that day. Mr. Jones could have been more
careful. The steps might not have been slippery (somebody could have
done something about that). So the insurance company, by opening up
an infinite space of conditions jointly sufficient for the accident, at the
same time limits the search to those states of affairs that are accessible
to human perception and intervention.
Let us now look at Mr. Jones himself How would he explain his fall
(supposing he survived it)? Again it would depend on a number of con-
tingent factors what would be a satisfactory explanation of the accident
for Mr. Jones himself So Mr. Jones might (most likely) blame the slip-
pery stairs, but it should be clear that he can only do so under certain
conditions. Suppose that Mr. Jones has been using these stairs for years,
they have always been slippery, he has never had an accident and has
never complained about the state of the steps. In that case it would be
somewhat difficult for Mr. Jones to blame the slippery steps. If he chose
18 Paul Georg Meyer
tion. It would be a step "back" towards Grice, Hom, and Searle, away
from Sperber and Wilson. It would mean that relevance, and, in conse-
quence, causality, are categories of communication and interaction, not
of cognition (Edwards and Potter 1993: 24).
For the time being, I am inclined to pursue the second line of reason-
ing, interpreting causality as a discursive-interactional rather than a
cognitive phenomenon. There are two reasons for this: first, this line is
more in accordance with my own previous research on discourse
(Meyer 1975, 1983, 1996) and the role of causality in it. Second, it
seems to be the tum that psychological attribution research is taking at
the moment (Edwards and Potter 1993), so that further linguistic re-
search in that direction opens up promising channels for an interdisci-
plinary exchange of ideas.
As a textlinguist who has been workin~ in a coherence relations
framework (though under a different name), I have always been con-
vinced that causality has a major role to play in the explication of text
constitution and text coherence. But until recently I had tended to be-
lieve that causality was a principle "out there", in the world, being re-
ferred to and being made use of in discourse.
To see research in science, logic, psychology and social sciences
converge on the idea that causality is primarily located in discourse does
not come as a complete surprise, though, given the discursive-
constructionist tum in social sciences in general (Woolgar 1988; Latour
1987; Bazerman 1988) which has left none of the sacred traditional no-
tions of epistemology untouched. Expressions such as fact, truth, nature
or world cannot be used naIvely anymore after taking notice of that dis-
cussion (in which they invariably tum up in "scare quotes", and are ridi-
culed more often than seriously discussed).
Nevertheless, this "discursive turn" of the causality discussion sets
new tasks for discourse analysts as well. We cannot shift responsibility
for the notion of causality to scientists, psychologists, or philosophers
any more, but find ourselves as the branch of knowledge most immedi-
ately competent for a characterisation of that notion. While I remain
sceptical concerning social constructionism in general (Meyer 1997: 67-
74), I have allowed myself to be convinced by the overwhelming evi-
dence in the case of causality. It remains an open question, though,
which is the more appropriate approach, a cognitive or a discursive one.
My decision for a discursive approach is preliminary and inspired by
research-practical considerations.
The relevance ofcausality 25
4. Causality in discourse
(4) Penal methods by themselves will not put an end to crime. Even at
their fiercest, as in nineteenth-century England, ( ..) they did not
succeed.
It seems that causality is such a basic principle that very often there
is no need to draw special attention to it, and the causal relation between
the two clauses is a matter of implicature. Even the direction of causal-
ity is left to the reader to infer in such cases. While in (3) to (6), the sec-
ond clause gives an explanation of some kind relating to the first clause,
the causal relationship is reversed in (7). The second clause states a con-
sequence of the fact reported in the first clause:
26 Paul Georg Meyer
The precise character of the explanations in (3) to (6) has also been the
subject of much debate, but is of little concern for the present contribu-
tion (see Dancygier and Sweetser; Pander Maat and Sanders [this vol-
ume]).
Kortmann (1991: 118-121) has suggested a scale of informativeness
for adverbial relations. Causally based adverbial relations (concession,
condition, instrument, and cause) are to be found above a dividing line
in the middle of Kortmann's scale which separates more informative
from less informative adverbial relations, contrast being the only non-
causal relation figuring in the top group. Kortmann uses this scale in
explaining the semantic interpretation of free adjuncts and absolutes,
i.e., participle constructions in English. It could also be used for a rank-
ing of coherence relations. It is obvious, e.g., that a causal relation is
more informative than a temporal one because a causal relation implies
a temporal relation and only adds further information to this.
A standard analysis of examples (3) to (7) within the coherence rela-
tions approach would be to find specific coherence relations such as
"explanation" or "justification" to describe the relation between the two
clauses. Against such analyses much scepticism has been voiced, e.g.
recently by Dahl (1995: 259):
5. Summary
Notes
7. Other approaches with a similar set of basic assumptions, are Grimes (1975);
Mann and Thompson (1988); Hobbs (1983); Graustein and Thiele (1979)
and many others.
8. Klare, Hugh 1., Stress violence and crime, in: Mayne, Richard (ed.): Europe
tomorrow, London 1972,48-63.
9. Dahl uses Mann and Thompson's (1988) terminology, where "rhetorical
relation" corresponds to what is now called "coherence relation" in most of
the literature.
10. I use capital letters to indicate that I am talking about principles within a
framework that are not necessarily identical with their colloquial counterparts.
11. Grice himself (1975: 46) speaks about "different kinds and focuses of rele-
vance" whose exact nature is not made very clear. The maxim of relevance is
the least explicit one in Grice's account. Maybe the five principles discussed
here could contribute to a clarification.
References
Alm-Arvius, Christina
1993 The English verb see: a study in multiple meaning. (Gothenburg
studies in English 64.) Goteborg: University.
Anscombe, G.E.M.
1971 Causality and determination. Cambridge: Cambridge University :
Press.
1975 Causality and determination. In: Ernest Sosa (ed.), Causation and
conditionals, 63-81. Oxford: Oxford University Press.
Bazerman, Charles
1988 Shaping written knowledge: the genre and activity ofthe experimen-
tal article in science. Madison, WI: University of Wisconsin Press.
Beckermann, Ansgar
1977 Grande und Ursachen: zum vermeintlichen grundsiitzlichen Unter-
schied zwischen mentalen Handlungserkliirungen und wissenschaft-
lich-kausalen Erkliirungen. Kronberg: Scriptor.
Breul, Carsten
1997 Grammatik und Bedeutung der kausalen Satzverbiinde. (Linguisti-
sche Arbeiten 368.) Tubingen: Niemeyer.
Dahl, Osten
1995 Causality in discourse. In: Gert Rickheit and Christopher Habel
(eds.), Focus and coherence in discourse processing, 251-260. (Re-
search in text theory 22.) Berlin: de Gruyter.
Danes, Frantisek
1970 Zur linguistischen Analyse der Textstruktur. Folia linguistica 4: 72-
79.
Edwards, Derek and Jonathan Potter
1993 Language and causation: a discursive action model of description
and attribution. Psychological review 100: 23-41.
32 Paul Georg Meyer
1975 Causes and conditions. In: Ernest Sosa (ed.), Causation and condi-
tionals, 15-38.
Mann, William C. and Sandra Thompson
1988 Rhetorical structure theory: toward a functional theory of text or-
ganization. Text 8: 243-281.
McCawley, James D.
1968 Lexical insertion in a transformational grammar without deep struc-
ture. Papers from the Regional Meeting of the Chicago Linguistic
Society 4: 71-80.
Meyer, Paul Georg
1975 Satzverknupfungsrelationen. Ein Interpretationsmode II fur situation-
sunabhtingige Texte. (Tubinger Beitrage zur Linguistik. 61.) Tubin-
gen: Narr.
1983 Sprachliches Handeln ohne Sprechsituation: Studien zur theoreti-
schen und empirischen Konstitution von illokutiven Funktionen in
'situationslosen' Texten. (Linguistische Arbeiten. 135.) Tubingen:
Niemeyer.
1996 Textfunktion-Textstruktur-Textanalyse: Zur Linguistik des Sach-
textes. In: Wolfgang Bomer and Klaus Vogel (eds.), Texte im
Fremdsprachenerwerb: Verstehen und Produzieren, 21-44. (Tubin-
ger Beitrage zur Linguistik 418.) Tubingen: Narr.
1997 Coming to know: studies in the lexical semantics and pragmatics of
academic English. (Forum fur Fachsprachenforschung 35.) Tubin-
gen: Narr.
Reichenbach, Hans
1991 The space problem in the new quantum mechanics. Erkenntnis 35:
29-47.
Russell, Bertrand
1912 On the notion of cause. Proceedings of the Aristotelian Society 13:
1-26.
1921 On the notion of cause. In: Mysticism and logic and other essays,
180-208. London: Longman.
Searle, John Rogers
1969 Speech acts: an essay in the philosophy of language. Cambridge:
Cambridge University Press.
1983 Intentionality: an essay in the philosophy ofmind. Cambridge: Cam-
bridge University Press.
Sosa, Ernest
1975 Introduction. In: Ernest Sosa (ed.), Causation and conditionals, 1-14.
Oxford: Oxford University Press.
Sosa, Ernest (ed.)
1975 Causation and conditionals. Oxford: Oxford University Press.
Sperber, Dan and Deirdre Wilson
1986 Relevance: communication and cognition. Oxford: Blackwell.
1995 Relevance: communication and cognition. 2. ed. Oxford: Blackwell.
34 Paul Georg Meyer
Stegmtiller, Wolfgang
1983 Probleme und Resultate der Wissenschaftstheorie und Analytischen
Philosophie. Bd. 1: Erkltirung Begrundung Kausalittit. 2., verb. und
erw. Aufl. Berlin: Springer.
Strawson, P.F.
1985 Causation and explanation. In: Bruce Vermazen and Merril B. Hin-
tikka (eds.), Essays on Davidson: actions and events, 115-135. Ox-
ford: Clarendon.
Swales, John
1981 Aspects of article introductions. (Aston ESP research reports 1.)
Birmingham: University of Aston in Birmingham / Language Studies
Unit.
Von Wright, Georg H.
1973 On the logic and epistemology of the causal relation. In: P. Suppes et
al. (eds.), Logic, methodology and philosophy of science, 293-312.
Amsterdam: North Holland.
1975 On the logic and epistemology of the causal relation. In: Ernest Sosa
(ed.), Causation and conditionals, 95-113. Oxford: Oxford Univer-
sity Press.
Wierzbicka, Anna
1972 Semantic primitives. (Linguistische Forschungen 22.) Frankfurt:
Athenaum.
1991 Cross-cultural pragmatics: the semantics of human interaction.
(Trends in linguistics: studies and monographs 53.) Berlin: Mouton
de Gruyter.
1996 Semantics: primes and universals. Oxford: Oxford University Press.
Woolgar, Steve
1988 Science: the very idea. ChichesterlNew York: Tavistock/Methuen.
Ziv, Yael
1988 On the rationality of 'relevance' and the relevance of 'rationality'.
Journal ofpragmatics 12: 535-545.
1993 Causality in language and in discourse. Journal ofliterary semantics
22: 177-185.
1993 Conjunction and causality: pragmatics and the lexicon. Pragmatics
and language learning monograph series 4: 17-27.
On the processing of causal relations
This study deals with how different kinds of causal sentences are under-
stood. The central hypothesis is that sentences that more directly reflect a
causal relation in the world and in the cognitive representation of this
situation are easier to understand than other causal sentences. This hy-
pothesis is tested in reading-time experiments: longer reading times reflect a
greater complexity in the processing of the underlying infonnation. Three
factors are discussed that detennine how directly a causal sentence matches the
cognitive representation of causality: content vs. epistemic relations, linear
order of cause and effect in the sentence, and conceptual order of cause and
effect. It is demonstrated that each of these factors makes an independent
contribution to the complexity of understanding causal sentences. In addition,
it is demonstrated that these factors have an effect independently of a fourth
factor, namely the causal constraint between cause and effect.
1. Causality in cognition
(2) Because John passed the exam, he must have worked hard
(2) Because John passed the exam, he must have worked hard
Conceptually, the cause is derived from the effect. The effect is considered
as a sign for the cause. (Therefore it was identified as an epistemic relation
in the first place). Sentence (2) has the conceptual effect-cause order.
There is an incongruence between what is cause and effect in the real
world and what is expressed in the sentence as antecedent and consequent.
It should be noted that the distinction between conceptual cause-effect
order and conceptual effect-cause order is not the same as the distinction
between content and epistemic relations. In fact, sentence (3) expresses an
epistemic relation, but the effect is derived from the cause.
40 Leo Noordman and Femke de Bli.Jzer
Therefore, sentence (3) has the conceptual cause-effect order. On the other
hand, the two distinctions are not completely independent of each other, as
is shown in Table 1. Content relations only occur in the conceptual cause-
effect order, because content relations directly reflect real-world causality.
Epistemic relations can have either conceptual cause-effect order or effect-
cause order.
conceptual order
cause-effect effect-cause
linear cause-effect effect-cause cause-effect effec~cause
order
content Because John John passed the -
relation worked hard, he exam, because
passed the he worked hard.
exam.
epistemic John worked John must have John must have John passed the
relation hard, so he must passed the worked hard, exam, so he
have passed the exam, since he since he passed must have
exam. worked hard. the exam. worked hard.
Tversky and Kahneman conclude from their studies that people "infer
effects from causes with greater confidence than causes from effects"
(page 118). People make use of cause-effect schemas, and causal infer-
ences that follow the normal cause-effect sequence are easier to make than
diagnostic inferences that reverse this sequence.
The conceptual order of the clauses should be distinguished from the lin-
ear order. Both sentences (1) and (4), repeated below, express the concep-
tual order of cause-effect.
The sentences differ in the linear order of the clauses. In (1), the first
clause expresses the cause and the second clause the effect. In (4), the first
clause is the effect and the second clause is the cause. Sentence (1) thus
has an order which is iconic with the world; sentence (4) is non-iconic.
What can we predict with respect to the processing of causal sentences?
An obvious prediction is that an iconic order facilitates processing. The
reason is similar to the one given in the section on conceptual order. If the
order of the clauses corresponds to the order of causality in the world (or
more specifically, to a model of causality), the matching process between
language and knowledge is easier (and thus the understanding of the sen-
tence).
However, on the basis of the literature one might make a different pre-
diction. Magliano et al. (1993) argue that inferences are made only about
causes and not about effects. They argue that readers make inferences
about causes because these explain the current situation (and the sentence
that expresses this situation). The inference forms a backward link. For-
ward inferences tend not to be made, because, at any point in a text, there
are numerous inferences one can make about the possible effects of the
situation expressed in the current sentence. Lexical decision data did in-
deed support the conclusion that inferences about causal antecedents are
made but inferences about effects are not. However, in this experiment the
causal constraints were rather weak (.30 and .26). We predict that in a
situation in which the cause strongly determines the effect and the effect
42 Leo Noordman and Femke de Blijzer
"In order to earn some money, John was cutting out weeds in his mother's
garden. It was a tough job, because the stinging-nettles were a metre high.
After two hours, he took a short break. The job was half finished. The sun
stood high in the sky and it was sweltering hot. He wiped away the sweat from
his forehead with his hand. He had touched the stinging-nettles with his hand.
His hand itched terribly."
In this fragment the cause sentence ("He had touched the stinging-nettles")
preceded the' effect sentence ("His hand itched terribly"). In the other ver-
sion of the same fragment, the cause sentence and the effect sentence were
reversed ("His hand itched terribly. He had touched the stinging-nettles").
It should be noted that the causal relations in both conditions have con-
ceptual cause-effect order. If we make the causality of the relation explicit,
we obtain the sequences: "because he had touched the stinging-nettles, his
hand itched terribly", and "his hand itched terribly, because he had
touched the stinging-nettles".
The materials in this experiment were constructed in such a way that,
first, the effects could very well be predicted on the basis of the causes
and, second, that the causes could very well be predicted on the basis of
the effects. This was achieved by two separate pilot studies to test and
improve the materials. The experiment showed that a "cause sentence
speeds up the processing of a subsequent consequence sentence, but a
consequence sentence does not speed up the processing of a subsequent
cause sentence... What this experiment demonstrates is a kind of iconicity
between cognitive structure and language" (Noordman and Vonk 1998:
205). If the linear order of the clauses is iconic with our model of causality
in the world, processing the sentence is speeded up. This has been demon-
strated in texts in which the effect is highly predictable given the cause,
and the cause is highly predictable given the effect. In the experiment of
Section 3, reported more completely in de Blijzer (1999), predictability
will no longer be held constant, but will be an independent variable.
Summarizing, we have made three distinctions: content vs. epistemic
relations, conceptual order, and linear order. In principle, these three dis-
tinctions yield a 2x2x2 matrix of eight different kinds of causal sentences
On the processing ofcausal relations 43
(see Table 1). As has already been noted, two cells are empty. A content
relation always has cause-effect order. If the conceptual order is effect-
cause, the relation is always epistemic. In Section 2.1, we discussed an
experiment that compared two kinds of sentences that differ on all three
factors. What we wish to do in the remainder of this paper is to tIy and
find evidence for the factors separately. Also, we would like to further
investigate a fourth factor, the predictability of causes and effects.
A fourth factor that may affect the processing of causal sentences is the
predictability of causes and effects in causal relations. How likely is the
effect given the cause and how likely is the cause given the effect? In par-
ticular, if a speaker draws a conclusion on the basis of an epistemic rela-
tion, it is crucial that there is a solid ground for drawing that conclusion
and that there are few possible alternatives for the conclusion. More spe-
cifically, when the derivation is from cause to effect, it is crucial that there
are few conditions that prevent the cause from having an effect. Following
Cummins et al. (1991), we will call such conditions "disabling condi-
tions". When the derivation is from effect to cause, it is crucial that there
are few other causes. We will call these causes "alternative causes". In the
previous experiment these factors were held constant.
Sentences (5) and (6) illustrate that disabling conditions affect the
plausibility of a sentence that expresses a derivation from cause to effect
(i.e., where the effect is derived from the cause).
(5) The woman must have got a sunstroke, since she was in the sun for
a long time.
(6) The girl must have got wet, since she jumped into the river.
In (5) it is quite likely that there are disabling conditions which could pre-
vent the effect (e.g., wearing a hat, sitting in front of a ventilator, throwing
water over your head once in a while or only being in the sun when it is
not too hot). So when a person has been in the sun for a long time, it is not
at all certain that the effect occurs and that this person does indeed get a
sunstroke. This line of reasoning, expressed by sentence (5), is rather im-
plausible. In (6), however, it is rather unlikely that·there are disabling con-
ditions that could prevent the effect. If someone jumps into the river, it is
44 Leo Noordman and Femke de Blijzer
highly likely that this person gets wet. Therefore, when we know the cause
we can easily draw the conclusion that the effect must have taken place.
When we derive the cause from the effect, the plausibility of a sentence
is affected by alternative causes, as is illustrated by sentences (7) and (8).
(7) The woman got a sunstroke, so she must have been in the sun for a
long time.
(8) The girl got wet, so she must have jumped into the river.
In sentence (7) it is unlikely that there are alternative causes for a sun-
stroke. Therefore, if someone has a sunstroke, we can easily draw a con-
clusion about the cause: this person has been in the sun for a long time.
This line of reasoning is very plausible. In sentence (8), however, it is
likely that there are alternative causes for being wet (e.g., rain, snow, tak-
ing a shower or a bath, a bucket of water). So, when a person is wet, it is
not at all certain that this person has jumped into the river. The line of
reasoning is not very plausible. It should be noted that these observations
do not regard the acceptability of the sentences as linguistic entities (they
all are acceptable), but the plausibility of the sentences as descriptions of
the world.
Cummins (1995) and Cummins et al. (1991) have shown that disabling
conditions and alternative causes affect logical reasoning. For example,
people have difficulty with the following deduction:
(9) If you liein the sun for a long time, you get a sunstroke. Mary has
been lying in the sun for a long time. So Mary will get a sunstroke.
(10) If you lie in the sun for a long time, you get a sunstroke. Mary got a
sunstroke. So Mary must have been lying in the sun for a long time.
On the processing ofcausal relations 45
Many subjects judged such a conclusion correct, because there are few
alternative causes. Cummins et al. showed that disabling conditions and
alternative causes affect judgments on correct and incorrect conclusions.
But in this study we are not interested in true/false judgments concerning
logical conclusions. We are interested in the understanding of causal sen-
tences. The question then is whether disabling conditions and alternative
causes also affect the understanding of causal sentences. Earlier we argued
that language understanding implies an implicit evaluation of the sentence
on the basis of a model of the world. Therefore, it is quite likely that the
understanding of causal sentences is sensitive to factors that affect the
validity of conclusions, and therefore that the understanding of causal
sentences is sensitive to disabling conditions and alternative causes. For
this reason, we incorporated causal constraint, operationalized in terms of
disabling conditions and alternative causes, as a variable in the experiment
to be reported in Section 3.
Sentences (5) to (8) illustrate two causal scenarios. The scenario underly-
ing (5) and (7)-if one lies in the sun for a long time, one gets a
sunstroke-has many disabling conditions (DC) and few alternative
causes (AC); the scenario underlying (6) and (8)-if one jumps into the
river, one gets wet-has few disabling conditions and many alternative
causes. These two kinds of scenarios will be used to study the role of
causal constraint. We collected a set of 71 causal scenarios. In order to
find appropriate sentences for the experiment, we presented these
scenarios to twenty participants in two judgment tasks. In one task,
participants had to assume that the causal event occurred and estimate on a
four-point scale how certain the effect was. For example: Suppose that the
girl jumps into the river. How certain is it that she will get wet? The higher
the score, the higher the causal constraint, the smaller the impact of
disabling conditions. The second task was just the reverse. Participants had
to assume that the effect event had occurred and to estimate on a four-
point scale how certain the cause event was. For example, suppose the girl
has got wet, how certain is it that she jumped into the river? The higher the
score, the higher the causal constraint, the smaller the impact of alternative
causes. For each scenario, the mean DC (disabling conditions) and AC
(alternative causes) scores over the twenty participants were computed as
well as the mean difference between the AC scores and the DC scores
46 Leo Noordman and Femke de Blijzer
The aim of this experiment was to find out whether the processing of con-
tent relations differs from that of epistemic relations and whether the proc-
essing of relations that have cause-effect conceptual order differs from
relations that have effect-cause conceptual order. Linear order was kept
constant: in all cases the first clause mentioned the effect and the second
clause the cause.
The experimental conditions in this experiment are illustrated in Table
2. The experiment was conducted in Dutch, but the table presents English
translations as well. There is only one type of content sentence: a sentence
type that expresses a cause-effect conceptual order. There were three types
of epistemic sentences. The fact that a sentence expresses an epistemic
relation was marked by the modal must have (in Dutch zal we/), or by the
conjunctions so or since (in Dutch: dus and want). In sentence type 3, the
epistemic character was indicated only by the conjunction so. The four
sentence types were applied to the two categories of scenarios.
On the processing ofcausal relations 47
ample, to find out whether the context can overrule the effects of these
factors.
What kind of reading task is appropriate in this situation? We are inter-
ested in understanding causal relations. Understanding requires that the
sentence is related to a model of the world. Accordingly, when readers
understand a sentence, they should be able to evaluate whether the causal
relation expressed by the sentence is a possible causal relation in the
world. In the experiment, that was the task the readers were instructed to
perform: Is a causal relation possible between the two clauses? Since this
task requires that there are also impossible causal relations, so-called filler
items are incorporated in the experiment that expressed impossible rela-
tions (see Section 3.3.)
3.1. Procedure
3.2. Predictions
The four types of sentences (see Table 2) define the four conditions in this
experiment. The predictions for the reading times for the identical second
clauses (excluding the conjunction) in conditions 1, 2, and 3 are as fol-
lows. On the basis of conceptual order, the reading times for conditions 1
and 2 are expected to be shorter than for condition 3. The reading times for
On the processing ofcausal relations 49
the content condition 1 should be shorter than for the epistemic condition
2. This leads to a predicted increase in reading times for the conditions 1 to
2 to 3. The second clause in condition 4 is two words longer than in the
other conditions, so these reading times cannot be compared to the reading
times for the other second clauses, but the total sentence in condition 4 has
the same length (in terms of words and letters) as in condition 2. Based on
conceptual order, the reading time for sentences in condition 2 is predicted
to be shorter than in condition 4.
It is likely that these predictions are qualified by the causal constraint:
disabling conditions may make the cause-effect order more difficult, and
alternative causes may make the effect-cause order more difficult. More
specifically, for the scenarios that have low DC scores and high AC
scores, the low DC scores will make the cause-effect items easier, and the
high AC scores will make the effect-cause order more difficult. For the
scenarios that have high DC scores and low AC scores, the high DC scores
will make the cause-effect items more difficult and the low AC scores will
make the effect-cause items easier. So the difference between conditions 1
and 2 on the one hand and 3 on the other hand will be greater for the sce-
narios with low DC scores and high AC scores than for the scenarios with
high DC scores and low AC scores. In other words, an interaction is pre-
dicted between conceptual order and the two categories of causal con-
straint. No effect of causal constraint is expected on the comparison be-
tween content and epistemic sentences (condition 1 versus 2), because
both sentence types have the conceptual cause-effect order.
The effect of causal constraint can be investigated in an additional way.
The hypothesis is that the cause-effect order becomes more difficult when
the DC scores are higher. Consequently, there should be a positive corre-
lation between the DC scores and the reading times for the causal clauses
of conditions 1 and 2. The effect-cause order is supposed to become more
difficult when the AC scores are higher. Consequently, there should be a
positive correlation between the AC scores and the reading times for the
causal clauses of conditions 3 and 4.
3.3. Materials
DC/high AC. For each scenario, four sentences were formulated corre-
sponding to the four experimental conditions. Four lists of materials were
composed, by pairing each of the 48 scenarios with a different experi-
mental condition in each list, so as to balance the lists with respect to the
scenarios and experimental conditions. Although sentence length does not
affect the way the hypotheses are tested, it is advisable to keep the sen-
tence length as constant as possible in order to reduce experimental noise.
The mean lengths of the causal clauses for the low DC/high AC and the
high DC/low AC scenarios were 8.25 and 8.33 syllables respectively. The
mean lengths of the complete sentences were also rather similar: 15.21 and
16.21 syllables respectively.
Given the fact that the task was to judge whether the causal relation
was possible, and all the experimental sentences required an affirmative
answer, filler sentences requiring a negative answer were needed. The
filler sentences expressed two clauses between which no causal relation
was possible. The clauses themselves made perfect sense and were quite
possible, so that the negative answer could only be based on the impossi-
bility of a causal relation, as it should be. An example is: "The letters have
become blurred because someone held the book upside down", and "Her
nose was bleeding because she had a hair cut". There were 32 filler sce-
narios. The filler scenarios were incorporated in the four lists and were
balanced with the experimental conditions in the same way as the experi-
mental scenarios.
Each of the four lists was presented to 8 participants. The participants
were students ofTilburg University, between 18 and 27 years of age.
1 2 3 4
content epistemic epistemic epistemic
cause-effect cause-effect effect-cause effect cause + modal
low disabling conditions / high alternative causes
1567 1730 2029 1985
Mean
1473 1610 1824 1829
The reading times for the clauses in condition 1 are a significant 137 msec
shorter than for clauses in condition 2: F 1(1,28)=8.44, p<.Ol~
F2(1,40)=8.82, p<.Ol, supporting the hypothesis that content relations are
easier to process than epistemic relations. The reading times for clauses in
condition 2 are 214 msec shorter than for clauses in condition type 3:
F 1(1,28)=27.72, p<.OOl~ F2(1,40)=13.37, p<.OOI, supporting the hypothe-
sis that conceptual order plays a role: cause-effect relations are easier to
process than effect-cause relations.
There was a significant interaction between the experimental conditions
and the causal constraint, just as predicted. The difference between cause-
effect and effect-cause conceptual orders is greater for low DC/high AC
scenarios than for high DC /low AC scenarios. Specifically, the difference
between clauses in conditions 1 and 3 is greater for the low DC/high AC
than for the high DC/low AC scenarios: F 1(1,28)=7.53, p=.OI~
F 2(1,40)=4.66, p<.05. A similar pattern is observed for the difference be-
tween clauses in conditions 2 and 3, although the F 1 shows only a trend
and the F2 is not significant: F 1(1,28)=3.69, p=.065~ F2(1,40)=2.03, p=.16.
These interactions support the hypothesis that the cause-effect order is
more difficult when there are more disabling conditions and that the effect-
cause order is more difficult when there are more alternative causes.
As predicted, the difference between the clauses in conditions 1 and 2
is constant for the two categories of causal constraint. That should be the
case, since sentences in both conditions express cause-effect relations. One
prediction was not confirmed. The reading time for the complete sentence
2 was not shorter than for the complete sentence 4, as emerges from the
data in Table 4.
52 Leo Noordman and Femke de Blijzer
Table 4. Reading times (msec) for the complete sentences in conditions 2 and 4
2 4
epistemic epistemic
cause-effect effect-cause
low disabling conditions / high alternative causes
3770 3709
high disabling conditions / low alternative causes
3621 3441
mean
3695 3575
The basis for this prediction was that the conceptual order in sentence 2
but not in sentence 4 corresponds to the order of cause-effect in the world.
The prediction was justified by the fact that sentences 2 and 4 had the
same length. In sentence 2, the modal must have occurred in the first
clause; in sentence 4 it occurred in the second clause. The assumption was
that the effect of must have is constant. But this assumption appears to be
wrong. In fact, the reading times for the second clauses in condition 3 and
4 did not differ. This is surprising, since the Dutch clause in 4 (ze zal weI
lang in de zon hebben gezeten; "she must have been in the sun for a long
time") is two words longer than clause 3 (ze heeft lang in de zon gezeten;
"she was in the sun for a long time"). Interestingly, the same difference in
length occurred (in the original Dutch examples) in the first clauses of
conditions 2 (De vrouw zal weI een zonnesteek hebben; "The woman must
have got a sunstroke") and 4 (De vrouw heeft een zonnesteek; "The
woman got a sunstroke"). Here the longer clause did indeed require a
longer reading time; the reading times for the first clause in the conditions
2 and 4 were 1464 msec and 1130 msec respectively.
The interpretation is that must have in the second clause (condition 4) is
an indication that the clause expresses a conclusion on the part of the
writer. So must have facilitates the epistemic interpretation. If must have is
absent (in condition 3), the reader has to infer on the basis of so and hislher
knowledge of the world that the sentence expresses an epistemic conclu-
sion. A similar effect has been observed by Cozijn (in preparation) for the
conjunction because. The causal relation between two clauses can, but
need not be expressed by the conjunction because. If there is no conjunc-
tion, the sentence is shorter. But Cozijn found that the presence of because
speeds up reading, counteracting the increase in length. The facilitating
effect of must have is not expected for the first clause (conditions 2 and 4).
The first clause of condition 4 is initially interpreted as a content relation,
and during the first clause there is not yet any evidence that the sentence
On the processing ofcausal relations 53
Table 5. Correlations between clause reading times (msec) and scores for
alternative causes (AC) and disabling conditions (DC) respectively.
1 2 3 4
content epistemic epistemic epistemic
cause-effect cause-effect effect-cause effect-cause + modal
AC .41* .37* .51 * .50*
DC -.01 .00 -.08 -.01
* p<.Ol
This makes perfect sense, since the effect clause was presented first, lead-
ing the reader to anticipate the cause sentence. The greater the predictabil-
ity ofthe cause sentence, the shorter, in fact, the reading time.
4. Conclusion
Third, the experiment also showed that the causal constraint has an ef-
fect. Processing from effect to cause is easy for scenarios that score low on
alternative causes. Processing from cause to effect is easy for scenarios
that score low on disabling conditions. But in both cases, the effect of
content vs. epistemic and the effect of conceptual order was observed.
Fourth, there is an effect of linear order. Presenting the cause first fa-
cilitates processing, as was indicated by the results of the experiment re-
ported in Section 2.3. An effect of linear order in processing was also
demonstrated in the experiment described in Section 3. In that experiment,
the effect clause was always presented as the first clause. The reading time
for the causal clause was shorter as it was the more predictable. Appar-
ently, readers try to anticipate the clause coming next.
Acknowledgment
References
Blijzer, F. de
1999 Causale scenario's. Het effect van de conceptuele volgorde en de
sterkte van de oorzaak-gevolg koppeling op de verwerking van seman-
tische en epistemische relaties [Causal scenarios. The effect of concep-
tual order and causal constraint on the processing of semantic and
epistemic relations]. MA thesis, Tilburg University, The Netherlands.
Cozijn, R
in preparation Integration and inference processes in understanding causal
relations. Ph.D. dissertation, Tilburg University, The Netherlands.
Cummins, D.D.
1995 Naive theories and causal deduction. Memory and Cognition 23: 646-
658.
Cummins, D.D., T. Lubart, O. Alksnis and R Rist
1991 Conditional reasoning and causation. Memory and Cognition 19: 274-
282.
Guenthner, F.
1989 Discourse: understanding in context. In: H. Schnelle and N.O. Bernsen
(eels.), Logic and Linguistics. Research Directions in Cognitive Science,
127-142. (European Perspectives 2.) Hove: Erlbaum.
56 Leo Noordman and Femke de Blijzer
Johnson-Laird, P.N.
1983 Mental Models. Cambridge: Harvard University Press.
Magliano, J.P., W.B. Baggett, B.K. Johnson and A.C. Graesser
1993 The time course of generating causal antecedent and causal conse-
quence inferences. Discourse Processes 16: 35-53.
Meyer Viol, W.P.M.
1984 Foregrounding and causality in inference making. M.A. thesis, Max
Planck Institute for Psycholinguistics, Nijmegen, The Netherlands.
Michotte, A.
1954 La Perception de la Causalite. Louvain: Publications Universitaires de
Louvain.
Noordman, L.G.M.
1979 lnferringfrom Language. Berlin: Springer.
Noordman, L.G.M. and W. Vonk
1998 Memory-based processing in understanding causal infonnation. Dis-
course Processes 26: 191-212.
Sanders, T.J.M., W.P.M. Spooren and L.G.M. Noordman
1992 Toward a taxonomy of coherence relations. Discourse Processes 15: 1-
35.
Sweetser, E.
1990 From Etymology to Pragmatics. Metaphorical and Cultural Aspects of
Semantic Structure. Cambridge: Cambridge University Press.
Traxler, MJ., A.J. Sanford, J.P. Aked and L.M. Moxey
1997 Processing causal and diagnostic statements in discourse. Journal of
Experimental Psychology: Learning, Memory, and Cognition 23: 88-
101.
Tversky, A. and D. Kahneman
1982 Causal schemas in judgments under uncertainty. In: D. Kahneman, P.
Slovic and A. Tversky (008.), Judgment under Uncertainty: Heuristics
and Biases, 117-128. Cambridge: Cambridge University Press.
Domains of use or subjectivity? The distribution
of three Dutch causal connectives explained*
1. Causal connectives
(1) The neighbours suddenly left for Paris last Friday. As a result /
That's why / So they are not at home.
(2) The lights in the neighbours' living room are out. # As a result / ?
That's why / So they are not at home.
In other words, there are restrictions on the use of a cue phrase and
connectives. Ever since Halliday and Hasan's (1976) seminal work, it has
repeatedly been argued that the lexicon of connectives and cue phrases is
ordered according to the type of relation they can express (e.g. Martin
1992; Knott and Dale, 1994). For instance, temporal connectives can be
distinguished from causal connectives. Yet, as the examples (1) and (2)
show, the restrictions on the use of connectives and cue phrases are also
more subtle, i. e. they also hold within the class of causal relations as such.
Not all causal markers express the same type of causal relation. How then
can these restrictions be described?
In this chapter we report a study on the meaning and use of three
Dutch causal connectives: daardoor, daarom and dus. All three
connectives express causality in a 'forward' direction, that is, cause
precedes consequence. These connectives can best be translated in
English by means of the phrases we used in examples (1) and (2):
daardoor is similar to as a result / as a consequence, daarom can best be
translated as that's why and dus is quite similar to so. In (3) and (4) we
have translated examples (1) and (2) in Dutch.
All three connectives seem to fit in (3), although they do express different
relations. Daardoor expresses a simple cause-consequence relation in
which the second segment (S2) might even be the unintended result of the
Domains o.fuse and subjectivity 59
(7) What are you doing tonight, because there's a good movie on.
In (5) John's love was the real-world cause of his coming back. (6)
means that my knowledge of John's return leads to the conclusion that he
loved her. An adequate paraphrase of (7) would be "I am asking what
you are doing tonight because there is good movie on". Distinctions like
these have been used by many linguists to account for differences in the
use of connectives and cue phrases (e.g. Halliday and Hasan 1976;
Martin 1992; Knott and Dale 1994), even though there are clear
differences in terminology and definitions (Bateman and Rondhuis, 1997;
Moore and Pollack 1992; Oversteegen 1997; Sanders and Spooren
1999).
How do such distinctions apply to the three Dutch causal connectives
introduced earlier? If we reconsider the two examples, (3) and (4), it was
already noted that in (3) all three connectives seem to fit in, although they
do express different relations. Daardoor expresses a content (Cause-
Consequence) relation, just like daarom (although this connective gives
rise to a volitional, a Reason-Action relation), and with dus the sequence
is of an epistemic nature: S2 is a conclusion based on S1. In (4), which
can only be interpreted epistemically, dus fits in very well, whereas
daardoor leads to an unacceptable sequence and daarom is at least odd.
Hence, at first sight, the domain theory seems relevant for the three
connectives. A first conclusion might be that daardoor can be described
as a connective used exclusively for content relations.
Similar conclusions of domain-specificity are reported for causal
connectives in other languages. For instance, it has been suggested that
German denn can only be used to express epistemic relations (among
others, Giinthner 1993; Keller 1995), just as French donc (Rossari and
Jayez 1996), Hungarian hat (Nemeth 1995) and English since are
supposedly impossible to use in the content domain (Sweetser 1990).
The attraction of the domain theory becomes even more apparent if we
consider the similarity between the domain distinctions and the categories
of coherence relations that have been proposed in cognitively inspired
work on relation categorization. In Sanders, Spooren and Noordman
(1992), for instance, it is shown how the set of coherence relations that
can hold between discourse segments can be decomposed in more basic
categories by making use of only four parameters. One of these
parameters, the Source of Coherence, which distinguishes between
semantic and pragmatic relations, shows great resemblance to Sweetser's
content-epistemic-speech act distinction (Sanders 1997; Knott 2000).
Given the 'basicness' of these conceptual distinctions it is attractive to
Domains ofuse and subjectivity 61
The use of dus, daarom and daardoor was studied in a corpus of 150
fragments, 50 for each connective. The corpus consisted of different text
types, argumentative / persuasive as well as descriptive / informative. It
contained annual reports and publicity brochures of Dutch companies,
public information brochures, articles from the readers' opinion page of a
Dutch quality paper (De Volkskrant 1994) and news articles from this
same paper.
Our methodology involved three steps. We first determined the
possible relational interpretations of these fragments without connectives.
Then we counted how often each connective expressed a certain relation.
In the final step, we checked whether the original connective could be
replaced by another. This substitution method is a way of testing
semantic intuitions (Knott and Dale 1994; Knott and Sanders 1998). The
questions are: Does substitution lead to a sequence which is still
62 Henk Pander Maat and Ted Sanders
(9) Vaste klanten zijn voor de Bijenkorf van vitaal belang: zij
besteden per jaar twee maal zoveel in de Bijenkorfwinkels als
andere klanten. Daarom heeft de Bijenkorf aan de Vaste
Klantkaart een aantal voordelen verbonden
'Regular customers are of vital interest to the Bijenkorf: in a year
they spend twice as much as other customers do. Daarom 'that's
why' the Bijenkorf has added a number of advantages to the
Regular Customer-Card. '
(10) Er blijkt weer een toename van het aantal besmettingen met het
HIV-virus te zijn. Ais AIDS-verpleegkundige zie ik de gevolgen
ervan. AIDS-preventie, daar komt het dus op aan.
'There appears to be an increase of infections with the HIV-virus.
As an AIDS-nurse, I see the consequences. Dus 'so' prevention of
AIDS, that's what it's all about. '
Domains ofuse and subjectivity 63
These results imply that only the use and meaning of daardoor can be
described in terms of relational domains, but even there we need the
additional criterion of volitionality, that is we need to determine whether
the relation concerns a reason for an intentional action or not. Daardoor is
restricted to the content domain, more specifically to relations of the non-
volitional content type. The difference between daarom and dus is
difficult to describe in terms of domains: both connectives regularly
express both kinds of relations. Hence we are faced with the situation that
the purported distinction between content and epistemic causality remains
unmarked in the field of most common Dutch connectives, while the
distinction between volitional and non-volitional relations is marked
linguistically by commonly used causal connectives. Does this imply that
(non-)volitionality has to be considered a more fundamental criterion than
the distinction between content and epistemic meaning?
In proposals for the categorization of coherence relations, volitionality
is indeed considered relevant but it is treated as a second-order distinction
within the content domain relations. For instance, Sanders et al. first
distinguish between semantic (equivalent to content) and pragmatic
(roughly equivalent to epistemic and speech act) relations, and then
separate volitional from non-volitional relations and epistemic relations,
as do Mann and Thompson (1988). Similarly, in the linguistic literature
on connectives and adverbial subordination, the first relevant distinction is
between content and epistemic.
It seems we might have to allow for the idea that, as far as the
typology of causality is concerned, a distinction of the type ["volitional]
may be as fundamental as the one between content and epistemic
domains (Stukker, Sanders and Verhagen 1999). Adding to this the
observation that volitional and epistemic relations are often lexicalized by
the same connectives, it makes sense to ask what the common
denominator in these relations might be.
So far, the only clear contrast we have seen in the data from actual
language use is the one between daardoor on the one hand, and dus and
daarom on the other: daardoor only expresses non-volitional relations,
while daarom and dus may express both volitional and epistemic
relations. This contrast constitutes a problem for the 'domain theory',
since it cuts right through the content domain. Hence our task is to
64 Henk Pander Maat and Ted Sanders
3.1. Subjectivity
The question now is: what is it that 'epistemic' causality has in common
with 'volitional' causality, so that speakers of Dutch easily use the same
vocabulary to express these two relation types, but not to express
relations of non-volitional causality? We want to suggest that what
epistemicity and volitionality have in common is that both crucially
involve an animate subject, a person, whose intentionality is
conceptualized as the ultimate source of the causal event, be it an act of
reasoning or some 'real-world' activity. That is, it seems that to us
humans a very fundamental distinction is the one between events
ultimately originating from some mind, and events that originate from
non-intentional causes; between causes that are crucially located in a
subject of consciousness, and those that are located in the inanimate
world. This distinction is so fundamental that it shows up in similar ways
at different linguistic levels and is often the only one marked explicitly by
means of some linguistic form (Verhagen 1995 and other contributions to
Stein and Wright 1995).
We will use the notion of subjectivity to specify this idea further. In a
sense, every linguistic utterance is subjective, that is, connected to the
point of view of some 'subject'. However, this use of the concept is not
very informative. A more interesting question regarding this notion is
whether this 'subject' is included in the semantic characterization of the
utterance. Langacker (1990), who applies the notion of subjectivity to
several linguistic phenomena, distinguishes between three situations (Pit
1997). First, the ground-the term he uses to refer to the speech event,
its participants and its immediate circumstances-may be entirely
external to the semantics of the utterance. This situation is exemplified by
(11) below. Second, the ground may be included in the scope of
predication as an offstage, unprofiled reference point. This is the case
when deictics like yesterday, tomorrow, etc. are used. Another example
is (12) below, in which the modal adverb probably invokes the present
speaker as the source for the probability judgement. Third, the ground
may be put onstage, as in (13). In this case, the ground is more or less
,objectified', that is, it is made part of the situation referred to in the
utterance.
Domains ofuse and subjectivity 65
(16) It was a small case, completely rusty. So it was made out ofiron.
The causal relation in (16) does not relate real-world situations, but two
beliefs of an SOC (Knott 2000). In this case the SOC is a concluder, that
is somebody making inferences. In (16), the concluder is identical to the
speaker. And the relation between these two beliefs is again situated
within the causal domain of this SOC. Hence, the contrast between non-
volitional and epistemic relations seems rather clearcut in that the
representation of epistemic relations assigns a very prominent role to an
SOC, while such a participant is entirely absent in our interpretation of
non-volitional relations.
As was indicated above, the distribution of causal connectives in
Dutch suggests a similarity of volitional and epistemic relations: dus and
daarom both express volitional and epistemic relations, whereas
daardoor encodes non-volitional relations. How should we conceptualize
this similarity? In our View, the representation of a volitional relation also
involves an SOC, namely the actor of the action in the second segment.
Consider fragment (17).
(19a) The lights in the neighbours' living room were out. So they were
not at home.
(19b) Harry saw that the lights in his neighbours' living room were out.
He concluded they were not at home.
For these reasons the distance between the SOC and the present speaker
is larger in (20b) than in (20a).
In section 2 it was found that, although daarom and dus may both
express volitional and epistemic relations, dus is much more frequent in
epistemic relations than it is in volitional relations, while daarom shows a
slight preference for volitional relations. The hypothesis we want to
explore below is that this differential distribution is to be explained by a
difference between the two connectives with regard to the distance they
express between the SOC involved and the present speaker. More
specifically, our claim will be that daarom encodes a larger distance than
dus does. We will now derive some specific predictions from this general
claim, so that they can be put to the test in a corpus study.
Our first prediction presents itself in a rather straightforward way.
When dus really encodes a smaller distance between the SOC and the
speaker, the subject responsible for the action or the conclusion in dus-
segments should more often be identical to, or close to the speaker than
the actors and concluders in daarom-segments are. In order to test this
claim, we distinguish between four types of sacs, ranging from minimal
(1) to maximal (4) distance between the SOC and the speaker.
In this example, there are two possible interpretations. In the first one,
Fritz decides to go home on the basis of his knowledge of the time and his
knowledge of the need to go home at five. Under this interpretation, Fritz
is the SOC. In the second interpretation, the speaker presents Fritz'
behavior as an habitual action: always when the clock strikes five, Fritz
forgets all about his work and gets ready to leave. Under this
interpretation, the speaker is the SOC, because Fritz does not necessarily
consciously decide to leave work, he just does what he does every day.
If the segments in example (21) are connected by dus, the second
interpretation is preferred. If we want listeners to construct the first
interpretation, and we still use dus, we would need to mark the first
segment as being presented from the perspective of the non-default SOC:
the actor, not the speaker. Examples containing such non-default SOCs
are (21a) and (21b). In (21a) the SOC is explicit in the first segment; in
70 Henk Pander Maat and Ted Sanders
(21 a) Frits hoorde de klok vijfuur slaan. Dus pakte hi} zi}n aktentas.
'Fritz heard the clock striking five. Dus 'so' he took his briefcase.'
(21b) Frits keek op zi}n horloge. Het was al laat. Dus pakte hi} zi}n
aktentas.
'Fritz looked at his watch. It was late already. Dus 'so' he took his
briefcase. '
4. Corpus analysis
4.1. Corpus
For reasons of exposition, we start with the results regarding the second
hypothesis, the continuity hypothesis. In the analysis, we determined the
perspective of the first segment in the third-person volitional relations for
both connectives (daarom: n=2I; dus: n=I4). The results (see Table 1)
show that the proportion of fragments characterized by an explicitly or
implicitly continuous perspective for dus is significantly larger than that
2
for daarom (Chi =15.26, p < .001, df=2).
(la) She had sent some poems to Maatstaj, (lb) and he had loved
them-(lc) found her poetry had an immediate perfection. (2a)
Dus 'so' he wrote to ask her if she had any more (2b) and this
became her first collection of poems. '
(24) (1) Maar al te vaak wordt het beeld van het buurland bepaald
door de media. (2a) Berichten over positieve zaken zijn geen
nieuws, (2b) dus concentreert men zich op het negatieve.
'(1) All too often the image of a neighbouring country is
determined by the media. (2a) Information about positive things is
not news, (2b) dus 'so' one concentrates on what is negative. '
Domains ofuse and subjectivity 73
'''In ' 68 culture was banned from there and degraded to 1,3 billion
spitting Chinese who cannot bring themselves to any respect for
beauty. Bags full of trash and shit simply go out of the window.
That is too much for me, dus 'so' I take the second class. '"
'that's why' it is not contradictory that the voters voted for the new
constitution and at the same time for Zjirinovski. '
(28) (1) De VPRO-quiz was echter een open quiz. (2) Het stond
iedereen vrij om allerlei feiten op te zoeken in encyclopedie of
bibliotheek. (3) Of om overleg te plegen met familie, vrienden of
collega's. (4) Het hebben van een hoge score impliceert dus niet
noodzakelijkerwijs dat men ook veel verstand van wetenschap
heeft.
'(1) However, the VPRO-quiz was an open quiz. (2) Everybody
was allowed to look things up in an encyclopedia or a library. (3)
Or to consult family, friends or colleagues. (4) Dus 'so' having a
high
.
score
,
does not necessarily mean that one knows a lot about
SCIence.
(29) (la) Hi} had ervaring op die circuits, (lb) ik niet. (2) Hi} maakte
daarom theoretisch meer kans om punten te halen.
'(la) He was experienced on those circuits, (lb) I was not. (2)
Daarom 'that's why' in theory he had a better chance to score
points. '
(30) (1) Het punt is echter dat de verdachten die schuldjuist met klem
ontkennen en daarom bi} het hof hoger beroep hebben
aangetekend tegen hun veroordeling bi} de rechtbank. (2) Die
strafis dus nog niet definitief
'(1) The issue, however, is that the suspects vehemently deny their
guilt and have lodged an appeal against their conviction by the
court. (2) Dus 'so' this sentence is not final yet.'
This result fits very well with our account, which claims that dus and
daarom encode different speaker-SOC distances. In some contexts, no
SOC is present, either explicitly or implicitly (in the form of, for instance,
subjective elements). In these situations the use of dus is unproblematic.
The reason is that the processing instruction constituted by dus, i.e. 'find
an SOC close or identical to the speaker', can always be carried out
successfully because every utterance presupposes the presence of a
speaker. However, the instruction encoded by daarom, i.e. 'fmd an SOC
at a certain distance from the speaker' presents problems in situations
where the utterance itself does not present an SOC at all. After all,
selecting the speaker is impossible. In order to support selection of the
speaker, daarom needs linguistic elements indicating his presence,
though not referring to him as such.
5. Conclusion
Notes
References
Ariel, Mira
1988 Referring and accessibility. Journal ofLinguistics 24: 65-87.
Bateman, John A. and Rondhuis, Klaas-Jan
1997 Coherence relations: towards a general specification. Discourse
Processes 24: 3-49.
Coates, Jennifer
1983 The semantics ofmodal auxiliaries. Beckenham: Croom Helm.
Cornelis, Louise H.
1997 Passive and perspective. Amsterdam!Atlanta: Rodopi.
Degand, Liesbeth
1998 Het ideationele gebruik van want en omdat. [The Ideational Use of
want and omdat]. Nederlandse Taalkunde 4: 309-326.
Degand, Liesbeth and Henk Pander Maat
1999 Scaling causal relations in terms of speaker involvement. In: Levels
of Representation in Discourse, 45-53. Working notes of the
International Workshop on Text Representation. University of
Edinburgh.
Fauconnier, Gilles
1994 Mental spaces. Cambridge: Cambridge University Press.
Fauconnier, Gilles and Eve Sweetser (eds.).
1996 Spaces, worlds and grammars. Chicago: University of Chicago
Press.
Gtinthner, Susanne
1993 "...weil-man kann es ja wissenschaftlich untersuchen"-
Diskurspragmatische Aspekte der WortstellUng in WEIL-Satzen.
Linguistische Berichte 143: 37-59.
Halliday, M.A.K. and Ruqaiya Hasan
1976 Cohesion in English. London: Longman.
Hanks, William F.
1992 The indexical ground of deictic reference. In: Alessandro Duranti
and Charles Goodwin (eds.) Rethinking context. Language as an
interactive phenomenon, 43-76. Cambridge: Cambridge University
Press.
Keller, Rudi
1995 The epistemic weil. In: Dieter Stein and Susan Wright (eds.)
Subjectivity and subjectivisation. Linguistic perspectives, 16-30.
Cambridge: Cambridge University Press.
Knott, Alistair and Robert Dale
1994 Using linguistic phenomena to motivate·a set of coherence relations.
Discourse Processes 18: 35-62.
Knott, Alistair and Ted Sanders
1998 The categorization of coherence relations and their linguistic
markers: an exploration of two languages. Journal of Pragmatics
30: 135-175.
80 Henk Pander Maat and Ted Sanders
Knott, Alistair
2000 Semantic and pragmatic relations and their intended effects. In:
Ted Sanders, Joost Schilperoord and Wilbert Spooren (eds.),
Cognitive approaches to coherence. Amsterdam: Benjamins.
Kortmann, Bernd
1997 Adverbial subordination. A typology and history of adverbial
subordinators based on European languages. BerlinlNew York:
Mouton de Gruyter.
Langacker, Ronald W.
1985 Observations and speculations on subjectivity. In: John Haiman
(ed.), Iconicity in syntax, 109-150. Amsterdam/Philadelphia:
Benjamins.
1990 Subjectification. Cognitive Linguistics 1: 5-38.
Lemke, Jay L.
1998 Resources for attitudinal meaning. In: Functions ofLanguage 5: 33-
56. Amsterdam: Benjamins.
Lyons, John
1977 Semantics 2. Cambridge: Cambridge University Press.
Mann, William C. and Sandra A. Thompson
1988 Rhetorical structure theory: Toward a functional theory of text
organization. Text 8: 243-281.
Martin, J. R.
1992 English text. System and structure. Philadelphia: Benjamins.
Moore, Johanna D. and Martha E. Pollack
1992 A problem for RST: The need for multi-level discourse analysis.
Computational Linguistics 18: 537-544.
Nemeth, Enik6
1995 On the role of pragmatic connectives in Hungarian spoken
discourses. In: B. Warvik, S-K. Tanskanen and R. Hiltunen (eds.),
Organization in discourse. Proceedings from the Turku conference,
393-402. Anglicana Turkuensia 14.
Nuyts, Jan
1994 Epistemic modal qualifications: on their linguistic and conceptual
structure. Antwerp Papers in Linguistics 81. Antwerpen:
Universiteit Antwerpen.
1992 Subjective vs. objective modality: what's the difference? In:
Fortescue, Michael D., Peter Harder and Lars Kristoffersen (eds.),
Layered structure and reference in a functional perspective, 73-99.
Amsterdam: Benjamins.
Oversteegen, Leonoor E.
1997 On the pragmatic nature of causal and contrastive connectives.
Discourse Processes 24: 51-85.
Pander Maat, Henk
1998 Classifying negative coherence relations on the basis of linguistic
evidence. Journal ofPragmatics 30: 177-204.
1999 The differential linguistic realization of comparative and additive
coherence relations. Cognitive Linguistics 10: 147-184.
Domains ofuse and subjectivity 81
Verhagen, Arie
1995 Subjectification, syntax and communication. In: Dieter Stein and
Susan Wright (eds.), Subjectivity and subjectivisation, 103-128.
Cambridge: Cambridge University Press.
Causal relations in spoken discourse: Asyndetic
constructions as a means for giving reasons*
Christine Gohl
1. Introduction
Earlier research on causal relations has dealt mainly with explicit lexical
markers, above all with subordinating conjunctions. More recent studies
on causal relations in spoken English draw attention to the paratactic use
of because (e.g. Ford 1994; Ford and Mori 1994; Couper-Kuhlen 1996;
Schleppegrell 1991), while studies on causal conjunctions in spoken
German focus, for example, on weil ('because') and the use of main
clause constructions in weil-clauses (e.g. Giinthner 1996; Pasch 1997;
Scheutz 1998; Uhmann 1998). This focus on the use of conjunctions
reflects the fact that they are indeed an important means for signaling
causal relations. That is not to say, however, that they are the most
frequent means of signaling causal relations in spoken interaction.
Especially in informal talk, a frequent device is the use of asyndetic
linkage. In these cases, speakers produce a reason for a previous action
without explicitly linking (e.g. via a causal connector) the utterance that
carries out the action to the utterance that is used to provide the account.
Although one can find various remarks in the literature on such asyndetic
causal structures (e.g. Blakemore 1987: Ch. 4.2.), they have not been
84 Christine Gohl
, 1Maria: uwe dear do you want anything else (.) look there' re
bo[nes to pick. 4
2Erik: [he'll get nothing more we have to smoke. (.)
3 ?Erik: cheers. «clinking of glasses))
4?: [(4 syllables)]
5Anna: [we'll come home totally undernourished,
6 just because these guys want to smoke. '
The tum produced by Erik in line 2, nix kriegt er me: mir musset rauche
'he'll get nothing more we have to smoke', consists of two successive
tum-constructional units which together form a cohesive unit or
construction. The cohesiveness between these two units is of course
supported by the fact that a causal relation can be construed between
them. Yet this causal relation would not be sufficient to create a
construction, were the two units not in sequential proximity.
However, it is not only units produced by one speaker, as in the case
above, which may form a construction, but also units produced by two or
even more than two speakers. The latter case is rare in my data, but it
does exist (cf (14) CAT II below). The interactive production of
constructions by two speakers occurs more often (cf (10) CAR below).
Yet the majority of causal constructions in my data are produced by a
single speaker.
Another means of signaling connectedness is via the prosodic design
of two utterances, in particular the pitch level at the end of the first
utterance. There are two possibilities for signaling connectivity: (i) the
two utterances are produced in one intonation unit, (ii) the two utterances
are produced as two intonation units with the cadence of the first unit
signaling continuity. 5 The latter is found most often in my data in cases
where a causal construction is produced by a single speaker. In these
instances, the fmal pitch movement of the first unit does not reach the
bottom of a speaker's pitch range, i.e. does not end in 'period intonation'
86 Christine Gohl
snoring, i.e. falling asleep, will not be possible; it is not possible because
standing is not a good position for falling asleep. In (1), the knowledge
that is required in order to discover and make sense of the causal relation
springs from cultural conventions rather than from knowledge of, for
example, particular biological constraints upon human beings. The
speaker uses the argument mir musset rauche 'we have to smoke' as an
account for the production of the sentence nix kriegt er me: 'he'll get
nothing more:'. For reasons of politeness one should not smoke while
someone else is eating. Therefore, mir musset rauche 'we have to smoke'
justifies the first part of Erik's utterance (nix kriegt er me: 'he'll get
nothing more') by giving a reason for why he says that Uwe should not
have anything else to eat.
The analysis of these examples brings us back to question (3): what
counts as a causal relation in spoken interaction? When is a construction
a causal construction? When is it a construction with a different, e.g.
temporal relation? Which constructions are used by participants in
communication as a means for producing a causal relationship? How do
we know that participants understand a relation between two or more
adjacent utterances as causal? And how can we, from an analytic point of
view, give a plausible interpretation of this relation that goes beyond
intuitive judgment?
One possibility is to look at the way speakers treat and recipients
respond to the utterance in which the construction in question occurs. In
these cases, we can make use of the 'next-tum-proof-procedure', i.e. use
the participants' own interpretations as strong confirmation of the analytic
interpretation (cf Bergmann 1991: 218; Sacks, Schegloff, and Jefferson
1974: 728-729). There are several instances in my data in which a next
speaker refers to a prior asyndetic causal relation with an utterance that
makes it clear that he or she has understood the preceding construction as
containing a causal relation. In these cases, next speakers (or same
speakers in their next tum) use a causal conjunction, a causal adverb, or a
causal phrase like das ist der Grund warum 'that is the reason why' with
respect to the same relation that has been constructed asyndetically
before. These lexical elements or phrases give the analyst a clear
indication that the causality is indeed present for the participants.
Consider excerpt (1) SMOKING again: Erik's utterance consists of a
10
dispreferred second action (nix kriegt er me: 'he'll get nothing more')
and a following account for this dispreferred action (mir musset rauche
'we have to smoke' ): Erik, although not the addressee of Marie's offer,
rejects her offer to Uwe by implying that if Uwe continues eating, he
(Erik) and the other smokers will not be able to smoke. Anna, in her tum,
Causal relations in spoken discourse 89
to convince him that his problem is his own fear of coming to a decision:
she advises him to stand by his girlfriend and speak openly to his mother.
The caller, however, rejects this by saying that his mother does not want
him to see his girlfriend at all (line 9):
12Psychologist: aha:~
13 (0.5)
14Psychologist: and
15 (0.5)
16Psychologist: why doesn't she like it?
17 (-)
18Caller: well she says h
19 my girlfriend has three children,'
Here the account is only produced after the psychologist has explicitly
asked for it (line 16). Between the caller's utterance (line 9)-an
objection to the psychologist's insisting that a decision is necessary-and
the psychologist's question in line 16, there are three long pauses and two
short utterances which do not pursue any further topical talk. That is, the
psychologist can be heard as pursuing an account from her caller by
providing him multiple opportunities at which he can provide such an
account.
The analysis of such deviant cases leads to the conclusion that
"account giving is not merely an empirically common feature that is
associated with unexpected or unlooked for actions, but is a normatively
required feature of such actions. Since failures to provide accounts attract
either overt pursuits of them or sanctions, we can conclude that the giving
of accounts in such contexts is itself a morally accountable matter"
(Heritage 1988: 135).
In sections 4.2.-4.6., different environments will be described in which
accounts, realized by asyndetic linkage, regularly occur in my data. The
environments will be presented in the order of their frequency of
occurence: accounts after dispreferred seconds are most frequent, then
accounts after assessments, requests, and finally after complaints and
reproaches.
Various conversation analytic studies (e. g. Atkinson and Drew 1979, Ch.
2~ Heritage 1984: 265-269~ Pomerantz 1984) have shown that particular
actions, often organized as first parts of adjacency pairs, make a second
action relevant, i. e. these actions call for a second action by virtue of their
sequential implications. These second actions have alternative, but
nonequivalent, realizations: one realization is preferred, the other
dispreferred. After requests, suggestions, offers and invitations, an
Causal relations in spoken discourse 93
With his utterance in line 4, da komm i mit der schere nicht durch 'I
can't get through with these scissors', Fritz rej ects Anne's suggestion
nimm halt die () schere; 'take the (.) scissors;'. The preferred, and thus
structurally unmarked, second part following a suggestion would be its
acceptance. In such a case no further talk would be necessary. In (6),
however, Fritz offers an account for refusing to follow Anne's
suggestion, i. e. for why he does not put it into action. In providing a
94 Christine Gohl
reason for not taking up his daughter's suggestion, Fritz need not make
his refusal explicit: the action is evident although non-verbal. In this case,
asyndetic linkage is the only possibility for the production of the account
because there is no previous utterance or implication to which a
conjunctional clause could be linked.
Accounts occur after various kinds of such dispreferred second
actions. In excerpt (7), from a doctor-patient interaction, an account
follows a request for information that cannot be fully met by the second
speaker.
The doctor is a specialist in the field of reproduction technological
medicine and the patient, a woman with a hormonal disorder who is
seeking advice about problems that might occur if she wants to get
pregnant.
In lines 1 and 2, the doctor is asking the patient for some information
about her problem. After correcting one of the doctor's assumptions (line
3), the patient starts reporting on earlier treatment, thus producing the
answer that has been made relevant, i. e. sequentially necessary by the
doctor's request for information. She cannot completely comply with the
doctor's request, however (line 9). Immediately following this partly
dispreferred second action, she produces an account, explaining why she
does not have the findings from her first gynecologist yet: der war
einfach noch in urlaub; 'he was still on holiday;'. i.e. there was no way
of collecting the findings or having them sent.
The excerpt in (8) presents an interactive construction of a sequence
with the dispreferred-second-account structure. This construction shows
that recipients are sensitive to dispreferred actions even if a request for
information and a following dispreferred action occur within the context
of a narration and are embedded in a reported speech situation. Five
family members are talking about a television program on the fight
against crime. In the passage transcribed, the topic is undercover agents.
Fritz is telling the story of a man he knows personally because he used to
be a customer in Fritz's shop.
16 (1.5)
17 na diirfe se 10s lege.
, 1Fritz: « »
2 he used to work as a policeman.
3 ie gJ, ludwigsburg. «name of policestation and town»
4 (1.5)
5 and then he went underground.
6 (1.5)
7 only recently his parents came in, (---)
8 and then I asked them well: eh:,
9 how is your son these days e::,
10 «p> he said (.) we don't know anything,>
11Erik: ohyeah
12 they get .h [they get (.) a completely new personal record,
13Fritz: [<<f> we don't know anything, (.»
14Erik: they get a three fifty s e, ((name of a car» (. )
15 and a large bank account,
16 (1.5)
17 and then they can get started. '
While Fritz is reporting on the conversation he had with the parents of the
former policeman (lines 8-13), Erik comes in (lines 11-12) and produces
an account for the parents' inability to give Fritz the information he had
asked for. Erik's utterance explains the production of a dispreferred
second by the characters in Fritz' story: the parents are not able to answer
Fritz's question wie gehts ihrem sohn und so e::, 'how is your son these
days e::,' because their son, after having gone undercover, was given a
completely new identity (die krieget () en vollstandig neue lebenslauf,)
and various other things that cut him off from his former life (lines 14-
17).
The accounts offered after dispreferred seconds in examples (6), (7),
and (8)-whether they follow requests for information (excerpt (7) and
(8» or suggestions (excerpt (6»-all refer to some kind of inability to
"carry out the proposed or required conversational action" (Heritage
1988: 136). In (6), the speaker accounts for why he is not able to use the
scissors for cutting the meat as Anne has suggested; in (7) the patient
accounts for her inability to completely comply with the doctor's request
for information, and in (8) a third speaker accounts for the inability of two
other-distant-speakers to carl)' out the action that has been called for
by the question asked, namely to provide an answer.
Causal relations in spoken discourse 97
Accounts are not only offered after dispreferred seconds; they also
emerge regularly after assessments. 15 These assessments 0 f en t ·
contaIn
lexical markers or phrases that make them 'strong assessments' in the
sense of what Ford has described as 'strong evaluation' (Ford 1994:
548). In German, some of these lexical markers are: besonders
'especially', wirklich 'really', sehr 'very', ganz 'totally', as well as
particular adjectives like abscheulich 'dreadful', wunderbar 'wonderful',
genial 'brilliant' or adjectives in their superlative form, e.g. am
schlimmsten 'worst' and nominal expressions like da war die holle los
'all hell broke loose' .
Accounts after strong evaluative assessments are more frequent in my
data than accounts after assessments that do not express a strong
evaluative stance. This observation leads to the conclusion that it is not
particular lexical items that are responsible for the production of an
explanatory unit, but the action carried out by the evaluative statement,
the assessment.
Let us look at transcript (9). The participants are talking about the
family's cat. The starting point for this discussion is Marie's utterance.
She thinks that the cat is sad because it did not get anything for
Christmas.
'lMarie: I don't want to (say) anything (but) the cat is really sad -
2 she didn't get anything.
3Fritz: «f> just now I gave her something>, (.)
4 «p> she didn't even look at it>.
5Erik: it's the cat's first Christmas, you see -
6 she doesn't know that [you get something at Christmas= -
98 Christine Gohl
With her utterance die hat nix kriegt 'she didn't get anything' Marie
gives an account for her assessment die katz isch ganz traurig- 'the cat is
really sad-' .
The following excerpt illustrates the interactive production of an
assessment-account sequence. The participants, mother and son, are
talking about the car the son has used to drive to his parents' home. This
car is obviously not his own, but belongs to the company he works for.
With her utterance, Marie is doing two things that call for a second action
by the recipient. She is asking a question and is also making an
assessment of the car the son has used to visit his parents: mit dem
schone auto 'that beautiful car'. The content of this question is marked on
the prosodic level as something extraordinary: first, during its production
the utterance increases in volume; second, there is remarkable pitch
movement on auto Jahre, with the pitch falling from a very high level.
Marie's question - which constitutes the first part of an adjacency pair
- calls for an answer. This is produced by Erik immediately after Marie
has finished her tum (ja:: 'yeah::'). The assessment, however, requires a
separate acknowledgment. This is provided when Erik gives a reason for
what Marie has said (line 2). Note that Erik not only reacts to the content
of Marie's utterance, but also to its prosodic design, which conveys a
somewhat surprised and approving attitude towards the fact that Erik was
allowed to drive this beautiful car. This attitude, however, has a slight
ironic undertone, which is taken up by Erik in his utterance: both Marie's
question and Erik's answer are thus produced in a teasing, slightly ironic
style-judging from content and, especially in Marie's utterance, from
the prosody.
Causal relations in spoken discourse 99
In line 3, Erik gives a reason for asking the question ja sind die zigarre
scho was? 'are these cigars okay?'. He asks if the cigars are okay,
because he felt sick after smoking one of them, implying with his
question that the cigars might not be okay. By offering an account, Erik is
giving the recipients some insight into the motive he has for asking the
question.
In the speaker makes a request in the form of an imperative. The
participants are talking about the food, the meat they are having for dinner
in particular.
100 Christine Gohl
Erik asks Uwe if he wants to swap pieces of meat. Immediately after this
request, he offers a reason for requesting: deins isch kleiner 'yours is
smaller'. In his next tum (line 3), Uwe refuses to accept Erik's request
(noi nemm du 'no you take it') and accounts for the production of this
dispreferred second action in lines 3 and 5: i will blo) e schtiick; 'I just
want a piece', and i will doch net solche brocken 'I don't want such big
pieces' .
Excerpts (11) and (12) show that requests-in these cases formulated
as a question and as an imperative-are followed by accounts. Requests
are face-threatening activities because they put some obligation on the
recipient he or she has to do something for the speaker, either in the form
of a verbal action (provide the information requested) or in the form of a
non-verbal action (do the thing requested). This format makes requests
conversationally sensitive actions-and these are likely to be followed by
accounts.
i. e. the persons who are doing the reproaching, offer accounts for
reproaching someone or complaining about something.
In (13) the topic is once again the family cat. Anna is blaming Fritz for
having given his cold to the cat.
From the external, visible fact der nie:st scho de ganze dag 'he's been
sneezing all day', Anna draws the conclusion that Fritz, who has a cold,
must have given it to the cat. She formulates this conclusion as a
reproach: both the content of this utterance and the way it is expressed
(e.g. the term mensch, which is very often used at the beginning or at the
end of an accusatory utterance) give it its reproachful character. Right
after this unit, Anna offers an explanation and a justification for the
conclusion and the reproach: the cat's been sneezing all day. 16
Reproaches are face-threatening activities, i.e. activities which are
likely to threaten both the face of the person the reproach is directed at
and the social relationship between the parties. Offering an account after
such an action makes a reproach more understandable because it reveals
the reasons behind it and gives the person blamed the chance to react to
the reproach in a way which is more customized than a mere refusal
would be: he or she can exculpate him- or herself by referring to the
reason for the reproach. This happens in line 3 in (13): Fritz makes it
clear that he refuses to take responsibility for the cat's cold because the
cat has been sneezing all morning. He implies that it must have caught
the cold elsewhere.
It will not come as a surprise that accounts after reproaches emerge
not only after a speaker has blamed some other participant for something,
but also after complaints about a third party. Consider the following
transcript. The content is the same as in CAT I: three family members are
talking about the meal they have just had and are considering the 'poor
cat', who seems to be sad because it did not get anything.
102 Christine Gohl
'lMarie: I don't want to (say) anything (but) the cat is really sad-
2 she didn't get anything.
3Fritz: «f> just now I gave her something>, (.)
4 «p> she didn't even look at it>.
5Erik: it's the cat's first christmas you see -
6 she doesn't know that [you get something at christmas= -
7?Marie: [what's going on;
8Erik: next year you'll probably have to give her something;'
With his utterance des hat er net emol aguckt 'she didn't even look at it',
Fritz complains about the cat's behavior with respect to his action: he has
given the cat some food, but the cat did not appreciate it. Erik and Marie
account for this behavior with the utterances that follow. Thus, once
again, the causal relation is constructed interactively. Fritz produces an
utterance which is treated by a second speaker, Erik, as a consequence:
the cat does not appreciate the food because it is her first Christmas.
Therefore, you cannot expect her to react in a particular way. The second
part of Erik's utterance (die weifJ gar net dajJ mer do was kriegt- 'she
doesn't know that you get something at christmas-') provides another
potential reason for the cat's behavior. At the same time, another speaker
(probably Marie) comes in and completes another explanatory unit (die
weifJ gar net was los isch; 'she doesn't know what's going on;') for what
Fritz has said earlier (des hat er net emol aguckt 'she didn't even look at
it' ). Both explanatory units, Erik's and Marie's, contain reasons for the
cat's behavior, as well as consequences with respect to the first part of
Erik's utterance (des isch doch dere katz ihr erschtes weihnachte- 'it's
the cat's first christmas-'): hence, because it is the cat's first Christmas,
she does not know that you receive something on this occasion and she
does not know what is going on. The interactive nature of this production
of accounts shows that both speakers, as recipients, consider the action of
Causal relations in spoken discourse 103
blaming the cat for not appreciating the food as an action that calls for
accounting.
5. Concluding remarks
This study shows that speakers may give reasons without usmg a
conventionalized causal connective element.
Different sequential environments were investigated in which speakers
produce such asyndetically linked causal relations in order to give
accounts. These environments are constituted by different actions which,
for various reasons, call for accounting. One group of these activities-
dispreferred seconds, requests, and reproaches-is particularly sensitive
to the social relation between participants, either because these activities
are face-threatening or because they put some obligation on the recipient.
The second group, including first (strong) assessments, is also socially
risky. Accounts following strong assessments make the production of a
preferred second more likely because they support what is often a highly
subjective attitude on the part of the speaker and thus make it easier for
recipients and possible next speakers to understand the speaker's stance
and react to it in a preferred format.
Ford has shown in her studies (1994 and this volume) that accounts or
explanations and justifications often emerge in contexts involving some
form of contrast, which mayor may not be marked, for example, by
contrastive markers or negation particles. The analysis of the asyndetic
causal constructions in my data, however, shows that only about half of
the accounts occur in contexts involving contrasts. This means that
accounts have a much broader context of occurrence. In fact, much
Causal relations in spoken discourse 105
[] overlap
[]
(.) micro-pause
(-), (--), (---) short, middle or long pauses of approx. 0.25-0.75 seconds, up to ca.
1 second
(2.0) estimated pause of more than 1 second
and=uh slurring within units
lengthening, according to its duration
uh,ah, etc. hesitation signals, so-called "filled pauses"
? intonation phrase-final: high rise
intonation phrase-final: mid-rise
intonation phrase-final: level pitch
106 Christine Gohl
Notes
dialogic nature." For the interactive function of the particles doch, eben, and
ja in spoken discourse see also Liitten (1979).
7. Occasional remarks on the possible causal meaning of modal particles can be
found for doch (Feyrer 1998: 187-192), auch (Paul 1966), halt (Abraham
1983, Hinrichs 1983) eben andja (Abraham 1983).
8. This investigation is based on the analysis of approximately 7 hours of
spoken German comprising about 4 1/2 hours of informal conversational
data, i.e. family conversations, 1 1/2 hours of interactions in a medical
setting, and 3/4 of an hour of radio phone-ins. Most of the data was recorded
in the Swabian area of Southern Germany. I wish to thank Ute Lacher-
Laukenikat and project C4 of the SFB 471 for making their data available to
me.
9. For the interpretation of asyndetic conditional structures, i.e. paratactic
conditionals, see Thumm (1999).
10. For the notion of 'dispreferred second' and a more detailed discussion of
accounts after dispreferred second actions, see section 4.2.
11. In fact, in excerpt (3), the utterance'.h .h Ooh I've no idea:.' also constitutes
an account: M accounts for her failure to provide J with the information
requested (Heritage 1984: 250). Compare also excerpt (6) in section 4.2.
12. For more details, including the marked-unmarked distinction, and a survey
of the most important issues concerning preference organization, see
Levinson (1983: 332-345). On the use of causal markers in English and
Japanese conversations in dispreferred environments, see Ford and Mori
(1994).
13. For another example illustrating the production of a dispreferred second-
account-sequence after a suggestion, see excerpt (1) SMOKING. While in
excerpt (6) SCISSORS the declining component is not verbalized, in excerpt
(1) SMOKING the speaker formulates the declining of another speaker's
suggestion in a very explicit and highly marked way.
14. 'Pock-pock' is presumably the name for a kind of a fowl.
15. For a discussion of assessments and their sequential implications in spoken
German interaction, see Auer and Uhmann (1982). Some of their transcripts
show that second speakers may explicitly ask for the reason of an assessment
given by a first speaker if the latter does not account for the assessment him-
/herself (Auer and Uhmann 1982: 8-9).
16. The formulation der nie:st scho de ganze da:g ('he's been sneezing all day')
falls into the category of Extreme Case Formulations (Pomerantz 1986).
Such formulations are used by speakers when they justify complaining about
something or somebody to "legitimize a complaint and portray the
complainable situation as worthy of the complaint" (Pomerantz 1986: 228).
References
Abraham, Werner
1983 Starke Satzverkniipfer-Sinnsog und Partikelenquete. In: Harald
Weydt (ed.), Partikeln und lnteraktion, 25-39. Tiibingen: Niemeyer.
108 Christine Gohl
Hinrichs, U.
1983 Konnen Abtonungspartikel metakommunikativ funktionieren? In:
Harald Weydt (ed.), Partikeln und lnteraktion, 274-290. Tubingen:
Niemeyer.
IckIer, Theodor
1994 Zur Bedeutung der sogenannten 'Modalpartikeln'. Sprach-
wissenscho:ft 19: 374-404.
Levinson, Stephen
1983 Pragmatics. Cambridge: Cambridge University Press.
Lindner, Katrin
1991 'Wir sindja doch alte Bekannte' The use of Germanja und doch as
modal particles. In: Werner Abraham (ed.), Discourse Particles,
163-201. Amsterdam: Benjamins.
Lutten, Jutta
1979 Die Rolle der Partikeln doch, eben, und ja als Konsensus-
Konstitutiva in gesprochener Sprache. In: Harald Weydt (ed.), Die
Partikeln der deutschen Sprache. 30-38. Berlin: de Gruyter.
Pasch, Renate
1997 WElL mit Hauptsatz-Kuckucksei im DENN-Nest. Deutsche
Sprache 3: 252-271.
Paul, Hermann
[1966] Deutsches Worterbuch, 6.A. Tubingen: Niemeyer.
Pomerantz, Anita
1984 Agreeing and disagreeing with assessments: some features of
preferredldispreferred turn shapes. In: 1. Maxwell Atkinson and
John Heritage (eds.), Structures of Social Action, 57-101.
Cambridge: Cambridge University Press.
Pomerantz, Anita
1986 Extreme Case Formulations: A way of legitimizing claims. Human
Studies 9: 219-229.
Sacks, Harvey, Schegloff, Emanuel A., and Jefferson, Gail
1974 A simplest systematics for the organization of turn-taking for
conversation. Language 50/4: 696-735
Scheutz, Hannes
1998 Weil in spoken German. Presentation given at the conference on
'Cognitive and Discourse Perspectives on Cause, Condition,
Concession and Contrast'. Oktober 1998. UniversiUit Konstanz.
Schegloff, Emanuel A.
1996 Confirming allusions. American Journal of Sociology 102/1: 161-
216.
1997 Practices and actions: Boundary cases of other-initiated repair.
Discourse Processes 23: 499-549.
Schleppegrell, Mary J.
1991 Paratactic BECAUSE. Journal ofPragmatics 16: 323-337.
110 Christine Gohl
Thumm, Markus
1999 How to do some things without 'if. Eine empirische Untersuchung
zu parataktischen Konditionalkonstruktionen im gesprochenen
Englisch. MA-thesis, Department of Linguistics, University of
Konstanz.
Uhmann, Susanne
1998 Verbstellungsvariation in Weil-Satzen: Lexikalische
DifIerenzierung mit grammatischen Folgen. Zeitschrift .fur
Sprachwissenschaft 17/1: 92-139.
Constructions with if, since, and because:
Causality, epistemic stance, and clause order*
1. Introduction
(1) If (as you say) he is in town now, maybe we should phone him?
(2) If I take my pills regularly, the symptoms go away.
In (1), imagine that the hearer has said earlier in the conversation that
the relevant person is in town now. In the speaker's reference to this
previous statement, she could use since instead of if, with quite similar
112 Barbara Dancygier and Eve Sweetser
effect. (2) can naturally be interpreted to mean that taking pills regularly
makes the symptoms disappear-a causal interpretation related to
readings of because and since. There is thus a significant degree of
overlap among the uses of the conjunctions involved. But the potential
for such interpretations will often depend on factors other than the
conjunction itself; notice that when is a plausible substitute for if in (2)
but not in (1), suggesting that the lexical and constructional contexts are
important in determinining possible construals of the relations between
conjoined clauses.
In positing functional overlap between these conjunctions, we are by
no means suggesting that they are synonymous--or even that they have
specific senses which are synonymous. On the contrary, we will try to
specify the source of the similarities as well as to clearly delineate the
differences. Both similarities and differences are best accounted for not
by focusing only on the conjunctions themselves, but by describing the
range of constructions they participate in. The main goal of our analysis
is thus to distinguish aspects of constructional meaning which are
involved in various uses of each of the conjunctions, and to show how
particular formal parts of constructions combine with each other to
contribute to varying interpretations of the whole. 1
We will bring up three relevant parameters in trying to compare and
contrast the four kinds of constructions. First, we will consider the role
of causal models in interpretation. Causality is involved in the
constructions in a variety of ways and on various levels of
interpretation. Constructional aspects of meaning, and the cognitive
domains in which these constructions are interpreted, can be as
important as the choice of conjunction, in bringing causality into the
interpretation. The conjunctions will also be compared in terms of the
type of mental space set-up they engage in; this is often also specifically
signalled by the use of verb forms marking negative epistemic stance.
Finally, we will review the constructions with respect to their roles in
backgrounding or foregrounding elements of interpretation in different
clause order configurations.
Our analysis of meaning will be framed in terms of Mental Spaces
theol)' (Fauconnier 1985, 1997), an extremely general formal theory
which provides mechanisms for talking about the ways in which we
connect cognitive structures with each other. To some extent, traditional
treatments of conditional semantics in terms of Possible Worlds have
aimed at capturing similar insights. Somewhat in the way that an if-
clause can be said to set up a Possible World within which a then-clause
Constructions with if, since and because 113
We will start this section by outlining types of mental spaces which are
particularly relevant to our interpretation of the use of conjunctions. We
will then show how these types of mental space set-ups correspond to
levels of interpretation of constructions and how causal relations are
established at each of these levels.
114 Barbara Dancygier and Eve Sweetser
Figure 1. If his computer gets repaired, he'll finish the paper by Friday.
Constructions with if, since and because 115
(4) If! don't see you before Thursday, have a good Thanksgiving!2
(5) If he finished the paper by Friday, his computer must have gotten
repaired.
(6) Chris wants you to meet her fiance, if that's the right word for
him.
Constructions with if, since and because 117
the speaker perhaps thinks that Chris is using fiance as a euphemism for
'live-in partner.' Crucially, the speaker is not here construed as
negotiating either the facts of the case (the content), or conclusions to be
drawn from them in the epistemic domain; nor is she conditionally
performing the speech act of stating that Chris wants the hearer to meet
a particular individual. She is apparently metalinguistically commenting
on, or presenting as merely conditional, her use of the label fiance to
refer to that individual.
In this case, we need a somewhat more complex mental space
structure, which we have referred to (Dancygier and Sweetser
1996,1997) as a metalinguistic space. A metalinguistic space
(generically diagrammed in Figure 3 below) consists of a speech-
context space, within which there is a space of content being discussed
and a space of linguistic forms available for referring to that content.
Metalinguistic conditionality concerns mappings between the content
space and the form space within a particular context. Here, the speaker
constructs a particular form-content mapping (uses fiance to refer to
Chris' partner), and then sets that in the context of a particular
metalinguistic space by calling the mapping into question explicitly (if
that's the right word). By implication, such a conditional usage at least
suggests that other metalinguistic spaces could be imagined, with
different form-content mappings; however, in (6) there is no explicit
attempt to evoke other specific mappings (e.g., partner).
(7) .Joe turned down the stereo because Sam was studying.
(Causal relation is between state of affairs described in P and
event described in Q, i.e. between contents of the clauses.)
(8) Sam is (must be) studying, because Joe turned down the stereo.
(Causal relation is between speaker's knowledge about content
ofP and speaker's conclusion about content ofQ.)
Constructions with if, since and because 119
(9) Could you turn down the stereo, because I'm trying to study.
(Causal relation is between the contextual state expressed in P
and the speech-act performance of the request in Q.)
and Spada (1993) observe, children learning their first language may in
fact hypothesize that principles of iconicity override meanings
introduced at the lexical level and produce sentences like You took the
towels away because I can't dry my hands. In such cases, the strategy
used is to present events in their sequential iconic order (the towels
disappeared from the bathroom first, and consequently the child had
nothing to dl)' her hands with later on). The child needs to learn not only
the causal meaning of because, but also the way to use it
constructionally to mark causal connections correctly. 5
In many content-domain causal conjunction examples, only the
causal relation itself is asserted, while the two events or situations
between which the causal relation holds are presupposed background. In
the classic I'll do it because I want to, and not because you told me to,
the speaker does not assert either that she will do as she is told, or that
she wants to do so, or that she was told to do so: but she is necessarily
asserting that her actions are caused by her wishes and not by the
addressee's commands. .[fclauses generally set up a new mental space
distinct from the Base space of the speaker's assumed construal of
"reality", but because-clauses do not set up such a space. Main clauses
of content-domain because constructions decribe situations which may
already be known to hold in the interlocutors' current base space. 6 The
because-clause then offers a causal explanation of how that situation
came about, also not necessarily on the basis of brand-new information.
The causal relationship is at the core of the message.
Since is similar to because in being semantically causal, and in that
both are often said to presuppose the truth of their complements. These
differences interact with other parameters, such as information structure
and typical clause order. Since will be discussed in more detail in
sections 3 and 4, where we focus on epistemic stance and information
structure.
The conjunction if presents an interesting case. Unlike because, it is
not lexically causal. In the content domain, however, if-sentences
typically express causal or enablement relations among events. (12), for
example, would typically convey the message that the conditioner will
cause the hair to get softer.
Dancygier (1993, 1998) has argued that causal meanings enter the
interpretation of content conditionals via conditionals' primary function,
prediction. 7 Prediction in conditionals is a type of reasoning which
consists in setting up a hypothetical (typically future) mental space and
attempting to predict its consequences based on knowledge of typical
cause-effect chains and general world-knowledge. In (12) above, the
speaker sets up a space in which the hearer applies the recommended
conditioner, and predicts in that space that the hearer's hair will become
softer as a result. Such causal interpretations tend to arise in conditionals
which set up content mental spaces, where causation-based predictions
can be made on the basis of assumptions of sequential correlations
among events or states of affairs and general knowledge. But the
interpretation of speech-act or epistemic conditionals also involves
setting up causal relations: the speaker of Since you're a linguist, what
do you think of Chomsky? constructs relationships of causality and
enablement as part of the background to the act of asking a question.
The most notable formal feature distinguishing English predictive
conditionals is the combination of verb forms: for future-reference
conditionals, a present-tense verb form is used in the If-clause, and the
modal will in the main clause (cf Fillmore 1990a,b). This verb form
pattern characterizes the whole sentence (not the two clauses in isolation
from each other) as representing a certain kind of reasoning.
Predictiveness (and, consequently, causality) is part of the interpretation
of certain If-constructions as wholes, rather than of the conjunction if or
any other lexical element of the sentence. Let us also note that such
constructions do not assert any of the clauses. The If-clause cannot
express an assertion because the conjunction sets up a hypothetical
mental space, different from the base space where assertions could be
made. The main clause describes the predicted result in that same
hypothetical space, so it is not asserted either. What a predictive
conditional asserts is the causal dependency and correlation between the
events or states of affairs described by its clauses.
To sum up, content-domain conditionals are interpreted causally as a
result of predictive uses which are explicitly marked by formal markers
such as verb forms. This is one of the basic differences between
sentences with because, and those with if. In the former, the conjunction
itself introduces causality; in the latter, constructional meaning is the
source of the causal interpretation. Though the constructions with
because and if are different in a number of respects (verb forms, clause
order, mental space set-up, to mention just a few), they can both convey
Constructions with if, since and because 123
example, a sentence like If Tim went to the embassy, he has his visa
documents can have two epistemic interpretations. If the speaker
believes that Tim was going to obtain his visa documents at the
embassy, she may use the information about Tim's visit there to infer
that he now has his documents. If, on the other hand, the speaker
believes that acquiring the documents enabled Tim to go to the embassy,
she concludes that he had them prior to the visit. In either case,
knowledge of cause-effect links in the content domain is used in a
reasoning process carried out in the epistemic domain.
Causal models can thus participate in the construction of meaning on
various levels. This is due not only to our understanding of causality,
which allows us to locate it in various types of mental spaces, but to our
understanding of the relations among those spaces, and, more generally,
to our understanding of speech interaction. This understanding is
dependent on what we may see as a nested complex of mental spaces.
At the center of this complex is some representation of the content of
what is said: events or states of affairs which are referred to by the
speaker. The content space need not make any reference to the speaker's
reasoning processes or to the speech exchange itself (although these
may also be addressed as explicit COlltent: I am asking you why you need
an extension on your paper). However, reasoning processes relevant to
the speech interaction almost inevitably make reference to the content
space: the speaker may, for example, be reasoning specifically about the
structure of the content space. In He must be home; I see his coat, the
speaker's use of must refers to her understanding that her reasoning
processes are under compulsion, not that the described subject's actions
are compelled; but the reasoning processes about the act of returning are
necessarily based on the speaker's understanding of the content world,
where both the return and the coat's presence reside.
The structure of the speech exchange will necessarily involve both
the idea that some content is expressed, and the understanding that the
speaker has reasoning processes related to that content, plus the added
understanding that there are a speaker and a hearer interacting in the
current speech setting, with all the social interactional issues attendant
upon conversational exchange. As we have seen, each of these levels of
mental space structure may involve causal models appropriate to the
level. But the nesting effect explains how different causal models can
enter the interpretation at the same time. This includes the possibility of
exploiting a cause-effect chain wherein P causes Q, as the necessary
Constructions with if, since and because 125
3. Epistemic stance
(13) If/Since he's (so) hungry (as you say he is), he'll want a second
helping.
Constructions with if, since and because 127
Sentences like (13) certainly suggest that the truth of the proposition
('he's hungry') has been asserted in earlier discourse. Rather than
setting up a novel space, they evoke a space already contextually
accessible. However, referring to a space available in previous discourse
does not necessarily mean that the speaker always aligns herself with its
content to the same degree. As argued by Dancygier (1998), ~f indicates
non-commitment to the clause it marks: but non-commitment covers a
wide range of possible attitudes, from strong disbelief to near-
commitment. Even in an example such as (13), where ~fand since seem
interchangeable, since sounds more completely positively aligned than
if, although in the presence of other strong indications of positive
attitude, the degree of non-commitment conveyed by if may be
negligible.
So there are important differences between if and since. If maintains
a neutral stance, while since takes a positive one. Also, since, unlike ~f,
has explicitly temporal senses. However, in appropriate contexts, there
is predictable functional overlap between the two conjunctions-
predictable, in the sense that it falls out naturally from a mental space
analysis of the constructions and contexts in question. But in defining
the possible contexts of overlap, we need a clear understanding of the
factors which make some particular assertion or claim accessible to, or
assertable by, some particular participant in the speech interaction. What
licenses a speaker to treat a particular content as already accessible to
the hearer, or as not being subject to questioning by the hearer? We shall
make use of the standard constructs of speech act theory to define
contexts where if and since are naturally isofunctional. As will become
evident, the relevant factors have a general application to the analysis of
the relationship between space-builders and informational structure:
space-building background information has to be brought up in a way
appropriate to its status in the speech context.
When fills a genuinely temporal space-building function, as does
since when used to mean 'from some past time to the present.' Non-
temporal since, however, builds up causal relations; it states a causal
relationship between the space which it evokes (rather than constructs)
and the main clause. An ~clause may, as in (13), merely re-enact the
building of an already accessible space--or it may build a truly novel
space, as in many predictive constructions. A causal since-clause,
however, always evokes a space which is presupposed by the speaker,
as accessible already to the audience.
128 Barbara Dancygier and Eve Sweetser
(16) Since you're wearing those shoes, we'd better not go to the
beach.
(17) If/#Since it's not rude to ask, what made you get interested in
linguistics?
(since would be acceptable only if the hearer had already made it
plain that the question was acceptable.)
(18) If/#Since you don't mind, could you hold this stack ofbooks for a
moment?
(since is acceptable only if the hearer has indicated he doesn't
mind.)
(20) Since/#If I'm stranded without a car, could you drop me at the
BART station?
If for some reason the speaker were not the most knowledgeable person
about her own access to a car, then if would seem more acceptable in
(20). (Imagine a situation where H, a service station employee, has just
told her that her husband picked up the car an hour ago, while she was
expecting to pick it up and drive home in it.)
Speaker and hearer are each presumed to be authorities on their own
states of mind and body, and not on each other's. Hence the speaker
should normally use since when evoking her own mental and physical
states, and ifwhen referring to the hearer's:
(21) If/#Since you're really tired, maybe we should stop work now.
(since is acceptable if the hearer has himself indicated his
tiredness. )
(24) Since he's phoning from London, his plane must have been
delayed.
(Epistemic: the presupposed knowledge of the phoning causes the
conclusion about the plane's delay.)
(26) Since you're so picky, I'll just put another "trout" on the grill.
(Metalinguistic: The addressee has just corrected the speaker's
use of the term salmon to refer to the wrong species of fish. )
(29) Q ifP I'll work at my office if the home computer breaks down.
some sense pre-built by the discourse; (27) could respond to What 'II you
do ifyour home computer breaks down?
you do if the home computer breaks down? Its major assertion is the
identity of the conditional relation under which the working in the office
would happen; the main clause content is already "on the floor." In
mental space terms, (29) begins by expressing a prediction Q, but in an
intonation pattern which makes it clear that Q is not being
independently asserted, and is therefore unlikely to belong to the base
space. The focal stress (marking new information), and the final
intonation fall (marking the end of an assertion), happen only once we
are identifying the space P within which Q is to be placed. Q is "old
information," while its space-setting is not.
So (27)-(29) correspond to different sequences of instructions for
building what are rather similar final mental space constructs. (27) is the
"canonical" sequence of setting up a space and then making a prediction
within it. (28) might be said to involve at least a potential pre-built P-
space, to which the utterance adds Q, subsequently confirming with an
~clause that yes, P was the intended space for elaboration by Q. (29)
could be seen as involving a presupposed Q, and asserting the
alternative space structure which links that Q with P rather than with ,-..;P.
These differing sequences of instructions are useful in different
contexts, just the way physical instructions to construct the same object
might usefully differ depending on which components of the object
were already pre-assembled or ready to the addressee's hand.
The three space set-up options presented in (27)-(29) differ in
foregrounding or backgrounding causal relations. (27), being a typical
predictive conditional, asserts the causal relation (rather than any of the
clauses) in the way discussed above. (28) is different in that it initially
asserts Q, and thus pushes the nature of the relation between P and Q
into the background. In fact, as Dancygier (1998) claims, this clause
order in conditionals is perhaps most common among speech act and
metalinguistic conditionals, which often perform a speech act in the
current space, and use the ~clause to negotiate the background structure
of that space. In (29), the stress on identifying the space in which the
"old" prediction is valid gives additional focus to the causal relation on
which the prediction is based. Dancygier (1998) argues that this clause
order and intonation pattern is available only to predictive, causality-
asserting conditionals, while all the other types of space-settings are
excluded. (29) thus represents a clause pattern which constructionally
requires causality to be asserted. 11
Similar clause order options are available for because-constructions,
but the preferences are distinctly different. While ~clauses typically
Constructions with if, since and because 135
precede their main clauses, clauses with because typically follow theirs.
This regularity (observed by Ford 1993) is influenced by the fact that
because-constructions do not set up new spaces, but establish causal
relations in the base space. It is thus common for these constructions to
first mention the fact requiring explanation, and then establish the cause
of the state of affairs in question. Consequently, the main contrast
among because-sentences is between two intonation contours available
in the Q because P pattern.
Sweetser (1990, 82ff) discusses parallel examples involving causal
conjunction, as examples of Chafe's (1984) generalization about
"bound" (commaless) and "unbound" (comma) intonation: comma
intonation demands an interpretation involving assertion of both the
main clause and the causal connection between the clauses, while
commaless intonation is readily given an interpretation which
presupposes the assertion (in this case, prediction) embodied in the main
clause, and newly asserts only the causal connection between the
clauses. Thus (30) is likely to be understood as asserting that the reason
for Anna's love is Victor's resemblance to her first love, but
presupposing that Anna loves Victor; (31), on the other hand, asserts
that Anna loves Victor, as well as the reason why.
(30) Anna loves Victor because he reminds her o..f herfirst love.
(32) I love you because you're sweet, not because you're rich.
(35) #He invited me since I live next door, not since I'm his boss.
(36) It's because you're sweet that I love you, not because you're rich.
(37) It's when Sue arrives that we'll eat, not now (/not when Joe
arrives).
(38) It's after lunch that they had coffee, not before dinner.
(39) #It's since I live next door that he invited me, not since I'm his
boss.
Various causal and adversative conjunctions fall on one side or the other
of this contrast: though and while behave like since, and phrases like in
spite of the fact that or despite the fact that behave more like because.
Interestingly, since in its temporal sense does allow assertion and focus
of the temporal relation:
Constructions with if, since and because 137
(40) It's since the pollen came out that I've been sniffling, not since I
moved to the new apartment.
(41) I've been sniffling since the pollen came out, not since I moved to
the new apartment.
5. Conclusion
Notes
6. Or, of course, in whatever space is currently being added to; the crucial fact is
that no new space need be set up when because is used.
7. For discussion of the relationship of then to predictive function, see
Dancygier and Sweetser (1997), Schiffrin (1992), and Iatridou (1991, 1994).
8. Conditionals with distanced form have been discussed extensively as
"counterfactual" or "unreal". The discussion of this issue exceeds the limits of
the present paper. The reader is referred to Sweetser 1990, Dancygier 1993,
1998, Dancygier and Sweetser (1996, 1997). For reasons why distanced
conditionals should not be treated as counterfactual, see Comrie (1986); for
discussion of the crosslinguistic connection between past tense forms and
distanced forms, see James (1982) and Fleischman (1989).
9. Cf. Dancygier and Sweetser (1996, 1997), for discussion of predictive and
non-predictive epistemic and metalinguistic conditionals, and for analysis of
reasons why content conditionals are necessarily predictive and speech-act
conditionals necessarily non-predictive (see also Sweetser (1990), Dancygier
(1998).
10. Cf Haiman 1980, 1986; also Ford 1993 on because vs. if.
11. All the patterns just discussed are also found among even if constructions,
with the same priorities and restrictions involved.
12. As Elizabeth Couper-Kuhlen has pointed out to us, this incompatibility with
non-comma intonation brings us back to Sweetser's (1990) original concern
about the preference of since for non-content uses, and raises doubts as to
whether examples like (23) should really be treated alongside other content
examples.
13. For discussion of parallels with paratactic conjunction, see Lakoff (1971),
Grice (1978), and Sweetser (1990).
References
Akatsuka, Noriko
1986 Conditionals are context-bound. In: Traugott et al. (eds.), 333-52.
Chafe, Wallace L.
1984 How people use adverbial clauses. BLS 10: 437-449.
Comrie, Bernard
1986 Conditionals: a typology. In: Traugott et al. (eds.), 77-99.
Dancygier, Barbara
1998 Conditionals and prediction: time, knowledge, and causation in
conditional constructions. Cambridge: Cambridge University Press.
1992 Two metatextual operators: negation and conditionality in English
and Polish. BLS 18: 61-75.
1993 Interpreting conditionals: time, knowledge and causation. Journal of
Pragmatics 19: 403-434.
Dancygier, Barbara and Eve Sweetser
1996 Conditionals, distancing and alternative spaces. In: Adele Goldberg
(ed.). Conceptual structure, discourse and language, 83-98. Stanford
CA: CSLI Publications.
1997 Then in conditionals. Cognitive Linguistics 8/2: 1-28.
140 Barbara Dancygier and Eve Sweetser
Davis, W.
1983 Weak and strong conditionals. Pacific Philosophical Quarterly 64:
57-71.
Fauconnier, Gilles
1985 Mental spaces. Cambridge MA: MIT Press [Reprinted 1994,
Cambridge: Cambridge University Press].
1996 Analogical counterfactuals. In: Fauconnier and Sweetser (eds.), 57-
90.
1997 Cognitive mappings for language and thought. Cambridge:
Cambridge University Press.
Fauconnier, Gilles and Eve Sweetser (eds.)
1996 Spaces, worlds and grammars. Chicago: University of Chicago
Press.
Fillmore, Charles J.
1990a The contribution of linguistics to language understanding. In: Aura
Bocaz (ed.). Proceedings of the first symposium on cognition,
language, and culture, 109-128. Universidad de Chile.
1990b Epistemic stance and grammatical form in English conditional
sentences. CLS 26: 137-162.
Fillmore, Charles J., Paul Kay, and Mary Catherine O'Connor
1988 Regularity and idiomaticity in grammatical constructions: the case of
let alone. Language 63/3: 501-538.
Fillmore, Charles and Paul Kay
1999 Grammatical constructions and linguistic generalizations: the what's
X doing Y? construction. Language 75/1: 1-33.
Fleischman, Suzanne
1989 Temporal distance: a basic linguistic metaphor. Studies in Language
13/1: 1-50.
Ford, Cecilia E.
1993 Grammar in interaction: Adverbial clauses in American English
conversations. Cambridge: Cambridge University Press.
Goldberg, Adele
1995 Constructions. A Construction Grammar approach to argument
structure. Chicago: University of Chicago Press.
Grice, H. P.
1978 Further notes on logic and conversation. In: Peter Cole (ed.), Syntax
and Semantics, Vol. 9: Pragmatics, 113-127. NY: Academic Press.
Haiman, John
1980 The iconicity of grammar: isomorphism and motivation. Language
56: 515-40.
1986 Constraints on form and meaning of the protasis. In: Traugott et al.
(eds.), 215-227.
Hom, Larry
1984 Toward a new taxonomy for pragmatic inference: Q-based and R-
based implicature. In: Deborah SchifIrin (ed.), Meaning, form and
use in context: linguistic applications (GURT 1984), 11-42.
Washington, D.C.: Georgetown University Press.
Constructions with if, since and because 141
Estrella Montolio
1. Preliminaries
1999). The first group, which we shall call the affirmative group (or
ACCCs), is made up of constructions such as:
The hypothesis put forth here is the following: whereas the construction
if p designates a very broad and generalized condition, roughly
paraphrasable as 'in the event that p,' CCCs express highly specific
Complex conditional connectives 149
'provided that' and a menos que/a no ser que 'unless' convey unique or
exceptional circumstances that will allow (or block) the event expressed
in the main clause to occur. This explains why such CCC-constructions
cannot be iterated in coordinate structures:
The reason for the unacceptability of the iterated CCCs lies in the fact
that it is simply not possible to present the only circumstance under
which something will or will not happen and then add another
exceptional condition. In short, CCCs disallow the presence of other
exceptional conditions.
situation by which q will not be realized. 7 That is, NCCCs convey the
instruction [q only if not pl. Hence, NCCCs convey something to the
effect of "in any situation other than p." In other words, NCCCs express
the exceptional circumstance by which the proposition expressed by the
main clause will not be fulfilled. Both types of CCCs express tff
relations. Both types of CCCs present a modal and argumentative value,
since they instruct the hearer whether a given circumstance is favorable
or disfavorable to the proposition in the main clause.
The perspective that the speaker adopts in the face of the possibility
of q 's realization is therefore quite different: while ACCCs signal that q
can only be realized if p, NCCCs express that q will not be realized only
in the case of p. This is what determines the affirmative character of the
first group and the negative character of the second. ACCCs convey the
necessary basis for something to occur; NCCCs express the
circumstance under which something will not be realized. Dancygier
(1985: 68) proposes quite insightfully that what unless really denies is q,
not p. Her proposal seems to be clearer than the one advanced by Geis
(1973): except if p, q. To observe a case in which the main clause
displays negative polarity (for instance, example (4)), we can
extrapolate from Dancygier's observation as follows in (13b):
In other cases, the assertion hidden in the protasis is less explicit, but it
is information that is present in the context of the speech event (the term
context should be understood in the broad sense of Sperber and Wilson
19962) . The thematic or given nature of such rrotases is demonstrated
by the fact that the protasis cannot be postposed:
Nevertheless, the examples in (28) and (29) show that CCCs are
incompatible with entonces 'then' even when they appear in a preposed
position. Such a fact invites the question of just what the function of
then in a conditional is. There seems to be agreement that one of its
main functions is to act as an anaphoric emphatic correlate to the
protasis, i.e., something like a resumptive pronoun (Dancygier 1998;
Dancygier and Sweetser 1997; latridou 1992; Montolio 1991b, 1999).
One might say that the use of then constitutes an attempt on the part of
the speaker to insure that the interlocutor retains the hypothetical
framework on which the following clause or utterance is based. This is
clearly the case of conditionals in which various protases are linked to
one apodosis:
B: J.~i,
a condici6n de que acabes los deberes. Entonces te llevare
al cine.
'Provided that/so long as you finish your homework. Then I'll
take you to the movies. '
(35) E si por aventura el Rey fuere de tan gran piedat que 10 quiere
dexar vivir, non 10 pueda facer a menos que no Ie saque los ojos.
(Fuero Real, 350)
Complex conditional connectives 161
The greater force of (39b) lies precisely in that the threatening action
(i.e., kill the boyfriend) is expressed in an affirmative factual manner.
Complex conditional connectives 163
This seems to be more effective than posing the matter in the converse
terms, as in (39a), where the threatening action is negated and the
restriction on this negated threat appears later.
The same pragmatic explanation can be applied to resolve the greater
or lesser inadequacy of utterances (36b) and (36d). As can be seen,
exchanging connectives causes the speech acts to be exchanged too.
Thus, (3 6b) seems to be a much weaker version of a threat than that
formulated in (36a), though it also could be interpreted as an
unconvincing stimulus. Insofar as the clear stimulating promise in (36c)
is concerned, it is converted directly into a threat as soon as its
connective is changed (36d).
Besides the illocutionary value generally equated with CCC-
structures, we can also pose the question of what their pragmatic
behavior is as compared to that of if-clauses. Also worth considering is
the question of whether there exist differences between the two subsets
of CCCs from such a point of view. An if-clause can appear in any
context that a CCC can, and in fact, speakers prefer to use the unmarked
connector if even when the insertion of a CCC is possible. The very
subtle differences that these constructions convey with regard to if (and
if ... not) are so faint that most speakers probably do not consider them
to be very operative. Furthermore, in the Romance languages, CCC-
clauses trigger the subjunctive mood, which according to the literature
on the matter is in the decline. This seems to be why CCC-ciauses are
more characteristic of formal, written registers than of spontaneous, oral
ones.
CCC-structures are, in effect, noticeably less frequent than if-clauses
in oral and spontaneous registers. In this sense, we can see the
correspondence between ontogenesis and linguistic universals, because
such constructions are acquired much later than if-structures. Although
most authorities on the matter of the acquisition of conditionals only
deal with if-clauses, Wing and Kofsky (1981) point out that just as
common sense would lead us to believe, CCC-structures are among the
latest to be acquired. Given their formal complexity and the specificity
of the condition they introduce, CCC-structures surface in child
language at a later stage than if-clauses and other types of synthetic
conditional structures, such as paratactic structures (McCabe et al. 1983,
Wing and Kofsky 1981). In the case of NCCCs, this late appearance
seems to be even more justified. NCCCs imply a greater cognitive-
processing complexity in that to the uncertainty of the truth value they
add a negative relationship between the two clauses.
164 Estrella Montolio
My claim is that no matter what the speech act may be, ACCCs and
NCCCs are pragmatically very different from one another. When
speakers choose to use an unless-clause, they are often trying to avoid
committing themselves completely: q is said first, and afterwards, as a
kind of afterthought, the only situation in which their utterance will not
be true is signalled (P). This claim is consistent with the historical
corpus-based analysis of Traugott (1997), who concludes that in general
unless-clauses express 'possible events that will block promises, threats,
expectations, and even probable recurrence. '
On the other hand, this utterance would not be very likely the other
way around; that is, it would be precarious if the graduate student
declared something like the following to his superior:
(42) Muy bien, dare sus clases a condici6n de que usted me ayude a
preparar la bibliografia.
'Okay, I'll teach your classes provided that you help me with the
bibliography. '
10. Summary
Notes
* I would like to thank Maria Rosa Vila and Joseph Hilferty for the many hours
that they spent with me discussing many of the points expanded upon here, as
well as the editors of this volume, Bernd Kortmann and Elizabeth Couper-
Kuhlen, for their exhaustive observations on this chapter. I would also like to
thank lbon Sarasola for providing me with the Basque example in endnote 4. I
would further like to mention the help received from Joseph Hilferty in
making an intelligible English version. It goes without saying, however, that
all errors are of my own responsibility.
1. The lack of quality grammatical descriptions regarding CCCs is reflected in
the fact that most Spanish textbooks and reference grammars for foreign
learners give a rather jumbled treatment of these constructions. Furthermore,
they tend to present them as if they were mutually commutable with one
another as well as with ij=-clauses in general. This has the unfortunate side
effect, as I have seen in my own classes, of prompting foreign students to
produce unacceptable utterances.
2. It may be possible that the conclusions reached here are also largely
applicable to languages such as Italian and German. Given my unfamiliarity
with these languages, however, I leave it to others to confirm or disprove this
Complex conditional connectives 167
References
Akatsuka, Noriko
1997 Negative conditionality, subjectification, and conditional reasoning.
In: Angeliki Athanasiadou and Rene Dirven (eds.), On Conditionals
Again, 323-354. Amsterdam: Benjamins.
Complex conditional connectives 169
Austin, John L.
1956 1ft and cans. Proceedings ofthe British Academy: 109-132.
Bartol Hernandez, Jose Antonio
1992 Algunas conjunciones y locuciones condicionales en la Edad Media.
In: Jose Antonio Bartol Hernandez, Juan Felipe Garcia and Javier de
Santiago (eds.), Estudios jilol6gicos en homenaje a Bustos Tovar,
vol. 1,85-110. Salamanca: Ediciones Universidad de Salamanca.
Brown, Penelope and Stephen C. Levinson
1987 Politeness. Cambridge: Cambridge University Press.
Comrie, Bernard
1986 Conditionals: A typology. In: Elizabeth Traugott, Alice Ter Meulen,
Judy Snitzer and Charles Ferguson (eds.), On conditionals, 77-99.
Cambridge: Cambridge University Press.
Contreras, Lidia
1963 Las oraciones condicionales. Boletin de Filologia de la Universidad
de Chile XII: 273-290.
Dancygier, Barbara
1985 If, unless, and their Polish equivalents. Papers in Contrastive
Linguistics XX: 65-72.
1990 Conditionals: Sequence of events and sequence of clauses. In: Jaeck
Fisiak (ed.), Further insights into contrastive analysis, 357-373.
Amsterdam: Benjamins.
1993 Interpreting conditionals: Time, knowledge, and causation. Journal
ofPragmatics 19: 403-434.
1998 Conditionals and prediction. Time, knowledge and causation in
conditionals structures. Cambridge: Cambridge University Press.
Dancygier, Barbara and Eve Sweetser
1997 Then in conditional constructions. Cognitive Linguistics 8(2): 1-28.
Dostie, Gaetane
1987 Etude semantique de quatre connecteurs conditionnels: a condition
que, pourvu que, en autant que et d'abord que. Le jran9ais moderne
LV/3-4: 174-203.
Ducrot, Oswald
1973 La preuve et Ie dire. Paris: Marne.
Esgueva, Manuel and Margarita Cantarero (eds.)
1981 El habla culta de la ciudad de Madrid. Madrid: Consejo Superior de
Investigaciones Cientificas.
Fauconnier, Gilles
1984 Espaces mentaux. Paris: Minuit.
Fillenbaum, Samuel
1986 The use of conditionals in inducements and deterrents. In: Elizabeth
Traugott, Alice Ter Meulen, Judy Snitzer and Charles Ferguson
(eds.), On conditionals, 179-195. Cambridge: Cambridge University
Press.
von Fintel, Kai
1991 Exceptive conditionals: the meaning of unless. North Eastern
Linguistics Society 22: 135-148.
170 Estrella Montolio
Ramsey, Violeta
1987 The functional distribution of preposed and postposed if- and when-
clauses in written discourse. In: Russell S. Tomlin (ed.), Coherence
and Grounding in Discourse, 383-408. Amsterdam: Benjamins.
Smith, Neil V.
1983 On interpreting conditionals. Australian Journal ofLinguistics 3: 1-
24.
Sperber, Dan and Deidre Wilson
19962 Relevance. Communication and cognition. Oxford: Blackwell.
Sweetser, Eve E.
1990 From etymology to pragmatics: Metaphorical and cultural aspects
ofsemantic structure. Cambridge: Cambridge University Press.
Traugott, Elizabeth Closs
1997 Unless and but conditionals: a historical perspective. In: Rene
Dirven and Angeliki Athanasiadou (eds.), On conditionals again,
145-167. Cambridge. Cambridge University Press.
Visconti, Jacqueline
1994 Entre hypotheses et conditions. Sur la semantique des connecteurs
conditionnels complexes. Cahiers de linguistique fran~aise 15: 77-
108.
Wing, C.S. and Ellin Kofsky
1981 Children's comprehension of pragmatic concepts expressed in
because, although, if, and unless. Journal ofChild Language 8: 347-
365.
Pre- and post-positioning of wenn-clauses
in spoken and written German *
Peter Auer
This paper is concerned with the distinction between pre- and postposi-
tioned (initial and final) wenn-clauses in German, and with the distinction
between written and spoken language. A simple cross-tabulation of the two
features [spoken/written] and [pre-/postpositioned] shows that initial wenn-
clauses are preferred in spoken German, but final wenn-clauses are pre-
ferred in written German. An attempt will be made to explain these find-
ings.
1. Introduction
The fmdings and discussions in this paper are corpus-based. They are
partly quantitative, partly qualitative. With respect to both dimensions,
the claim is that a full understanding of the syntax of (particularly) spo-
ken language eludes the possibilities of a purely introspective method-
ology. Of course, no (quantitative or qualitative) corpus-based investi-
gation can do without a strong reliance on the analyst's knowledge ('in-
tuition') about the language being researched; in fact, fmding valid gen-
eralizations always involves Gedankenexperimente, playing with struc-
tural changes in and recontextualizations of the 'examples' found to be
used by the informants. On the other hand, not even the empirical start-
ing point of the present investigation (i.e., the (differing) preferences of
spoken and written language for post- and prepositioning) is available to
a purely introspective approach, since it is of a quantitative kind. In or-
der to reach an explanation of these findings, this quantitative analysis
has to be complemented by an in-depth analysis of individual cases of
usage. Such an analysis will pay attention (a) to the in-time emergence
of syntactic patterns, including the details of their delivery such as hesi-
tations, reformulations, break-offs, etc., and (b) to the interactional as-
pects of this emergence, including hearer feedback (or lack of it) and
sequential placement. In this respect, spoken language research can
profit in important ways from conversation analysis.
174 Peter Auer
This, of course, is not the place for a full contrastive analysis of the two
constructions. In order to facilitate the reader's access to the examples
to be discussed below, and in order to link up the present investigation
with previous ones on tf-clauses in spoken (and written) Engiish, a short
sketch of some important similarities and differences may, however, be
in order here.
2.1. Syntax
For (2c), special conditions of use hold, and it may therefore be consid-
ered to be "marked" (cf Konig and van der Auwera 1988, Kopcke and
Panther 1985, Giinthner 1999, and below).
2.2. Semantics
Clearly, the temporal readings of wenn are not covered by English if,
but by when instead.
Some verbal and prosodic features of wenn-clauses may facilitate or
even enforce one or the other reading: (a) the temporal, non-iterative
reading is not available in sentences referring to past events; here, the
temporal conjunction als takes over (while English allows when); (b)
focussing adverbials such as stressed nur ('only') in the main clause
strongly suggest a conditional reading of the (following) wenn-clause;
(c) the particle schon (no English equivalent) in the wenn-clause sug-
gests a factual-conditional reading; (d) immer wenn ('always when')
instead of a simple wenn as a conjunction enforces a habitual-temporal/
contingent reading; (e) selbst wenn (' even if) and wenn ... iiberhaupt
('if ... at all') instead of a simple wenn enforce a conditional reading; (f)
subjunctive (Konjunktiv II) in the wenn-clause enforces a hypothetical-
conditional reading; (g) stressed wenn-conjunctions suggest a condi-
tional instead of a temporal reading.
Two special uses of wenn-clauses need to be mentioned. The first is
the expression of concessivity through the combination of wenn and
auch (wenn + auch or auch + wenn), roughly similar to Engl. even
if/even though: 5
(8) a. auch wenn sie KEInen job haben wollen, () miissen sie=n
bisschen da aufn PUNKT kommen. (fabricated)
'even if you don't want ajob, you need to get down to the point'
b. wenn sie auch keinen JOB haben wollen, () sie miissen n
bisschen da aufn PUNKT kommen. (fabricated)
'even though you don't want a job, you need to get down to the
point'
c. obWOIa sie KEInen job haben wollen, () miissen sie=n
bisschen da aufn PUNKT kommen. (fabricated)
178 Peter Auer
'although you don't want to have ajob, you need to get down to
the point'
(9) das EINfachste, da ham sie RECHT, das war fiir uns, wenn sie
mal==n MOnat (-) im teleFONmarketing ARbeiten wiirden.
'the simplest solution for us, and here you are right, would be if
you could work in our direct marketing sector for a month'
60,00
50,00
40,00
30,00
20,00
10,00
0,00
pre- paren- post-
single others
positioned thetical positioned
IJ percentage 56,00 4,00 32,60 5,40 2,00
ral or a clear conditional reading (based on the substitution tests and cri-
teria discussed in section 2.3).9
Of the n=203 disambiguated wenn-clauses, 24% have temporal, the
remainder conditional meaning. Exactly half of the wenn-clauses with
temporal meaning are prepositioned and postpositioned respectively.
(Of the 76% unambiguously conditional examples, about two thirds are
pre-positioned (55% of the total), one third is postpositioned (21 %).)
There is, then, a clear difference between conditional and temporal uses:
only for the former does the preference for pre-positioning hold. Since
the majority of German wenn-clauses are semantically ambiguous be-
tween a temporal and a conditional reading, this fmding also suggests
that, taken as a whole, they behave syntactically like (English) condi-
tional rather than temporal clauses.
Figure 2 shows the percentage of integrative, resumptive and non-
integrative constructions among the pre-positioned wenn-clauses in the
sample (n=280).
60,00
50,00
40,00
30,00
20,00
10,00
0,00
integrative resumptive non- others
integrative
25,72 47,83 18,48 9,42
(10) ich mein ich muss ihnen (-) ganz SCHNELL und GANZ: vehement
sagen wenns IRgendwie gEht (-) fahrns InN
'I mean I have to tell you without hesitating and very vehemently:
if you can make it at all, go there!'
(12) wenn auch die theoRIE; (-) eh (-) so IRgendwo mal gehOrt wurde
im KOPF? (-) eh das UMsetzen das ist ja das entSCHEldende,
'even though the theory (-) ehm (-) may have been heard
somewhere in one's head (-) the decisive thing is putting it into
practice'
(13) also wenn sie FRAgen ham zwischendufch, eh DIE konnen Sie
ruhig STELlen?
'well if you have any questions in between, ehm you can ask
THEM of course. '
(14) un wenn ich mein Eltem anrufh wUrde, ==DS wiirde AUCH nix
bringn.
'and if I called my parents, that wouldn't be any use either. '
(15) wenn WIRKlich==n ganzen tag das telefon klingelt, und acht
STUN' (-) man IS hinterher' < <acc>man WEISS was man>
getan hat. () geb ich ehrlich ZU
'if the phone really rings all day, and eight hOD' (-) afterwards
you are' you know what you have done. i have to admit that. '
I: mhm,
B: before [that we cannot (do) the total financing
I: [(h)
like with Kopenick 15
B: and (.) and he is just on the road then he says this isn't worth it
for me.
then I'd rather stay in North Rhine-Westphalia «a West German
state»'
This passage is embedded into a larger report the applicant gives of his
participation in a West German consultant agency in the New States,
which however closed down its East German office, making him redun-
dant. The interviewer does not seem to know the company and ques-
tions its importance on the market. The applicant counters by stating
that although small, the company had very professional consultants. At
the same time, he has to deal with the interviewer's innuendo that the
company withdrew from the East German market because it was not
working successfully. In this context, the case of "Mr. Keller" is men-
tioned, an experienced consultant who was disappointed by the
kafkaesque way in which state and bank authorities made it hard for
new enterprises to get subsidies, and returned to the Old States.
After he has been portrayed as a successful consultant who came to
East Germany mainly for idealistic reasons, "Mr. Keller's" dissatisfac-
tion with the situation is described in a complex tum construction which
starts out with a wenn-clause (wenn er nun aufder Straj3e sitzt... ). In the
given context, the interpretation is not hypothetical but refers to a (fac-
tual) state of affairs ('since he was always on the road... '), which is es-
tablished as the ground from which some conclusion can be drawn. Be-
fore this conclusion is reached, however, the speaker elaborates at con-
siderable length on the unfortunate situation in which "Mr. Keller" and
his clients found themselves; in four clauses each introduced by (und)
dann, the various fruitless journeys between the financing bank and the
state authorities in Schwerin are described. Towards the end of this
elaboration (securely produced by the speaker within the realm of his
own tum, since a syntactic projection - that of the when-clause - still
remains to be taken care of), the interviewer produces some recipiency
tokens which, although not claiming the tum (cf their reduced loudness,
indicating non-competitiveness), nevertheless acknowledge the
speaker's point: two laughter particles and one comment (wie mit
Kopenick) display understanding. Only after this feedback does the
Positioning ofwenn-clauses in German 187
hesitation pause to refute this solution: 'if you did that, there would be
trouble between the two of us' (i.e. between the mediator, B, and one of
the two people quarreling, i.e. himself). The utterance is linked to B's
proposal by the initial anaphoric da; it is semantically and syntactically
complete, and being marked by a final fall, it certainly is a candidate for
a complete tum. However, B does not pick up this refutation, but con-
tinues with the production of the unfinished syntagm in another piece of
simultaneous talk. Sequential structure and temporal development are
now out of phase: a response has been produced to an utterance which is
still in need of being completed, and is only completed after the re-
sponse. In this context, II's following wenn-clause, syntactically ex-
panding an already complete turn/syntagm, can be seen as a skillful way
of re-aligning sequentiality and timing: it re-instantiates II's refutation
of B' s proposal without repeating it, by retrospectively transforming a
simple construction into a hypotactical one with a post-positioned ad-
verbial clause. Semantically, this expansion adds nothing new: it just
restates what B herself has said before.
The possibility of such an expansion is not only available to the
speaker but also to the recipient, of course, who may become a co-
speaker and co-producer of the emerging syntactic pattern by adding a
wenn-clause himself/herself:
(23)
L: .h dann: eh () wir' der Hundwird auchjetz zunehmend ruhiger;
S: mHM (-) des GUT so; (-)
L: JAja des==also wird langsam (a)==richtiger HUND;
S: aHA
L: hm, (-)
S: wenn==er (nicht mehr) abhaut, (-)
hat (name) des ANgebot jetz fiir den zaun? ==
'L: then ehm beco the dog is becoming more and more calm now;
S: this is how it should be;
L: yeah slowly he's tuming into a real dog.
S: I see
L: hm,
S: if he doesn't escape (any more),
did NN get the offer for the fence in the meantime?'
Positioning ofwenn-clauses in German 193
In the last sections, it has been shown that wenn-clauses are preferen-
tially pre-positioned with respect to their main clauses in spoken Ger-
man, and that this serialisation has a number of cognitive and interac-
tional advantages. It has also been shown that the more marked struc-
ture, i.e. post-positioned wenn-clauses, which does occur in about a
third of all instances, has its own specific contexts of usage. These are
partly due to (a) syntactic constraints on pre-positioning in superordi-
nate clauses without a pre-field or in which two complementizers occur
in adjacent position; partly to (b) tum-taking (afterthought position);
partly to (c) semantic-syntactic reasons (wenn-clauses in complement
function are postpositioned); and partly to (d) pragmatic reasons (post-
positioned dependent clauses are thematic or subthematic, unless focus-
sing particles indicate the contrary).
In written German, wenn-clauses are generally less frequent than in
spoken discourse, a finding which contradicts the frequent claim that
spoken language avoids syntactically complex constructions: the fre-
quencies of wenn-clauses per 100 words in the corpus of spoken lan-
guage used above is 0.54, but in a corpus of written language, taken
from the newspapers DIE ZEIT (politics section) and Frankfurter All-
gemeine Zeitung (culture section), it is 0.33. 23 To put it differently,
Positioning ofwenn-clauses in German 195
every 186th word is wenn in our spoken corpus on an average, but only
every 300th word in our written texts. Once more, this fmding is in line
with comparative work on written and spoken English (Ford and
Thompson 1986, 354: 0.72 vs. 0.46; similarly: Beaman 1984 and Biber
1986), but also with previous work on German (Leska 1965: 450).
There are of course also qualitative differences between the wenn-
clauses used in the two corpora; in particular, certain rather idiomatic
patterns (constructions) seem to be more or less exclusively used either
in spoken or written language. For instance, the reduced wenn-(dann)-
constructions of spoken German (cf section 2.2. above) do not occur in
the newspaper corpus, while, on the other hand, the topicalizing causa-
tive construction wenn p dann (deswegen), weil q ('if p, then that is be-
cause of q'), as in (26), seems to be used exclusively in writing.
60,00
50,00
40,00
30,00
20,00
10,00
0,00
6. Conclusion
M .hh ich will UMgehend den (-) nachs- Ich will umgehend den nachsten An-
ten (-) ANrufer wieder einen Horer rufer, wieder einen Horer, begrii.6en:
begrii.6en guten Abend? Guten Abend.
A guten Abend, Guten Abend.
B guten Abend? Guten Abend.
(0.5)
A .hhh ja; ALso hh das proBLEM ah Ja, also das Problem aller steht auch
ALler steht ah auch ah so fib vor MIR so vor mir jetzt irgendwo. Wie und
jetzt irgendwo-wie wie und wo AN- wo anfangen am besten? Es ist eine
fangen am besten? etwas auBergewohnliche Problema-
B rnhm, tik, die vielleicht nicht so ganz haufig
A = ahrn: (--) es is eine etwas AUSser- in ihrer Sendung erscheint, obgleich
gewohnliche Proble ProbleMAtik, die das eigentlich eine total menschliche
vielleicht nicht nicht so ganz ah: : Angelegenheit ist. Ich lebe seit, urn
HAUfig in ihrer sendung erscheint, gleich mal auf Fakten zu kornrnen,
.hh obgleich des eigentlich ah::: ne ich lebe seit circa fiinfzehn Jahren
total MENSCHliche angelegenheit is- mit einern Mann zusarnrnen.
ich lebe: seit =um gleich mal auf
FAKten mal zu KOMM: ich lebe seit
hh fib circa FONFzehn JAHRN mit
einern MANM zusArnrnen? [.hh ]
B [ja] Ja?
A und fib auch ah recht GU:T eigentlich Und auch recht gut, eigentlich. Na-
=natiirlich mit den iiblichen ALLtags- tiirlich mit den iiblichen Alltags-
schwierigkeiten die iiberall existieren schwierigkeiten, die iiberall existie-
200 Peter Auer
Notes
* The corpus research on which this paper is based was supported in many and
substantial ways by Benjamin Stoltenburg. Thanks to Jessica Wallace for cor-
recting the worst blunders in my English, and to Susanne Giinthner as well as
to the editors of this volume for their comments on a previous version.
1. Transcription of the spoken extracts follows GAT-conventions (cf. Selting et
al. 1998); capital letters indicate stress positions. English translations are sim-
plified, particularly with respect to prosody and hesitation phenomena. In case
of conflict, less idiomatic versions have been chosen in order to give a better
impression of German syntactic structure.
2. The wenn-clause itself should be seen as adjoined to the resumptive particle,
i.e., as a co-constituent of the front-field; cf. Eisenberg (31994: 364f).
3. For a detailed discussion, see Metschkowa-Atanassowa 1983 and Zifonun et
al. 1997: 2280-2293.
4. A certain kind of ambiguity between a temporal and a conditional reading can
also be observed in the German question word wann (usually asking for tem-
poral information, 'at which time') which, when followed by a verb in the
subjunctive mood, often takes on a conditional meaning (wann wiirdest du
kommen = 'under which conditions would you come' or 'at which time would
you come'). Thanks to Bernd Kortmann for drawing my attention to this par-
allel.
5. For an analysis of these "concessive conditionals", cf. Konig 1985.
6. Among these wenn-clauses in the role of obligatory constituents, we may also
count comparisons using wie wenn and als wenn, as in: du kOmmsch dir
vielleicht vor wie wenn dir deine wErte verLORN gangn sin. ('maybe it seems
to you as ifyour values had been lost '). This usage of wenn will not be taken
into account in the following discussion, nor has it been included in the quan-
titative analysis.
A note in passing: some grammarians believe that wenn-clauses in comple-
ment function are obligatorily marked by a resumptive es (e.g. Eisenberg
31994: 365); this is not supported by my data, however.
7. Some 400/0 of the corpus are job interviews, mainly collected among north and
east German speakers, some 30% are therapeutic conversations, both in face-
to-face and in radio phone-in contexts, and the remaining 30% represent pri-
Positioning ofwenn-clauses in German 201
vate everyday conversations, partly on the telephone. In the latter two types of
data, southern German speakers prevail.
8. Embeddings of wenn-clauses into complex hypotactic constructions were not
counted as parenthetical.
9. The count excludes, in addition to all polyvalent cases, all factual conditionals
(Le. those expressing a positive epistemic stance), which are always non-
temporal, and all concessives, as well as wenn-clauses used as complements,
but includes counterfactual conditionals and reduced wenn-dann routines.
10. *Warum [wenn Du Kirschen magst] pfliickst Du Dir keine vom Baum? 'Why
[if you like cherries] don't you pick any from the tree?' therefore has to be
understood as parenthetical. Resumption by dann is also excluded here, Le.,
non-integration is the only option.
11. The term is used in a broader sense here than in Sweetser 1990. Details on this
construction may be found in Gfinthner 1999. Note that Sweetser's "epistemic
conditionals", although not "content conditionals", do not allow pre-front field
placement in German (* Wenn er sich jeden Tag voillaufen lasst, sie hat ihn
verlassen. 'If he gets drunk everyday, she has left him. ').
12. Cf. Konig and van der Auwera 1988: 128 ("assertive emphasis on a conse-
quent of a concessive allows non-integration"), Kopcke and Panther 1989:
700 ("high degree of ego involvement") and Gfinthner 1999 for details.
13. Cf., among others, Ford and Thompson 1986: 370; Ford 1993; Dancygier and
Sweetser 1996.
14. But see Ford 1993: 56.
15. Speaker I is most likely alluding to Zuckmayer's play (and a famous German
movie) Der Hauptmann von Kopenick, in which the Prussian state and army
authorities are caricatured.
16. This of course, leaves the question open why temporal adverbial clauses
should behave differently from conditional ones-a question which requires
an investigation of its own.
17. For a similar remark on English, cf. Ford and Thompson 1986: 369.
In some cases, however-though not in (26) with its oblique relative pro-
noun-the wenn-clause can follow the relative pronoun (das Essen ist wie ein
Teddybtir, der, wenn es hart ist, immer bei Ihnen ist, und der, wenn man ein-
sam ist, zum Festhalten da ist). But here we are dealing with parenthetical
placement in the middle field of the sentence; this is exceedingly rare in spo-
ken German.
19. Of course, wenn is not always stressed in fronted wenn-clauses. Cf. the fol-
lowing example:
ich (.) hab (-) FUNF jahre lang an der schule franzOsisch geHABTh, mir
jEhlts eigentlich an (.) PRAxis, .h aber: (-) ich bin aberZEUGT davon, =wenn
ich: eh eh OFters mal die geLEgenheit hatte zum beispiel in FRANKreich, eh
mich aufzuhaltn, .hh dass des: (-) eh SICherlich Ausbauftihig is.
'I had French at school for five years, actually I'm lacking practice, but I'm
convinced if I on occasion had the chance to spend some time for instance in
France, that I could work on it. '
20. The tendency to place the wenn-clause early in dependent constructions is also
evidenced by the fact that parenthetical placement immediately after dass is
202 Peter Auer
frequent (cf. Note 18). Often, a second, resumptive dass is added at the begin-
ning of the consequent:
kAnnst du ihm vielleicht (-) A USrichten dass ich ANgerufn habe?
und dass wenn er mit dem trelitz geSPROChen hat uber meine priifungk,
dass er sich dann irgenwie=mal=GANZ kurz bei mir mElden soli?
'could you perhaps tell him that I called?
and that if he has talked to Trelitz about my exams,
that he should give me a quick ring some time?'
21. The same was found in English conversations by Ford and Thompson 1986:
368.
22. In other-speaker produced post-positioned wenn-clauses this does not always
hold, since second speakers may choose this way of intimately linking their
speech to a preceding syntactic pattern but nonetheless produce unexpected
and even contradictory information under this 'disguise'.
23. More exactly, the corpus included the F.A.Z. Feuilleton-Glossen from Jan 8,
1993 (Ausgabe Nr. 6) to Dec 31, 1993 (Nr. 304) and DIE ZEIT politics sec-
tion of Dec 30, 1994 (No.1) to Feb 17, 1995 (Nr. 8). Transcriptions of con-
versational speech were regularized in order to make a comparative computer-
based word-count possible; see the appendix for an example.
24. The written corpus used by Ford and Thompson consisted of philosophical
essays, a professional text for automobile mechanics, and a personal narrative
account (1986: 355). The preference for pre-positioning held for all these
sources.
25. A separate count for the two newspapers shows that the quantitative results
are identical.
26. The resumptive particle is dann in 46 cases and so in 16 cases. Dann and so
are not freely interchangeable; in particular, so can be used with concessives
(introduced by auch/selbst wenn), while dann cannot. Cf.: Auch wenn das Ab-
geordnetenhaus erst im Herbst, voraussichtlich am 22. Oktober, gewtihlt
wird, so hat mit der Urabstimmung schon der Wahlkampf begonnen. ('Even
though parliament will not be elected until autumn, probably on October 22,
SO the electorial campaign has already begun with the strike ballot. ')
27. Of the 15 tokens in the spoken material, the three positions were about equally
distributed.
28. For a similar argument, cf. Thompson 1985.
References
Auer, Peter
1991 Das Ende deutscher Satze-Rechtsexpansionen im deutschen Ein-
fachsatz. Zeitschriftfur germanistische Linguistik 1991: 139-157.
1996 The pre-front field in spoken German and its relevance as a gram-
maticalization position. Pragmatics (Special issue, eds. Johannes
Wagner and Cecilia Ford), Vol. 6, No.3: 295-322.
Positioning ofwenn-clauses in German 203
Biber, Douglas
1986 Spoken and written textual dimensions in English: Resolving the
contradictory findings. Language 62: 384-416.
Couper-Kuhlen, Elizabeth
1999 Varieties of conditionals and their emergence in discourse. In: Aditi
Lahiri, Alexander Patschosvky and Christoph Schwarze (eds.), Is-
sues in interdisciplinary research on the lexicon = Working Paper
No. 99, University of Konstanz, Fachgruppe Sprachwissenschaft, 89-
130.
Dancygier, Barbara and Eve Sweetser
1996 Conditionals, distancing, and alternative spaces. In: Ann E. Goldberg
(ed.), Conceptual structure - Discourse and language, 83-98. Stan-
ford: CSLI Publications
Eisenberg, Peter
1994 GrundrifJ der deutschen Grammatik. Stuttgart: Metzler.
Fauconnier, Gilles
1985 Mental spaces. Cambridge: Cambridge University Press.
Fillmore, Charles J.
1990 Episternic stance and grammatical form in English conditional sen-
tences. In: Chicago Linguistic Society 26, University of Chicago,
137-162.
Ford, Cecilia E.
1993 Grammar in interaction. Adverbial clauses in American English
conversations. Cambridge: Cambridge University Press.
1997 Speaking conditionally: Some contexts for if-clauses in conversation.
In: Angeliki Athanasiadou (ed.), On conditionals again, 387-413.
Amsterdam: Benjamins,
Ford, Cecilia E. and Sandra A.Thompson
1986 Conditions in discourse: a text-based study from English. In: Eliza-
beth Traugott et al. (eds.), On conditionals, 353-372. Cambridge:
Cambridge University Press.
Greenberg, Joseph H.
1963 Some universals of grammar with particular reference to the order of
meaningful elements. In: Joseph H. Greenberg (ed.), Universals of
language, 73 -113. Cambridge, Mass. : MIT-Press.
Gtinthner, Susanne
1999 Wenn-Satze im Vor-Vorfeld: ihre Formen und Funktionen in der
gesprochenen Sprache. Deutsche Sprache 27(3): 209-235.
Haiman, John
1978 Conditionals are topics. Language 54: 564-89.
Konig, Ekkehard
1985 Where do concessives come from? On the development of conces-
sive connectives. In: Jacek Fisiak (ed.), Historical semantics, 263-
282. Berlin: de Gruyter.
204 Peter Auer
This paper calls into question the popular view of counterfactual thinking
under the influence of formal logic, which considers the utterance of a
counterfactual sentence as an instance of complex and intricate reasoning
skills. Using naturally occurring data from English, Japanese, and Korean,
we will demonstrate that there exists a pattern of prototypical
counterfactual reasoning which appears to be natural and spontaneous to
every human being. Speakers/writers express a particular stance of
desirability versus undesirability toward a particular event, based on their
subjective evaluation of reality. Counterfactual conditionals are invoked as
a necessary step in this line of thinking.
1. Introduction l
This chapter represents a continuation of our inquiry into the most typical
usage of counterfactual conditionals in everyday life across language and
culture (Akatsuka 1997, 1999). We use the word "typical" here in the
sense of "natural and spontaneous," and thus underscore our claim that
human beings appeal to counterfactual reasoning in dealing with the
many aspects of everyday life that we find ourselves facing.
In the tradition of mathematics, philosophy and formal linguistics, the
conditional sentence has been regarded as the epitome of Man's rational
capacity, the height of Man's ability to reason logically. In the same vein,
it has long been argued that underlying the counterfactual conditional is a
similar, and perhaps even more logically complex type of reasoning,
engendering such analytical sentences as in (1), from Fauconnier (1985:
118).
(1)
A: If Napoleon had been the son ofAlexander, he would have won the
battle of Waterloo.
B: But he would have died long before that.
A: Well, suppose he lived a very long life, without ever ageing, or that
Alexander was resurrected in Corsica in the eighteenth century.
206 Noriko McCawley Akatsuka and Susan Strauss
2.2. "Space-builders"
(3) a. If Lucky had won, I would be rich. I would have moved to Tahiti.
b. I wish Lucky had won. I would be rich.
c. Fortunately, the fire did not cross the highway. My house would
have been destroyed.
d. Luckily, the fire was prevented from crossing the highway. My
house would have been destroyed.
(5) a. ??Thefire did not cross the highway. My house would have been
destroyed.
b. ?? The fire was prevented from crossing the highway. My house
would have been destroyed.
(7) Fortunately, the fire didn't cross the highway. My house would
have been destroyed.
'I was lucky that the fire did not cross the highway. If it had, my
house would have been destroyed. '
The essential difference between the two examples is that the conditional
antecedent appears in the Japanese example, (8). That is, unlike English,
Japanese syntax does not normally tolerate monoclausal conditionals. And
once again, we see the overt representation of the speaker's attitude,
desirable, towards the past event/state of affairs described by the
proposition, p. In this case, p is synonymous with 'the fire did not cross
the highway' (=fact).
To illustrate in English, we repeat examples (Sa) and (5b) below.
(5) a. ??The fire did not cross the highway. My house would have
been destroyed
b. ??Thefire was preventedfrom crossing the highway. My house
would have been destroyed.
(5) a' The fire did not cross the highway. If it had, my house would
have been destroyed.
(5) b' The fire was prevented from crossing the highway. If it hadn't
been, my house would have been destroyed.
Counterfactual reasoning and desirability 211
Now, (Sa') and (Sb') both contain the explicit conditional antecedent
'if not P,' and it is here that we can begin to understand what the speaker
is actually doing by invoking counterfactuality: she is presenting the
reason why she considers it desirable that p, i.e., 'the fire did not cross
the highway' (=fact), happened. The counterfactual situation, i.e., 'not p'
or 'the fire's crossing the highway,' would have led to the loss of her
house. We informally represent this line of reasoning, in (9):
Contrary to Fauconnier's claim, then, it is not the lexical items not and
prevent that build the so-called counterfactual space. Rather, we maintain
that it is the speaker who invokes counterfactual thinking as an integral
step in the line of reasoning, as delineated in (9). Thus, the use of
negation figures strongly in the expression of p vs. 'if not P,' as well as q
vs. 'not q,' or in other words, in the speaker's expression of stance of
desirable vs. undesirable, and not in the construction itself of the so-
called 'counterfactual space. '
Let us now tum to Fauconnier's analysis of (3a) and (3b), repeated
below as (lOa) and (lOb) for convenience.
(lOa) IfLucky had won, I would be rich. I would have moved to Tahiti.
(lOa)' Unfortunately, Lucky did not win. If he had won, I would be rich.
IfI were rich, I would have moved to Tahiti.
(lOb)' Unfortunately, Lucky did not win. I wish he had won. If he had
won, I would be rich.
212 Noriko McCawley Akatsuka and Susan Strauss
It will be obvious that (9) and (11) represent the identical pattern of
reasoning, with the sole difference being the speaker's desirability
assessment taking opposite values. The line of reasoning delineated in (9)
and (11) will be referred to henceforth in this study as "prototypical
counterfactual reasoning."
We have demonstrated that Fauconnier's framework is an i1mpersonal
one, not unlike those proposed by scholars from the tradition of formal
logic and mathematics; the view is necessarily limited to an incomplete
observation of the distribution of counterfactuals. Crucially, the
framework fails to account for all four examples (i.e., (3a)-(3d» as
squarely representative of prototypical counterfactual reasoning.
Unwittingly, too, even example (1), based on an invented dialogue
positing absurd conditions ('if Napoleon had been Alexander's son'),
expresses a strong stance of desirability ('he would have won the Battle
of Waterloo'). And what Frenchman would find this anything but
desirable?
We now tum our discussion away from constructed examples and
examine authentic discourse excerpted from the earthquake data in
English, Japanese, and Korean, where "prototypical" counterfactuals
emerge spontaneously.
earthquake hit. Because of these and other factors, the casualties were
4
rather light, given the magnitude of the quake .
Approximately three weeks following the incident, researchers from
UCLA collected 'earthquake stories' from dyads of native speakers of
English, Japanese, and Korean. Subjects were predominantly UCLA
graduate students, though some undergraduates also participated. In all
cases, we attempted to ensure that the dyad participants did not know
each other previously, though here too some of the Japanese data reveals
that a handful of subjects did know their interlocutor to some extent. For
the most part, however, subject pairs were intended to be strangers.
The only instructions given to the subjects were to tell each other
about the experiences they had had during the earthquake, and that the
data recording session would last for approximately 20-25 minutes. Other
than these criteria, no explicit instructions were given. The interactions
were both audio and video taped and were transcribed using the
conventions of Conversation Analysis (Atkinson and Heritage 1984).
What is remarkable is that counterfactual conditionals emerged in all
three languages, and in a number of cases we find significant overlap in
terms of the types of issues captured with counterfactuals (e. g., 'if it had
happened during rush hour,' 'if it had happened in a place other than
LA,' 'if it had happened before the recent earthquake retrofitting of some
of the campus buildings,' etc.). None of the conversational dyads was
aware of what the other dyads had said in other recording sessions.
Further, because the study was designed solely to elicit spontaneous talk
in the respective languages, no mention of counterfactuals was ever made
at the inception. It was not until we transcribed and analyzed the data that
we discovered this parallelism.
In the discussion that follows, we limit ourselves to one representative
sample from each language. In all three examples, the subjects
systematically express their reactions to the relatively light damage and
small number of injuries and deaths, in the light of the strength of the
earthquake itself
Interestingly, the abstract line of reasoning of each case is quite similar
to what we have noticed in Fauconnier's examples (3c) and (3d), even
though Fauconnier's examples are monologic and constructed, while our
data derive from spontaneous, natural dialogue. Instead of the initial
occurrence of such attitudinal adverbs as fortunately and luckily as we
witnessed in the Fauconnier examples, we observe here the speakers
exchanging and building on the expression "we were lucky", in all three
languages. We will call this activity 'the opening.' Just following this
sequence emerges the counterfactual utterance, 'if not p' had happened,
214 Noriko McCawley Akatsuka and Susan Strauss
'not q,' would have happened in all three languages. Henceforth, we will
adopt the formula 'if not p, then not q' to represent counterfactual
conditionality.
Let us now examine each excerpt, starting with English.
The speakers, T and H, were just talking about the fact that friends living
in the dorms said that the building seemed to sway at the time of the
earthquake; T then adds that some of the UCLA dormitories had recently
undergone seismic retrofitting: "actually, it's good that this happened now
because..." In other words, "we were lucky," hence the opening
sequence.
Observe, too, that after this opening T immediately appeals to
counterfactuals. What is of particular importance here is T's initial clause
"I jus' kept thinking," which precedes the counterfactual construction,
underscoring just how naturally and just how spontaneously we appeal to
counterfactual reasoning.
(12) English:
«participants: T=Female, H=Male))
(a) Opening
In the stretch of talk in section (b) 'if not p, not q' we witness the
gradual emergence of a complex set of thoughts, all of which express
some counter-to-fact notion with respect to some aspect of the reality of
this earthquake. That is, first T proposes a time and a day different from
the actual time and actual day of the incident p and then indicates that the
consequences q would have been much worse (or undesirable), a line of
reasoning we encountered earlier in example (9), repeated below as (13):
(14) Korean:
«participants: H=Male, T=Male))
(a) Opening
H2: mannyak-ey..
if (lit. in a one in ten thousand chance)
'if
The Korean example represents a very neat case where one speaker
produces the conditional antecedent, 'if not p' clause and the second
speaker the consequent, 'not q.' And the interactional dynamic which
engenders such a collaborative construction is an interesting one.
Even more revealing, however, is the Japanese example, where we
find the most elaborate instance of collaboratively produced
counterfactual reasoning of the three excerpts under investigation.
218 Noriko McCawley Akatsuka and Susan Strauss
(15) Japanese:
«participants: H=Male, A=Female))
(a) Opening
happened in the library at a time when students were in it'). And finally,
the relevance of an altered time emerges in the talk, referring to the
freeway overpass which collapsed as a result of the quake, beginning at
line A3, and resumed again from H7 through the end of the excerpt.
(b) 'ifnotp,notq'
A7: zenbu
all
'all of them'
(c) Closing
HIO: nnn
'yeah'
All: nn=
'yeah.'
HI2: =nn nn
'yeah.'
3.3. Summary
To sum up, in the excerpts from all three languages, we find the
consistent parallel whereby speakers invoke counterfactual thinking to
express the pattern of reasoning originally introduced as (9) and (13), and
repeated below as (16):
3.4. Discussion
3.4.2. The speaker's belief: "if 'not p' actually happens, then terrible
consequences will result"
rather than grammar (e.g., Quine 1950, Grice 1975, Stalnaker 1975). The
position of these logicians has been more or less inherited by many
linguists (e.g., Geis and Zwicky 1972, Haiman 1978, Karttunen and
peters 1979, McCawley 1981). It is important to note that Fauconnier
(1985: 113-115) explicitly rejects the idea that there is an inherent
connection between p and q of both indicative conditionals and
counterfactual conditionals. Fauconnier even takes the position that the
counterfactuality of q in counterfactual conditionals is implicature and
therefore cancelable. Basically, then, Fauconnier (1985) treats natural
language conditionals very much like mathematical conditionals. Note
that in the case of mathematical conditionals, there is no connection
between p and q of 'p ::> q'; they are independent propositions. Further,
there is no sequential relationship between the two, either. However,
McCawley (1981) observed that in all English conditionals, p is
temporally and/or causally or epistemologically prior to q. We now come
to the conclusion that it is far from being an accident that Fauconnier's
mental space theory has failed to account for the prototypical
counterfactual reasoning phenomenon in our discussion in Section 2.
(20) Mother: You have to be a good girl. If you are not a good girl,
Santa Claus won't bring you anything.
Throughout the data, the caregivers often gave warnings, criticisms, and
advice to their children, constituting a particular type of persuasive
speech act. Typically, this pattern of persuasion takes the following form:
An initial utterance expressing "do X" or "don't do X" is followed by the
provision of a reason, couched in a conditional statement. In other words,
the message takes the general shape of "if children don't listen to what
adults say, undesirable consequences will result." What is being
contrasted here are desirable consequences versus undesirable
consequences. We can informally represent this line of reasoning involved
here as (21):
(21) It is desirable that p will happen. If 'not p' happens, it will lead to
undesirable consequences.
(23) None of this would have happened if Wayne Lo, at the age of 18,
had not been able to walk into a gun store, flash his driver's license
and $129 and walk out with a deadly weapon. Or if he had not been
able to have 200 bullets sent to him at Simon's Rock College by a
mail-order arms company. To my friend, Greg, there is a
straightforward conclusion to be drawn from the mystery of Galen's
death. "We've just got too many guns in this country. We've got to get
rid of them". Anyone who reads Gone Boy will find it hard to disagree.
This section will serve as a preview of our next project, which involves a
re-examination of Labov's (1972) seminal study of narratives in the light
of our prototypical counterfactual reasoning hypothesis. Specifically, we
are interested in Labov's treatment of counterfactuals in relation to his
key notions of "evaluation" and "comparators". Simply stated, Labov's
"evaluation" is a syntactic unit with which the narrator shows why a
particular event is a reportable experience; "comparators" are
grammatical devices that the narrator utilizes to evaluate the narrative.
According to the original framework, neither notion is related to the
speaker's evaluative stance of desirability. Consider the following excerpt
(1972: 387) (emphasis added), where Labov cites one of the most
dramatic danger-of-death stories to illustrate the notion of what
constitutes a comparator:
(24) One of the most dramatic danger-of-death stories was told by a retired
postman on the Lower East Side: his brother had stabbed him in the
head with a knife. He concludes:
And the doctor just says, "Just about this much more", he says,
"and you'd've been dead".
Comparators then include negatives, futures, modals (emphasis added),
quasimodals, questions, imperatives, or-clauses, superlatives, and
comparatives, more or less in increasing order of syntactic complexity.
5. Concluding remarks
Abbreviations
ATTR ATTRIBUTIVE
COMPLTV COMPLETIVE
COND CONDITIONAL
CONN CONNECTIVE
COP COPULA
EXCL EXCLAMATION
FLLR FILLER
GEN GENITIVE
GER GERUNDIVE
:MDL MODAL
NML NOMINALIZER
ONM ONOMATOPOETIC
PLN PLAIN FORM
PRT PARTICLE
PST PAST
SM SUBJECT MARKER
TAG TAG MARKER
TM TOPIC MARKER
Notes
2. See Akatsuka (1985) for a full discussion of the epistemic scale where this
relationship is discussed in cognitive terms.
3. The Japanese data were collected and transcribed by Shoichi Iwasaki and his
research assistants, Eri Yoshida and Yumiko Kawanishi. Data collection and
transcription for the Japanese data appearing here were partially supported
by the University of California's Pacific Rim Study Grant (1992-94). The
English and Korean used in this study data were collected and transcribed
independently from the Iwasaki data.
4. The magnitude of the Northridge earthquake was 6.6. In all, 51 people lost
their lives and over 5,500 people were injured. For the sake of comparison,
the earthquake in Kobe, Japan, which occurred exactly one year later at
approximately the same time of the morning, measured 7.2 on the Richter
scale~ the number of fatalities in the Kobe earthquake exceeded 5,000.
5. See Akatsuka (1986) for an earlier discussion of this point based on
introspective data
References
Akatsuka, Noriko
1985 Conditionals and epistemic scale. Language 61: 625-639.
1986 Conditionals are discourse-bound. In: Elizabeth Closs Traugott,
Alice ter Meulen, Judy Snitzer Reilly, Charles A. Ferguson (eds.),
On conditionals, 333-352. Cambridge: Cambridge University Press.
1991 Dracula conditional and discourse. In: C. Georgopoulos and R.
Ishihara (eds.), Interdisciplinary approaches to language: Essays in
honor of S.-Y. Kuroda, 25-37. Dordrecht/BostonILondon: Kluwer
Academic Publishers.
1997 On the co-construction of counterfactual reasoning. Journal of
Pragmatics 28: 781-794.
1999 Towards a theory of desirability in conditional reasoning. In: Akio
Kamio and Ken-ichi Takami (eds.), Function and Structure, 195-
213. AmsterdamlPhiladelphia: Benjamins.
Atkinson, J. Maxwell and John Heritage (eds.)
1984 Structures of social action: Studies in conversation analysis.
Cambridge: Cambridge University Press.
Clancy, Patricia M., Noriko Akatsuka and Susan Strauss
1997 Deontic modality and conditionality in adult-child discourse: a
cross-linguistic study. In: A. Kamio (ed.), Directions in functional
linguistics, 19-57. Amsterdam: Benjarnins.
Fauconnier, Giles
1985 Mental spaces. Cambridge, MA: MIT Press.
Counterfactual reasoning and desirability 233
TEXTS:
1994 Los Angeles earthquake data:
English data: Susan Strauss, 1994
Japanese data: Shoichi Iwasaki, 1994
Korean data: Susan Strauss and Yumiko Kawanishi
Adversative connectors on distinct levels of
discourse: A re-examination of Eve Sweetser's
three-level approach *
Ewald Lang
1. Introduction
The levels are distinguished depending on what the clauses are taken
to represent and hence on what sort of entities the causal relation is taken
to obtain. Thus, in (la) it is two content units between which a factual
causal relation is asserted to obtain. More exactly: the content units are
two propositions whose couching in declarative sentences enables them
to render assertable statements. The linking pattern for a causal relation
on the content level thus amounts to STATEI\ffiNT because STATEI\ffiNT.
In (1 b), the first clause is not asserted as a statement but only as an
assumption which is inferred from what is stated as fact in the second
clause. The causal connector marks the second clause as sufficient
evidence justifying the assumption rendered by the first clause. The
linking pattern on the epistemic level is ASSUMPTION, because
STATEMENT (EVIDENCE). The hypothetical status of what is asserted by
the first conjunct can be made explicit by embedding it under modals -
cf (1 b-i, ii) or by rephrasing it as an inference - cf (1 b-iii):
(1) b-i. John must have loved her, because he came back.
b-ii. John undoubtedly loved her, because he came back.
b-iii. From the fact that he came back,
I conclude that John loved her.
Adversative connectors on distinct levels ofdiscourse 237
In (Ic), the first clause serves the purpose of a speech act of asking
whose performance is being justified by the statement rendered by the
second clause. The linking pattern thus is SPEECH ACT, because
STATE:MENT (JUSTIFICATION). The because-clause does not form part
of the question but is linked instead to an understood assertion which, if
explicitly inserted, may read like this:
While 1 fundamentally agree with claim (I) One meaning, several uses (I
consider this to be the only promising way of dealing with connectors!),
and while 1 am basically sympathetic to the notion of "pragmatic
ambiguity", 1 see some serious difficulties in accepting the claims in (II)
and (ill).
238 Ewald Lang
(C-2) The lexical items, i.e. the connectors that are used to connect
clauses on the three levels of interpretation, display more level-
specific distributional restrictions than the sample of cases
discussed by Sweetser would suggest.
(CI-I) Are the clauses of the same or of different clause type? (in
morphosyntactic terms of being declarative, interrogative, or
imperative sentences)
This is the point where further criterial features enter the picture. One of
them, of course, is comma intonation. Due to its interaction with other
features the indicative role of comma intonation is more far-reaching than
Sweetser assumes. Below the level of sharing declarative mood there are
important indicators of clause-internal differences which interact with
comma intonation. Thus, in isolation, the first clause of (1 b) may render a
STATEMENT as does the first clause in (la). Both behave differently,
though, if we add an adjunct like presumably or a parenthetical like I
guess, which mark the hypotheticality or assumption status of the
assertion being made:
3. Adversative connectors
From the set of about 30 lexical items that may be subsumed under the
label "adversative connector" in German we will select only those needed
to illustrate (C-I) to (C-3).
Let us first turn to coordination involving the connector aber, which
(like but in English) is commonly considered as the standard adversative
connector. Compared with connectors such as trotzdem ('nevertheless'),
hingegen ('whereas'), or dennoch ('nonetheless', 'all the same'), aber is
much less restricted regarding distribution and hence to a much higher
degree prone to "pragmatic ambiguity". Nevertheless, the supposed
ambiguity of aber-conjunctions is drastically reduced by cues which
emerge from structural features of the coordinate clauses in the domain of
sentence structure. In addition to CI-I and CI-2, the relevant features
include:
Let us now look at some data that are representative of the range of cases
to be taken into account.
(3) a. Dein Vortrag war sehr lang, aber das soli keine Kritik sein.
'Your talk was rather lengthy, but this is not a criticism. '
Adversative connectors on distinct levels ofdiscourse 243
Sweetser comments: "But presents two conjuncts which clash with each
other in some way ... At the epistemic level, the available premises may
clash with an apparently necessary conclusion" (1990: 100). The first part
of the quote is unproblematic; but what exactly is "epistemic" about the
entities connected by aber/but? (4a, b) instantiate the well-known type of
"denial of expectation" interpretation of but-conjunctions (Lakoff
1971:133). In both cases, the clauses are mutually independent, hence
semantically compatible STATEMENTS. Given this, (4a, b) qualify as
coordinations of statements at the content level with the additional
requirement that the interpretation involves inferring an ASSUMPTION
244 Ewald Lang
which is in contrast with the two statements being presented side by side.
This analysis reduces the "epistemic" aspect of (4a, b) to the inferences to
be drawn and thereby does justice to the structurally determined level-
distinction. The same point was made in Foolen (1992:83).
What we are left with is the problem of where to locate (5a): it does
indeed indicate a different contrast from (5b). I doubt, however, whether
this contrast can be accounted for in terms of the "epistemic level"
without rendering this notion completely void. I would contend that the
alleged "epistemic" difference between (5a) and (5b) is that the
interpretation of (5a) requires some additional inferencing which the
interpretation of (5b) does not. Thus, (5a) and (5b), like (5c) when
compared with (5d), clearly differ as to the conditions they impose on the
context in order to get interpreted properly.
According to an observation first made by Bellert (1972), the
interpretation of but/aber-conjunctions involves at least two semantic
contrasts to be read off or inferred from the conjuncts. (5d) contains the
overt contrast John-Bill but fails to provide cues to reconstruct the
second contrast needed, i. e. an INFERRED ASSUMPTION which is in
opposition to the STATEMENT Bill is rich. This pragmatic insufficiency
has its grammatical counterpart: (5d) can be considered deviant because
it does not allow a contrastive stress pattern - just try and you·will see'
Moreover, contrary to most work done in the field, we claim that the
"semantic opposition" (more exactly, the occurrence of antonymous
Adversative connectors on distinct levels ofdiscourse 245
This section will provide an outline of the way in which sentence and
discourse level interact in determining the interpretation of adversative
conjunctions. To begin with, note that the semantic contribution of
aber/but in combining two clauses into a conjunction is twofold. In a
nutshell, it may be conceived of as:
(VIa) states what aber/but share with und/and; therefore (VIa) is the
basis for (Sa) and (Sb) having identical truth conditions. This is a feature
of aber/but that probably nobody will deny. (VIb) is what separates
aber/but from undland, and hence (VIb) is what causes (Sa) and (Sb) to
have distinct conditions of use as regards semantic coherence. As such,
(VIb) is perhaps not as uncontroversial as (VIa), so it may be worth
commenting on.
What has been stated in (VIa + b) should not be understood as
amounting to "but = and + pragmatics". This would defmitely miss the
target. The conditions of use of aber/but and other adversative connectors
are not simply a matter of "pragmatically motivated interpretation" (as
Sweetser suggests); rather they are rooted in the lexical semantics as well
Adversative connectors on distinct levels ofdiscourse 247
The structural differences are clearly brought out by the fact that none
of (7a-d) is replaceable with any of the others as an appropriate answer to
the respective question-context given in II. ..11. Assessing (517) once more,
we may say: in isolation and in written form these conjunctions are not
really ambiguous but are underspecified regarding the cues they need to
get interpreted.
Finally, we will look at those grammatical features of adversative
connectors that form the basis of the division of labour between grammar
and pragmatics just outlined. The most relevant point to note is this:
indicating a contrast in the way suggested in (VIb) is a semantically built-
in feature of aber/but which correlates with likewise built-in syntactic
properties of these connectors. In contrast to and, or, etc., adversative
coordinators like aber/but display the following features:
First, aberlbut allows for binary connection only. This is the basis for
marking the assertion following aberlbut as the source from which to
look for an appropriate CONTRASTING ASSUMPTION, that is, the target.
Second, aber Ibut is intrinsically asymmetric in marking what has
been called a "change of perspective" (Spooren 1989): for example, (4a)
or (7a) with the reversed order of conjuncts would not render the same
250 Ewald Lang
(11) II Wie geht's den beiden? II 'How are the two doing?'
like (12c), are inherently asymmetric regarding the relation between the
clauses they link.
To summarize: due to their distinct information structure, (12a-c) are
not replaceable with (1Ia,b) while keeping the respective discourse
contexts constant, and vice versa.
Furthermore, the examples in (11) and (12) reveal an important
difference in the way interclausal aber/but relate to the discourse. In the
non-symmetric cases, the target from which the ASSUMPTION that
contrasts with the STATE1vffiNT in the second conjunct ('His wife is at
work') has to be derived is supplied by the first conjunct. So, in (12a-c) it
is the assertion of John's being off sick that triggers the relevant
inference: sick people need care, and spouses make good candidates for
this job.
In (11a,b), which form symmetric continuations of the given discourse
topic, the target that licenses the use of aber/but is not as easy to
reconstruct. Structurally, the question Wie geht's den beiden? ('How are
the two doing?') does not provide an appropriate target per se; all it does
is to license a list of and-conjoined items, cf (11 b). This is where
pragmatics enters the picture. In interpreting (1Ia), we face the following
situation: if searching the wider context for a suitable target fails, we are
left with no alternative but to take the first conjunct Hans ist krank
('Hans is off sick') and generate from it the implicature that the news
about his wife will be of similar quality.
Given that, as claimed in (VIb), aber/but has the built-in feature of
triggering an inference, the notion of a but of "semantic opposition"
(introduced by Lakoff (1971» seems to be misconceived: semantic
opposition per se never suffices to license the use of aber/but.
4. Concluding remarks
Notes
level. I will take for granted that pragmatically (i) is asymmetric, while (ii) is
symmetric regarding the alternatives presented:
(i) Give me liberty or give me death! [= Sweetser's (46), p. 98]
(ii) Give me a salami sandwich or a hotdog!
What needs to be clarified, however, is this: is it justified to consider (ii) as
an or-{;onjunction of two speech acts? Speech acts normally resist being
presented as alternatives.
Another open question is whether we are to take (ii) as a conjunction of two
speech acts or rather as a single speech act of request which comprises an
alternative of how the request might be complied with?
Again, these questions show that a closer examination of the internal
structure of the conjuncts seems to be unavoidable.
3. The advantage of illustrating the issue at hand with German data consists in
the fact that aber is an unambiguous adversative connector. It does not share
the range of additional uses and readings that are covered by but, as in
German these are rendered by separate lexical items like sondern, auf3er etc.
Analysing but-constructions in comparison with aber-constructions thus will
help us to pick out only those having adversative readings.
4. Without going into details (see Lang 1984, 1991, in print), let me just point
out that the derivation of the Common Integrator (CI) of a given conjunction
(a) is sensitive to surface structure parallels shared by the conjoined clauses,
and (b) can best be illustrated by means of a (multiple) question to which the
given conjunction would serve as an appropriate answer.
So while (i)-(iv) share the propositional formula "someone transfers
something to someone" as constituent part of their respective Cis, the
specific make-up of the latter rests on the information structure of the
clauses, specifically on the (non-)coreference of the items instantiating the
argument places. (Non-)coreference is indicated by (non-) coindexing.
Coreference requires destressing: if e.g. the beast is coindexed with the dog
the former serves as an anaphoric epithet for the latter.
(i) Grandpa will bequeath the dogk to his grandchildreni and
Grandma (will)(bequeath) the beastv*k to us j/*i
CI: WHO will bequeath WHAT TO WHOM?
(ii) Grandpa will bequeath the dogk to his grandchildreni and/but
Grandma will sell the beastk/*l to us i/*j
CI: WHO will transfer the dogk to USi IN WHAT WAY?
(iii) Grandpa and Grandma will bequeath the dogkand the beastv*k
to their grandchildreni and to us j/*i, respectively
CI: WHO will bequeath WHAT TO WHOM?
(iv) Grandpa will bequeath, and Grandma will sell, the dog to us
CI: WHO will transfer the dogkto USi IN WHAT WAY?
Note that the internal structure of the conjuncts in (i) allows for regrouping
the parallel constituents into a respectively construction like (iii) which leads
to the same CI as (i) but is, for instance, unavailable for (ii). On the other
hand, (ii) can also be rendered by a right node raising construction like (iv)
which leads to the same CI as (ii) but is unavailable for (i). So (iii) and (iv)
may serve as overt syntactic diagnostics for information structure and CI
Adversative connectors on distinct levels ofdiscourse 255
differences that seem to be concealed when conjunctions like (i) or (ii) are
presented in written form only.
5. (8) and (9) point to an important difference between German aber and
English but regarding scope determination under negation. Since German
adversative coordinators for contrast (aber) vs. correction (sondern) are
lexically distinct, their scopal behaviour is built-in and need not be indicated
by other means-ef. (i, ii). English but-constructions, however, have to
render the distinction of contrast vs. correction under negation by distinct
conditions on deleting non-focussed material in the second clause. Thus, but-
constructions like (iii) expressing contrast disallow, whereas but-
constructions like (iv) expressing correction require, the deletion of repeated
material in the second conjunct (for details see Lang (1991).
(i) Hans ist nicht dumm, aber (er ist) faul [contrast]
(ii) Hans ist nicht dumm, sondem (er ist) faul [correction]
(iii) Hans is not stupid but *(he is) lazy [contrast]
(iv) Hans is not stupid but (*he is) lazy [correction]
The same point can be made to explain the difference in the following cases:
(v) John ist reich, aber Bill (ist) arm.
(vii) John is rich but Bill *(is) poor.
(vi) John ist reich, aber Bill nicht lauch.
(viii) John is rich but Bill *(is) not Itoo.
Despite the differences shown, aber and adversative but both mark scope
boundaries in preventing propositional operators from taking scope over
them.
6. The relationship between aber-conjunctions and concessive constructions
cannot be tackled here, but see the contributions of Barth and of Crevels (in
this volume).
Another point that deserves mentioning is the fact that German aber (unlike
English but) is not confined to interclausal positions, but can also float in the
second clause (this is why in HdK [=Pasch et a1. (forthcoming)] aber has
been classified as a "connective particle"). At first glance, this distributional
flexibility seems to support the view that aber is most prone to "pragmatic
ambiguity". It can be shown, however, that being allowed to float within the
second clause is what enables aber to syntactically mark the information
structure relation of the second clause with respect to the discourse topic as a
non-symmetric one, similar to (12a-c) above:
(12) d. F[[Hans] ist KRANK], F[[seine Frau] geht aber ARBEITEN]
'Hans is off sick, his wife is, nevertheless, at work'
The prosodic details of contrastive vs. concessive clause linking are being
investigated in a current DFG project of mine which forms the source of the
data discussed here.
256 Ewald Lang
R~ferences
Bellert, Irena
1972 On certain properties of the English Connectives and and but. In:
Senta Plotz (ed.), Transformationelle Analyse, 327-356. Frankfurt
am Main: Athenaum.
BrauBe, Ursula
1998 Was ist Adversativitat? aber oder una? Deutsche Sprache 26:138-
159.
Foolen, Ad
1992 Review of Sweetser 1990. Lingua 88:76-86.
HdK [= Pasch, Renate, Ursula Brausse, and Eva Breindl]
(to appear) Handbuch der deutschen Konnektoren. 2 vols. Ms. Mannheim:
Institut flir deutsche Sprache.
Kortmann, Bernd
1997 Adverbial subordination. A typology and history of adverbial sub-
ordinators based on European languages. (EALT 18.) Berlin/New
York: Mouton de Gruyter.
Lakoff, Robin
1971 If's, and's and but's about conjunction. In: Charles J. Fillmore and
D. T. Langendoen (eds.), Studies in linguistic semantics, 114-149.
New York: Holt, Rinehart & Winston.
Lang, Ewald
1984 The semantics of coordination. (SLCS 9.) ArnsterdamlPhila-
delphia: Benjarnins.
1991 Koordinierende Konjunktionen. In: Semantik/Semantics, Art. 26,
597-623. (HSK 6.) BerlinlNew York: Mouton de Gruyter.
(in print) Die Wortart "Konjunktion". In: Lexicology/Lexikologie, Art. 88,
(HSK 17.) Berlin/New York: Mouton de Gruyter.
Raible, Wolfgang
1992 Junktion. Eine Dimension der Sprache und ihre
Realisierungsformen zwischen Aggregation und Integration.
Heidelberg: Winter.
Rudolph, Elisabeth
1996 Contrast: adversative and concessive relations and their
expressions in English, German, Spanish, Portuguese on sentence
and text level. (Research in text theory 23.) BerlinlNew York:
Mouton de Gruyter.
Spooren, Willibrordus
1989 Some Aspects of the form and interpretation ofglobal constrastive
coherence relations. PhD Dissertation, University of Nijmegen.
Sweetser, EveE.
1990 From etymology to pragmatics. Metaphorical and cultural aspects
ofsemantic structure. Cambridge: Cambridge University Press.
Viewpoints and polysemy: Linking adversative
and causal meanings of discourse markers
Scott A. Schwenter
1. Introduction
2. Analytical framework
No. of "Viewpoints"
No. of Speakers One Two
(Monolo~ical ) (Dialo~ical)
One (Monologue) Monologue/ Monologue/
Monological Dialo£ical
Two (Dialogue) Dialogue/ Dialogue/
Monological Dialogical
3. Si as an adversative DM
The argumentative sequence of the first two turns in (5) can be summa-
rized as in (5'), where the parenthesized segments are conclusions left
implicit, but easily inferrable from the exchange.
Viewpoints and polysemy 263
It is important to notice that the utterer (=A) of the si-marked turn does
not dispute the "truth" of the proposition ("your stomach is going to
hurt"), but she does dispute its relevance and the relevance of the associ-
ated argumentative sequence within the discourse situation. Her response
to R signals the exclusive relevance of the "be thirsty" argument and, as a
consequence, justifies the conclusion "drink quickly".
Refutation with si may also be directed towards conversational impli-
catures that go beyond the literal meaning of the proposition, like the one
generated by A's utterance in (6):
In (8), M makes a hedged assertion about the dye job done on B's hair,
and characterizes it as done mal 'poorly'. B' s response disputes this
characterization of the dye job; but instead of characterizing the dye job
positively, B strengthens the negative orientation already made explicit by
the adverb in M's assertion, stating that the dye job was donefatal 'hor-
ribly, lit. fatally,.2 Such an example expresses some degree of adversativ-
ity, since there is a clear correction of one speaker's evaluation by another
speaker's evaluation. At the same time, because the two speakers actually
Viewpoints and polysemy 265
AGREE on the overall implicated point-the dye job was not a good
one-the rhetorical purpose of each utterance in the exchange is parallel.
Here there is no ambiguity between a dialogical, adversative interpreta-
tion and a monological interpretation with parallel arguments. Rather,
both components of meaning combine to produce the overall interpreta-
tion of the example.
The adversativity expressed by a particular si-marked utterance can be
diminished even more when a "correction" of (some aspect of) an inter-
locutor's utterance cannot be identified:
In the situational context of this example, S and A are inspecting the bi-
cycle, and S' s utterance is accompanied by the physical act of lowering
the bicycle seat a certain amount. A's utterance is in response to the (non-
locutionary) implication that the seat cannot be lowered beyond the point
to which S has lowered it. This is where the adversativity of the example
lies. Nevertheless, the contributions of both speakers are argumentatively
oriented in the same direction, each constituting positive assessments of
the bicycle in question, and implicating a conclusion like "we should buy
this bicycle" (which, in fact, they did). The co-oriented nature of the ut-
terances is made clear by the appearance of incluso 'even' in A's tum,
which marks the expressed proposition as a parallel, but stronger, argu-
ment for the aforementioned conclusion (which A took to be implicated
by what S said and did) (Kay 1990; Konig 1991).
The two preceding examples, along with (6) above, illustrate that,
even in dialogues, the degree of opposition marked by si is not fixed, but
sensitive to contextual factors, lexical relationships, etc. The "hybrid" na-
ture of each example is due primarily to the parallel argumentative orien-
tation of the asserted propositions: in (8) both the mal proposition and the
fatal proposition are directed towards a negative evaluation of B's hair,
while in (9) the propositions derivable from both utterances are being
used to evaluate the bicycle positively. The greater adversativity ex-
pressed in (8) as opposed to (9) results from the explicit correction of a
weaker predicate with a stronger one.
266 Scott A. Schwenter
4. Si as a "causal" DM
(10) [To the disgusted husband whose favorite soccer team has just lost
three goals to none]
No te preocupes, jhombrel, si eljUtbol no es importante.
'Don't worry about it, man! SI football [=soccer] isn't important.'
(Porroche 1998: 236)
In her analysis of this example, Porroche points out that the target of the
husband-directed utterance is not an actual utterance produced by the
husband, but rather "a belief that ... forms part of the mental universe of
the speaker, but which, at times, can also be attributed to the interlocutor"
Viewpoints and polysemy 267
(12) [Mother to her 18-month-old son, who is sitting at the dinner ta-
ble]
A: jVenga hombre! jCome!
'Come on man! Eat!'
A: te quedan dos bocaditos.
'there are two bites left. '
~ A: Si ya no te queda casi nada.
'SI there's hardly anything left. '
[ACA,3/21/99]
Here, both of A's final two utterances can be seen as trying to per-
suade the child to eat, or more accurately, to finish eating. Both of these
utterances can thus be considered "pragmatically subordinate" to the ini-
tial tum where the command is uttered; the si-marked utterance being the
"stronger" of the two. Indeed, it is interesting to note that if the proposi-
tional contents of A's latter two utterances are switched, the sequence
will become pragmatically anomalous, as shown in (12'):
The reason for this is that the sequence introduced by si must be inter-
pretable not only as motivation for carrying out the action demanded by
the imperative, but also as the "strongest" piece of motivation among
Viewpoints and polysemy 269
Like the exhortative examples seen in 4.1, in this section we will once
again consider discursive sequences consisting of two utterances with the
structure [VI + si V2]. Once again, there is a paratactic relationship be-
tween the two utterances, and in this relationship the si-marked utterance
V2 remains "pragmatically subordinate" to VI. Specifically, VI repre-
sents an inferential conclusion that the speaker presents as "warranted" or
"justified" by the content of an assertion in V2 (cf Schiffrin 1987: 202).
Because they refer to speaker-internal cognitive positions, these examples
are labeled "epistemic"; indeed, they appear to be good examples of cau-
sality in the epistemic domain, to use Sweetser's (1990) terminology.
Viewpoints and polysemy 271
(15) [A has been telling C how she likes the actor Mel Gibson]
A: Y a L tambien Ie gusta mucho.
,And L likes him a lot too. '
~ A: Si incluso Ie grabe yo una pelicula de el.
'SI I even recorded one of his movies for her.'
A: i,C6mo se Ilamaba?
'What was it called?'
C: i,EI hombre sin rostro?
'The Man without a Face?'
A: Si, esa es.
'Yes, that's it.'
[ALC]
conclusion (Kay 1990; Konig 1991; Schwenter 1999c), and to imply that
other, weaker, arguments for that same conclusion also exist. The combi-
nation of si + scalar particle thus signals very strong speaker commitment
towards an inferential conclusion, which, as in (14), is often a controver-
sial and/or critical assessment. Notice that, unlike the exhortative exam-
ples seen above, epistemic examples like (14) and (15) do not contain
explicitly addressee-directed utterances. To the extent that an opposing
viewpoint can be identified, or even inferred, it is construed in opposition
to the conclusion, not to any aspect of the justificatory utterance intro-
duced by si. As a result, it is in the speaker's best conversational interest
to support the conclusion to the greatest degree possible.
It is interesting to note that most of the examples of si introducing an
assertion that justifies (some aspect of) speaker viewpoint, as in (14) and
(15) above, are functionally quite similar to the epistemic use of the
causal connective porque 'because'. Indeed, porque could replace si in
(14) and (15) without affecting the propositional content in the least, with
the proviso that there would typically be a shorter intonation break
(comma intonation) between the connected segments. The most salient
semantic/pragmatic difference between the two markers is that speakers
intuit a tighter "fit" between the content of the two segments with si so
that, for instance, (14) would be more likely used in a situation in which
the speaker felt that there already was or possibly would be opposition to
her opinion ("her cellulitis is too much") stated in the first segment. The
use of si instead of porque thus tends to reflect emphatic epistemic com-
mitment to the truth of the conclusion. While it is possible that porque
could be used in such a context, the relevant point regarding the epistemic
porque construction is that it has a wider domain of application than, and
is pragmatically unmarked with respect to (cf Lambrecht 1994: 16-17),
the epistemic si construction.
Further evidence for this particular analysis comes from the two con-
nectives' distinct co-occurrence possibilities with epistemic expressions
like me parece 'It seems to me' or creo 'I think.' These epistemic ex-
pressions function to attenuate speaker commitment to the truth of the
propositional content of an assertion. While such expressions seem com-
pletely natural with porque, they are by contrast pragmatically bizarre
with si, as the following examples illustrate (one might imagine these ut-
tered by a speaker A in a context in which another interlocutor B assumes
that Juan is coming to their party that same night):
Viewpoints and polysemy 273
Here A's inferential conclusion that fixing the curtain "will be very
easy" is warranted by the si-marked assertion that follows. The DM total,
which is most plausibly translated in (18) as 'after all', indicates that the
following proposition represents the only argument relevant to the conclu-
sion. And nada mas 'only, lit. nothing more' further minimizes the de-
scription of the effort involved in fixing the curtain. A's argument for how
easy fixing the curtain will be, then, constitutes the only argument for
coming to this conclusion, i.e. si appears in a non-additive context. All of
these markers thus work in tandem to justify the conclusion, which "fol-
lows" from the assertion in A's fmal tum.
The exhortative examples in the preceding section included a strategy
of lexical minimization (e.g. through negation) in a si-marked utterance as
motivation for carrying out the action specified in an imperative-mood
command. But as the preceding example shows, such minimization is not
limited to exhortative examples-epistemic examples display it also. An-
other epistemic example that uses minimization in support of a previously
stated conclusion is (19). Here, Y is attempting to justify her opinion,
stated at the outset, that the man she's talking about doesn't pay attention
to anyone:
4.3. Summary
Implicit in the organization of this paper has been the position that the
adversative DM si and the causal DM si are distinct polysemies of the
7
same lexeme. This is supported on discourse-structural grounds: the ad-
versative DM presents, within a dialogue, a dialogical "retort" directed at
an interlocutor's utterance, while the causal DM introduces an assertion
which provides motivation/justification for a preceding utterance by the
same speaker. By contrast, the exhortative and epistemic uses of the
causal DM are not distinct polysemies but rather offer a case of prag-
matic ambiguity (Hom 1989; Sweetser 1990): what si is expressing in
each use is the same semantic content, the interpretation of which differs
due to the pragmatic domain of application. The two polysemies also dif-
fer with respect to collocational properties: perhaps most importantly,
only the adversative DM can be prefaced, perhaps in every case, by the
adversative DM pero 'but', which reinforces the contrast expressed by si
(Montolio 1999). Utterance-initial pero is not found with either sub-type
8
of causal DM si.
Despite these distinctions, however, it should also be obvious from the
foregoing discussion that the boundaries between each use of si as a DM
are not clear-cut, and that the strong dialogical nature of the adversative
276 Scott A. Schwenter
Dialogical Monological
Adversative +++
Exhortative ++ +
Epistemic + ++
This display helps illustrate what has been stressed above: exhortative
and epistemic uses of si do not lack dialogical features completely. The
importance of these features is somewhat diminished in these uses, but
paradigmatic contrasts with other forms like porque still indicate a dia-
logical residue: a more salient feature now is high scalar commitment to a
conclusion, as the examples in 4.2 illustrate. The same cannot be said for
adversative uses of si, which always appear to be exclusively dialogical
(even in monologue examples, see Montolio 1999). The greater dialogical
nature of exhortative uses of DM si, as compared with the epistemic uses
in Table 2, results from the addressee-directed characteristics of the for-
mer-most importantly, the assumption of strong opposition on the part
of the addressee-which are exhibited to a lesser extent by the epistemic
uses.
5. Conclusion
(20) They think he's crazy, (but) in fact he's quite sane.
Notes
1. See Schwenter (1999b) and Montolio (1999) for discussion of the pero si
combination.
2. The scalar relationship between the two adverbs in (8) is easily demon-
strated. Consider the results when they are inserted into a syntactic frame of
the type "X but not Y", where X represents the "weaker" adverb, y the
"stronger" one:
(ia) Te 10 han tintado mal, pero no fatal.
'They dyed it badly, but not horribly.'
(ib) *Te 10 han tintado fatal, pero no mal.
'They dyed it horribly, but not badly.'
The infelicity of (ib) results from the fact that any speaker of Spanish would
order these adverbs as <fatal, mal> along a semantic ("Horn") scale, such
thatfatal semantically entails mal, but not vice-versa.
3. The intonational break between the imperative and the utterance segment
justifying it tends to be longer when si introduces the latter. This is reflected
by the comma intonation in (11a) as opposed to the period in (11b).
4. Note too that the translation equivalent of because in Spanish, porque, could
not replace si in the examples seen above. In fact, upon translating examples
(10) through (13) into English I am unable to replace si with any English
connective form.
5. My use of "modally harmonic" with respect to discourse markers extends the
term beyond that of the scholars cited, who limit it to cases involving a mo-
dal verb and other devices like adverbs.
6. The differences between this connective use of si and epistemic porque are
very similar to those separating discourse marker uses of English after all
Viewpoints and polysemy 279
References
Garcia, Serafina
1996 Las expresiones causales y finales. Madrid: ArcolLibros.
Haiman, John
1978 Conditionals are topics. Language 54: 564-89.
Horn, Laurence R.
1989 A natural history ofnegation. Chicago: University of Chicago Press.
Jespersen, Otto
1940 A modern English grammar on historical principles, Volume 5:
Syntax. London: George Allen and Unwin.
Kay, Paul
1990 Even. Linguistics and Philosophy 13: 59-111.
Konig, Ekkehard
1986 Conditionals, concessive conditionals and concessives: Areas of
contrast, overlap and neutralization. In: Elizabeth Closs Traugott,
Alice ter Meulen, Judy Snitzer Reilly, and Charles A. Ferguson
(eds.), On conditionals, 229-46. Cambridge: Cambridge University
Press.
1991 The meaning offocus particles: A comparative perspective. Lon-
don: Routledge.
Kortmann, Bernd
1997 Adverbial subordination: A typology and history of adverbial sub-
ordinators based on European languages. Berlin: Mouton de
Gruyter.
Ladusaw, William A.
1996 Negation and polarity items. In: Shalom Lappin (ed.), The hand-
book ofcontemporary semantic theory, 321-41. Oxford: Blackwell.
Lambrecht, Knud
1994 Information structure and sentence form. Cambridge: Cambridge
University Press.
Montolio, Estrella
1999 iSi nunca he dicho que estuviera enamorada de ell Sobre construc-
ciones independientes introducidas por si con valor replicativo. Ora-
lia 2: 37-69.
Mori, Junko
1994 Functions of the connective datte in Japanese conversation. In: No-
riko Akatsuka (ed.), Japanese/Korean Linguistics, Vol. 4, 147-63.
Stanford: CSLI Publications.
Pons Borderia, Salvador
1998 Conexi6n y conectores: estudio de su relaci6n en el registro infor-
mal de la lengua. (Cuadernos de Filologia 27.) Valencia: Universi-
tat de Valencia.
Porroche Ballesteros, Margarita
1998 Sobre algunos usos de que, si yes que como marcadores discursivos.
In: Ma Antonia Martin Zorraquino and Estrella Montolio Duran
(eds.), Los marcadores del discurso: Teoria y analisis, 229-42. Ma-
drid: ArcolLibros.
Viewpoints and polysemy 281
Roulet, Eddy
1984 Speech acts, discourse structure, and pragmatic connectives. Jour-
nal ofPragmatics 8: 31-47.
Schiffrin, Deborah
1987 Discourse markers. Cambridge: Cambridge University Press.
Schwenter, Scott A.
1998 From hypothetical to factual and beyond: refutational si-clauses in
Spanish conversation. In: Jean-Pierre Koenig (ed.), Discourse and
cognition: Bridging the gap, 423-35. Stanford: CSLI Publications.
1999a The pragmatics of conditional marking: Implicature, scalarity, and
exclusivity. (Outstanding Dissertations in Linguistics.) New York:
Garland.
1999b Discourse markers and the PAISN distinction. Ms., The Ohio State
University.
1999c Two types of scalar particles: evidence from Spanish. Advances in
Hispanic linguistics. In: Javier Gutierrez-Rexach and Fernando
Martinez-Gil (eds.), 546-61. Somerville, MA: Cascadilla Press.
Schwenter, Scott A. and Elizabeth Closs Traugott
2000 Invoking scalarity: the development of in fact. Journal ofHistorical
Pragmatics 1: 7-25.
Sweetser, Eve E.
1990 From etymology to pragmatics. Cambridge: Cambridge University
Press.
Traugott, Elizabeth Closs
1997 The discourse connective after all: A historical pragmatic account.
Paper presented at the 16th International Congress of Linguists,
Paris, July.
1999 The rhetoric of counter-expectation in semantic change: A study in
subjectification. In: Andreas Blank and Peter Koch (eds.), Histori-
cal semantics and cognition, 177-196. Berlin: Mouton de Gruyter.
The treatment of contrasts in interaction *
Cecilia E. Ford
1. Introduction
A feature of RST which makes it especially useful for the broad study of
rhetorical relations is that it allows for relations to be either explicitly
marked or inferable based on the content of text spans in question; thus
no explicit linguistic marking is required for the interpretation of two
spans as being in an oppositional relation. Furthermore, RST places no
fixed limit on the size of text spans to be related, though in practice the
lowest unit has often been the clause.
Example (1) offers an initial sense of the type of pattern I will be ex-
amining. The quote is from a magazine interview, and the speaker is a
student who organized a protest at her high school. She is describing the
effort it took:
EXPLANATION
,..--------------------------,
'cause pep assembly's a real big deal,
especially at homecomdng; the whole school
gathers
The first part of the contrast ("It sounds easy") explicitly refers to an
assessment that the interviewer (or the magazine readers) could poten-
The treatment ofcontrasts in interaction 285
tially make;2 the second part of the contrast ("but it wasn't") stands in
opposition to the first. What follows the contrast is an explanation of the
opposition: the speaker explains that the setting for the protest was a
particularly important and consequential one. Here then, through her
subsequent talk, the speaker shows her understanding of the contrast as
presenting a problem needing resolution, and the resolution she provides
is an explanation supporting one pole of the contrast. At one level of
rhetorical analysis, we see that the incompatibility between the parts of
the contrast is addressed and resolved through the explanation.
At another level, when we take these rhetorical moves in their con-
text we see a contrast plus explanation doing particular persuasive work.
A contrast, here, functions as a format for presenting two perspectives,
of which the speaker identifies strongly with only one. Through this rhe-
torical combination, the speaker efficiently acknowledges a potential
incompatibility but builds support for her own perspective as she adds
an explanation for one pole of the contrast.
In this paper, I examine the functions of contrasts in interaction, with
a specific focus on the relationship between a contrast and the talk
which follows. While the rhetorical relation of contrast is a format for
the expression of diverse actions, a pattern is evident: across contexts,
contrasts are treated as needing resolution in subsequent talk; that is,
they are regular building blocks for larger actions involving explana-
tions or solutions for problematic states of affairs. In the data I have ex-
amined for this study, there is indeed a pattern of explanation and solu-
tion relations following contrasts; however, as we will see, it is also evi-
dent that the larger context of activity and the role of a speaker in an
activity are mediating factors affecting how and whether further resolv-
ing talk is treated as relevant after a contrast.
The remainder of the paper will be organized as follows: In section 2,
I describe my data sources; section 3 offers some examples of contrasts
in spoken English. Section 4 sketches the lexical and grammatical en-
coding devices associated with contrasts in the present data. In 5 and 6,
bearing in mind that the social and sequential context of any piece of
talk is central to the formulation and interpretation of what that talk is
doing, I discuss the rhetorical combinations found in the data:
CONTRAST + EXPLANATION and CONTRAST + SOLUTION. In
section 7, I examine cases where contrasts are not followed by explana-
tions or solutions. I discuss relationships between activity type, speaker
role, and the location of the contrasting action in an interactional se-
quence in relation to the absence of further talk in these cases. Section 8
286 Cecilia E. Ford
concludes the paper with a general review and discussion of the find-
Ings.
2. Data sources
3. Contrasts in conversation
CONTRAST=l-
r twenty six percent (1.4)
But we are: entitled to inquire
(0 • 4)
Who put i t up to twenty six percent=
Audience: hehh[eh
Audience: [x-xxXXXXXXXXXXXXXX ...
4. Forms of contrast
( 4) (repea ted)
?: ( (O.~) ) 7-
K: [WE PREFER THEM IOF~~
T: You prefer them aw- I prefer them~, I can see:.
5. Contrast + Explanation
(5)
T: As I move away, (.) does my height get
different?=
CONTRAST (1) ~ T: =No.
?: 0No. 0
EXPLANA- (1):::> T: I look smaller,because I'm farther away,
TION' I I I
First, the teacher answers his own question, "does my height get differ-
ent?" with the contrasting "No." Next, he explains this negation by con-
ceding that while there is an illusion of getting small, it is only because
of the added distance. And within that explanation, he introduces yet
another contrast, the contrast between his original denial of difference
and his subsequent concession that there is an appearance of difference.
This contrast is highlighted by the stress on look (as opposed to actually
being different). As with the first contrast, the second is followed by the
explanation, "because I'm farther away."
The contrasts in (5) are vehicles through which the science teacher
highlights expectations and assumptions, thus leading the students
through a discovery process. The first contrast draws attention to the
fact that there is an incompatibility between the image and the object
being reflected. This highlighted contrast then leads into the explana-
tion. Within that explanation is the more subtle contrast between ap-
pearance and reality, indexed through the contrastive stress on look.
And again, the opposition highlighted through the contrast leads into an
explanation component.
The treatment ofcontrasts in interaction 291
(6) R: 0: :kay.
R: W'll yeah. Look at that.
(0.5)
R: You were back here'n
CONTRAST:::) [we were way over there.
A: [Yau: :p?
A: Yup?
EXPLANATION :::) T: CAN YOU SAY ( . ) STACY?
T: [huh huh huh huh huh
A: [eh [ heh heh heh
R: [uh hah hah hah hah hah hah hah hah=
A: =I say we keep her
they are now. That A and R understand the joke is evidenced by their
laughter. A's "I say we just keep her" also demonstrates that the refer-
ence to Stacy's being on the team is interpreted as an explanation for the
team's success.
In (6), the contrast is presented by one speaker and the explanation
by another. By providing an explanation, T treats R's statement of the
contrast as unfinished in itself and calling for elaboration. Thus, in terms
of its treatment by the conversation participants, we can analyze this
contrast as expressing an opposition that is remarkable and puzzling, an
observation that seems to draw further talk providing a resolution, even
if it is done in a joking format.
Example (4), repeated once again, is from the setting of the science
class, but the activity is quite different from the one in (5), in which the
teacher was in more firm control of the direction of the talk. Here the
students are working in small groups on laboratory tasks. There is much
more student talk in this portion of the class; and even as the teacher
announces an action he's about to take, at least one student feels free to
object:
11 11
DISAFFILIATION ACCOUNT
A: Did you get your (.) first pay check from it?
( .)
A: [At least?
=>R: [NO: I won't get that for a couple of weeks yet.=
A: =Oh,
( .)
A: [W'l
=>R: ['Cause i t takes a long time.
(8)
s: I have a check.
S: Eight fifty.
CONT~T=> J: Nah. I won't take- I don't take second-
party checks.
S: eh huh huh huh huh
EXPLANATION=> J: I don't got no way of telechecking 'em,
trasts are relevant not only from a discourse analyst's perspective, but
also, and in a significant manner, to the interactants themselves. In (9) T
presents an assertion which contrasts with a prior assumption (line 8),
but she does not immediately provide an explanation that might resolve
the asserted incompatibility. Note that both A and R treat the absence of
an explanation as problematic and deserving of further attention.
A and R have been remarking on the fact that there are many pictures
of people and sheep on the walls of T and P's dining room, where they
are seated. At lines 4-7, R begins to refer to the people in the pictures as
relatives, "grandma" and "grandpa":
(9)
1 R: I see there's a lo:t of pictures of sheep.
2 A: Oh yeah. there's more in the other [r(h)oo:m
3 t(h)oo.
4 R: [There
5 there's sheep there, Right by grandma.
6 A: ha ha ha
7 R: Gramma, (.) an' (.) an' (.) and grandpa,and the
8 the sheep and [baby sheep.
9 =>T: [We don't know these people.
10 (. )
11 T: t>or< these people.
12 R: y~ dotn't?
13 ( .)
14 T: No.
15 ( .)
16 A: Do you know the one [just below that one
though?
17 ?: [just
18 (0.5)
19 A: the one in the mid[dle?
20 T: [No.
21 T: The little cryin' girl?
22 A: No?
23=>R: [You just like the >pictures.<
24=>T: [We just like- silver: frames and
25 pictures,
26 R: And what about the sheep.=Do you know the
27 sheep?
296 Cecilia E. Ford
limitations preclude the inclusion of those cases here, and while the au-
thors do not draw specific attention to a CONTRAST +
EXPLANATION rhetorical cluster in their data, I take its occurrence in
their examples as suggestive of a more general pattern, one that is likely
not exclusive to my data nor to English conversations.
In sum, in the present data, we find a rhetorical pattern involving
contrasts, in a variety of functions, regularly followed by explanations.
As shown here, contrasts may be used pedagogically to highlight puz-
zles or problems; they may be the vehicles for observing remarkable
states of affairs; and, through interactional sequences, they may involve
disagreements between speakers. Significantly, explanation is a treated
by interactants as normatively relevant subsequent to contrasts in a vari-
ety of functions. Whether produced by single speakers or across speak-
ers, and whether prefaced by because or not, speakers in these data
regularly treat explanations as relevant after contrasts.
In addition to the pattern of contrasts followed by explanations, an-
other set of cases from my data involve contrasts followed by solutions.
Examining these cases provides further perspective on sequential con-
texts in which elaboration is added after a contrast. These cases also
support the observation that contrasts are used to present incompatibili-
ties or problems which participants treat as needing resolution.
6. Contrasts + Solutions
A further pattern was evident in the data whereby contrasts were fol-
lowed either by solutions or by displays that solutions, if not actually
stated, were understood as relevant or expected. In Rhetorical Structure
Theory, a relation of SOLUTIONHOOD requires that one span present
a problem and that the "situation presented in [the other text span be] a
(partial) solution to the problem" (Mann, Matthiessen, and Thompson,
1992:72). While Mann et al. do not point to the rhetorical relation of
CONTRAST as a pattern used in expressing the problem in the solu-
tionhood schema, they do note that the "scope of a problem includes...
conditions that carry negative values, either expressly or culturally, in-
cluding...frustrations" (1992:72). This resonates with Longacre's (1983)
notion of frustration in relations between predications. Frustration in a
text relation involves an expectancy which is countered. Contrasts ap-
pear to be common textual expressions of situations that go against ex-
298 Cecilia E. Ford
(10) "she" =P
A: She's trying to decide if sh[e wants to SPEND=
R: [She has a lot of
sheep.
A: =thirty five dollars on her dad. for Christmas.
R: Oh.
A: And she really doesn't.
=> (0.4)
A: But- (.) the gift that she thought of was thirty
five dollars, and she can't find i t anyplace
else.
R: Mm-mrmn.
A: so,
=> R: Has she tried Chickadee Depot.
There are actually two contrasts in this example: (1) spending $35 or
not, (2) not wanting to spend $35 (for which finding the gift more
cheaply would be a solution) but being unable to find the item anywhere
else for less. Note the pause after the first contrast, an indication that R
is treating A's tum as not yet complete. Given the pattern I am observ-
ing, this would be a relevant place for an expression offering resolution,
either an explanation or a solution. A normative pattern whereby con-
trasts are elaborated in such ways could explain R's lack of uptake fol-
lowing this first contrast and during the pause at the first arrow; she may
be awaiting an explanation, for example.
The treatment ofcontrasts in interaction 299
The second arrow points to R's response to the second contrast. This
contrast is between P's not wanting to spend $35 but not being able to
find the birdfeeder elsewhere for less. By asking whether P has tried
"Chickadee Depot," a local bird store, R is proposing a potential solu-
tion. In suggesting this solution (in the form of a question), R displays
her interpretation of the contrast as expressing a problem, for which a
search for solution is relevant.
In the science class, not surprisingly, there was a great deal of talk
focusing on problems and solutions. In the following case, the teacher
treats a solution as relevant in the context of a contrast, but he suggests
that the students solve the problem on their own. This strategy seems
particularly appropriate to one of the pedagogical goals in this discourse
context: to teach the students to formulate and solve physics problems.
(11)
T: If I stand thi-Oand this is the mirror
th- the screeno here, (0.3) I'm at a
«pointing to wall in front of him»
certain height, right? Look in.
~ If I go back farther, am I smaller?
CONTRASTS I'm still the same height, right?
( .)
J: Yeah,=
After using contrastive devices to introduce the shift "in discourse time"
to a next item, the teacher does not go on to produce any kind of expla-
nation or solution. The contrast functions only at a discourse organiza-
tional level, and subsequent talk does not attend to any content-level
opposition.
More interesting than the discourse structuring uses of contrastive
marking are cases in which sequential context, and the related factor of
speaker role in a local interactional activity or setting, are associated
with unelaborated contrast. Contrasts which come at the close of story-
telling sequences, for example, appear to function more as concluding
assessment than as focused expressions of problems which might call
for further elaboration and resolution. (13) is such a case. T has been
telling a story about her partner P's nephew; R and A respond with turns
which deliver displays of appreciation, action types which are typically
produced in the sequential contexts of story endings. In this case, the
displays are done through negative statements:
(13)
T: He was cute,>you know, sending the thank you
card before Ch(h)r(h)istm(h)as(h) ,<
hh[uh huh huh huh
R: [Yeah (h) huh huh hah
R: Stra[tegic.
T: [( )
T: We saw right through him.
=> R: [He is no: t a dumb kid.
A: rYe [h
T: [eh heh huh huh huh huh huh.
(): .hhh
(0.2)
=> R: [No sense in sending i t in August.
=> A: [Well at least he didn't send his list to
Santa.=
T: =That's right
A: °Yeah. °
While there are contrasts implicit in these negations, they are not treated
as invoking any need for explanation or solution. This seems related to
their coming as contributions to an assessment sequence at the close of a
story telling. As such, rather than presenting focused topics of talk on
~02 Cecilia E. Ford
their own, they are specifically produced as positive and agreeing re-
sponses to the upshot of T's story, a story about a clever and somewhat
manipulative child (note T's "We saw right through him"). In the sense
that they are affiliative with the teller's displayed stance and with the
story's gist, contrasts in such contexts are probably less likely to be
treated as needing resolution or even any special attention or elaboration
in themselves. That is, they may well express incompatibility at some
level (incompatibility with the expectation that children are guileless,
and that thank you notes should be sent soon after the action for which
they express gratitude); however, these turns are fully aligned interac-
tionally with the stance displayed in the immediately preceding talk. 7
In another kind of unelaborated contrast, the lack of orientation to
resolution seems to serve as an enactment or indexing of authority. In
the classroom context, there are explicit rules of conduct and there is a
clear authority figure. There are instances in which the teacher makes a
negative command or when a student makes a command which is based
on institutional rules, but no explanation or solution follows. While
there may be incompatibility expressed, the call for elaboration is art-
fully ignored in a manner which claims or underscores institutionally
supported rules and authority.
In one case, a student says to her classmates, "No talking." As this is
a rule during specific classroom activities (though one that is often vio-
lated), it is not surprising that the student offers no further explana-
tion-the explanation being treated as self-evident from shared knowl-
edge. And when the teacher at one point says, "Don't complain", in fail-
ing to produce a subsequent explanation, he may well be enacting his
role as authority. These cases suggest there may be institutional contexts
and specific activities (issuing directives based in or with reference to
institutional authority) in which a lack of explanation or solution is in
fact a resource for aligning with or reinforcing authority. Thus, explana-
tions may be relevantly absent in certain types of turns produced by per-
sons who are either referring to institutional authority or enacting it. It
must be noted, however, that this pattern is by no means absolute; recall
example (4), in which the teacher offers an explanation for his prefer-
ence for having the lights on, "I can see."
Related to the enactment of authority are cases of correction, them-
selves expressions of special access to relevant knowledge. (14) is a
case of third tum repair, involving a speaker repairing a misunderstand-
ing of her tum after the recipient has produced a responsive tum (Sche-
gloff 1992):
The treatment ofcontrasts in interaction 303
(14)
R: Ohm (.) who drew the last time. me or her.
T: I did.
( .)
T: very poorly.
=> R: No.=Me or her.
In the game they are playing (Pictionary) the drawing of pictures alter-
nates between teams and between teammates. Interpreting R's question
is complicated by the fact that both teammates and teams alternate in
this way. T understands R's first tum to be asking who among all the
players drew the last time. It turns out that R does not want to know
which individual among all the players was the last to take a tum; R's
question pertains to her team only: Was it she or her teammate who
drew last? The overlapped "me or her" in the first line specifies this ver-
sion of the question, but T is already responding at that point.
In the contrasting tum, R offers a repeat of her own previous tum,
and in so doing, she speaks from authority-having exclusive access to
what she actually meant in the first place. An explicit explanation in this
context would likely seem irrelevant and redundant (e.g., "No. Me or
her, because that's what I meant").
Finally, as a corollary to the treatment of contrasts as problems, there
are cases where contrasts express complaints or troubles-tellings (Jeffer-
son and Lee 1992). It can be a delicate matter whether or not the recipi-
ent of a complaint should offer a solution ("advice giving") or should
take the role of a sympathetic listener ("troubles recipient"). In (15), A
is describing an ongoing dispute in her extended family. The contrast
involves the families' living remarkably close to each other but still ar-
guing about where to spend their time:
(15)
A: ... she an'- her husband, (.) her parents
and her husband's parents, live within
~ six blocks of each other here in town?
CONTRAST'iR: Oh (.) hm hmmm
A: and ih- (.) where they spend most of their
time is like this major bone of
>contention about whether they spend time
at ~ parents' house< or his parents'
house?
304 Cecilia E. Ford
( .)
A: Six blo[cks-
R: [Six blocks away?
A: °Yeah. °
(1.1)
A: °It's insane.o
The examples and discussion in this section underscore the fact that
contrasts are by no means always followed by explanations or solutions,
or even by clear orientations to the need for resolution. The contexts and
functions documented here provide a wider view of the uses of contrasts
in interaction, an area worthy of further study. Contrasts can be used in
text organizational functions, exploiting the contrastive format to
achieve a shift in focus without placing the contrast itself in focus. Such
subordination of the role of the contrast relative to a broader context of
activity is also seen in story closings. In these contexts, interactants may
use negative statements as presentation formats for otherwise agreeing
or affiliating turns without implicating any need to focus on the con-
trasts themselves as topics for further development.
Contrasts can also express commands and perform repairs. The
speakers of such turns act in the role of authority, act with reference to
shared institutional rules, or display their special access to knowledge
precisely through not explaining their contrastive moves. That is, the
authoritative stance can be seen as both associated with and enacted
through the absence of explanation. Finally, contrasts may be used in
the service of troubles telling. The lack of a move toward resolution is
one of the interactional features that distinguishes troubles-telling from
advice-giving sequences. Instead of proposing explanations or solutions,
interactants produce assessments which underscore treatment of the
contrast as indeed expressing trouble not easily remedied.
8. Conclusion
Notes
Re..ferences
Atkinson, J. Maxwell
1984 Public speaking and audience responses: some techniques for invit-
ing applause. In: 1. Maxwell Atkinson and 1. Heritage (eds.), Struc-
tures of Social Action, 370-410. Cambridge: Cambridge University
Press.
Atkinson,1. Maxwell and John Heritage (eds.)
1984 Structures ofSocia/Action. Cambridge: Cambridge University Press.
Cumming, Susanna and Tsuyoshi Ono
1997 Discourse and grammar. In: T. A. van Dijk (ed.), Discourse as Struc-
ture and Process, 112-137. London: Sage.
Eggins, Suzanne and Diana Slade
1997 Analysing Casual Conversation. London: Cassell.
Fairclough, Norman
1989 Language and Power. New York: Longman.
Ford, Cecilia E.
1986. Overlapping relations in text structure. In: Scott DeLancey and Rus-
sell S. Tomlin (eds.), Proceedings ofthe Second Annual Meeting of
the Pacific Linguistics Conference, 107-123. University of Oregon:
Department of Linguistics.
1993 Grammar in Interaction: Adverbia/ Clauses in American English
Conversations. Cambridge: Cambridge University Press.
1994 Dialogic aspects of talk and writing: because on the interactive-
edited continuum. TEXT 14(4): 531-554.
1999 Collaborative construction of task activity: Coordinating multiple
resources in a high school physics lab. Research on Language in So-
cia/Interaction 32 (4): 369-408.
Ford, Cecilia E., Barbara A. Fox, and Sandra A. Thompson
1996 Practices in the construction of turns: the "TCU" revisited. Pragmat-
ics 6(3): 427-454.
Fox, Barbara A.
1987 Discourse Structure and Anaphora: Written and Conversational
English. Cambridge: Cambridge University Press.
Given, Talmy
1978 Negation in language: Pragmatics, function, ontology. In: Peter Cole
(ed.), Syntax and Semantics, Vol. 9,69-112. New York: Academic
Press.
Goodwin, Charles
1979 The interactive construction of a sentence in natural conversation. In:
George Psathas (ed.), Everyday Language: Studies in Ethnomethod-
ology, 97-121. New York: Irvington.
1981 Conversational Organization: Interaction between Speakers and
Hearers. New York: Academic Press.
1995 Co-constructing meaning in conversations with an aphasic man. Re-
search on Language and Socia/Interaction 28(3): 233-260.
The treatment ofcontrasts in interaction 309
Selting, Margret
1998 TCUs and TRPs: The Construction of Units in Conversational Talk.
InLiSt NO.4. (Interaction and Linguistic Structures.) Sekretariat LS
Angl. Sprachwissenschaft. Konstanz, Germany: Universitat Kon-
stanz.
Tannen, Deborah
1979 What's in a frame? Surface evidence for underlying expectations. In:
Roy A. Freedle (ed.), New Directions in Discourse Processing, 137-
181. Norwood, New Jersey: Ablex.
Tomlin, Russell S., Linda Forrest, Ming Ming Put, and Myung Hee Kim
1997 Discourse semantics. In: T. A. van Dijk (ed.), Discourse as Structure
and Process, 63-111. London: Sage.
Whalen, Jack
1992 Conversation analysis. In: E. F. Borgatta and M. L. Borgatta (eds.),
Encyclopedia ofSociology, 303-310. New York: Macmillan.
Concessives on different semantic levels:
A typological perspective *
Mily Crevels
It will be argued (a) that concessive clauses are expressed on four different
semantic levels: the content level, the epistemic level, the illocutionary or
speech-act level and the textual level, and (b) that there exists a systematic
cross-linguistic correlation between the semantic level at which adverbial
clauses apply, on the one hand, and the way they are expressed, on the
other. An illustration of this correlation is the fact that languages may use
different conjunctions to express adverbial relations if these apply at
different levels. The discussion of three hypotheses stands at the centre of
this paper (section 3). These hypotheses have been formulated to investigate
whether, depending on the semantic level, concessive clauses show
distinctive formal properties and have been tested against data from a wide
range of languages.
1. Concessive relations
(1) Although p, q
(3) a. Even though he had not eaten for days, he looked strong and
healthy.
b. {f one does not eat for days, one normally does not look strong
and healthy.
2.3. Integration
(8) She's just given birth to a beautiful baby girl although she's fifty-
two.
(9) He left his wife and children, although he loved them very much.
In the speech-act domain the content of the concessive clause does not
form an obstacle for the realization of the event or the state of affairs
described in the main clause, but raises obstacles for the realization of
the speech act expressed by the speaker in the main clause.
(11) My favourite poster is, I think, a French one for Nesquik, which
shows a sophisticated-looking small boy leaning nonchalantly
Concessives on different semantic levels 319
Dutch (Indo-European)
(12) a. Ze komt vast en zeker; hoewel, hi} haar kan je maar nooit
weten ...
'She will certainly come; though, one can never be sure with
her ... '
b. Ze komt vast en zeker, hoewel je hi} haar maar nooit weten kan
'She will certainly come, although one can never be sure with
her ... '
3. Hypotheses
(13) Content level> Epistemic level> lliocutionary level > Text level
HYPOTHESIS 1
The probability that a concessive construction will be expressed by
asyndetic means increases as the construction pertains to a higher
semantic level.
In Lakxota, which shows a cut-off point between the content and the
epistemic level, syndetic means are used at the content level, but an
asyndetic strategy is possible at the higher levels. Consider examples
(17) and (18) with a Lakxota content concessive and an epistemic
concessive, respectively:
awich-ayushtan.
them-left
'He left his wife and children, although he loved them very much.'
speech-act and the text level. Consider the epistemic concessive in (19)
and the speech-act one in (20):
ro uba ai-r-erea.
yOU:SG bad FACT-sBJ-be
'Although I should be minding my own business, your behaviour
is a disgrace. '
Another cut-off point is the one between the illocutionary and the text
level. In Turkish, for example, syndesis is the only option on the
illocutionary level (21), while at the text level asyndetic linking is also
possible (22).
HYPOTHESIS 2
The probability that in an elicited subordinate concessive construction
the q will be marked instead of the p increases as the construction
pertains to a higher semantic level. 6
Kwaza (Isolate; Kyjkaw Made, and Hein van der Voort p.c.)
(24) e 'ka) dwy-le'te tso'roi=e-ki afY- 'na
leg break-coNc run=go-DECL house-Loc
'He ran all the way home although he had broken his foot.'
While (24) shows the same marking of the p as the example in the
questionnaire does, (25) shows a marking of the q. As observed above,
the marking of the q in this last Kwaza example goes hand in hand with
a more paratactic or coordinate-like type of construction.
The cut-off point between the epistemic and the speech-act level can
be observed in Bahasa Indonesia. Consider examples (26) and (27),
which exemplify an epistemic and an illocutionary concessive,
respectively:
HYPOTHESIS 3
A concessive linker can only be used to introduce constructions on
different semantic levels if these levels form contiguous categories in
hierarchy (13).
ho:-d-d
gO-PAST-3MSG
'He left his wife and children, although he loved them very much.'
As can be seen in the Appendix, -ar-ude is the only linker used at the
content and epistemic level in Kannada; the speech-act level, however,
shows the use of other linkers as well.
Finally, Table 4 also shows us proof for a cut-off point between the
speech-act and the text level. Consider, for example, (33) and (34) with
a Dargi (Daghestan) illocutionary and textual concessive, respectively:
d-erdi-i-I-ra
pl-drink-AOR-GER-l
332 Mily Crevels
4. Conclusion
Abbreviations
Notes
* Without the help and patience of numerous native speakers and linguists this
paper would not have had its actual form. I cannot thank them enough for their
generosity in providing me with valuable data. Furthermore, I am grateful to
Elizabeth Couper-Kuhlen, Verena Haser, Kees Hengeveld, Bernd Kortmann,
Manfred Krug, and Hein van der Voort for their thorough comments on an
earlier version of this paper.
1. The restriction on focusability as such may not be the manifestation of a
syntactic/semantic, but rather of a pragmatic constraint.
2. Giinthner (this volume) argues that concessives of this type with German
obwohl 'although' are not concessive in the sense of Konig, because the
presupposition 'Ifp, then normally --q' supposedly does not hold.
3. It should be noted that Though in (11) does not function as a subordinator, but
as an adverbial linker, or conjunctional adverb. For a detailed discussion on
subordinators, adverbial linkers and coordinators the reader is referred to
Crevels (2000).
4. The sample .languages have been selected according to the Diversity Value
(DV) method as proposed by Rijkhoff and Bakker (1998). Contrary to other
methods (e.g. Dryer 1987), in which typological and geographic independence
are a sample's most important criteria, the chance offinding maximal variation
is the greatest when the DV method is applied, precisely because this method
creates maximal genetic distance between the languages which are finally
included in a sample. For a detailed description of the sampling procedure the
reader is referred to Rijkhoff and Bakker (1998) and RijkhofI et ale (1993).
5. Note that the translations between single quotes correspond to the sentences in
the questionnaire.
6. Obviously, the marking of the apodosis (q) indicates a more paratactic or
coordinate-like type of construction.
7. It goes without saying that textual concessives are not relevant in this context,
since textual concessives, as we have seen before, do not modify the main clause
of a concessive construction, but a whole preceding text unit which may be
composed of various sentences.
8. Note that by marking the q with -wara 'but' the consultant has used a
coordinating strategy to express the questionnaire sentence. For a more detailed
account of the different clause combining strategies used by the consultants see
Crevels (2000).
Concessives on different semantic levels 337
9. The characters in Table 4 correspond to the concessive linkers listed for each
language in the Appendix. Thus, the A corresponds to the first listed linker,
the B to the second, etc.
References
Bolkestein, A. Machtelt
1991 Causally related predications and the choice between parataxis and
hypotaxis in Latin. In: R. Coleman (ed.), New Studies in Latin
linguistics, 427-451. AmsterdamlPhiladelphia: Benjamins.
Crevels, Mily
1994 La conexi6n concesiva en espanol: una aproximaci6n funcional. MA
thesis, Department of General Linguistics and Department of
Spanish, University of Amsterdam.
1998 Concession in Spanish. In: Mike Hannay and A. Machtelt Bolkestein
(eds.), Functional grammar and verbal interaction, 129-148.
Amsterdam/Philadelphia: Benjamins.
2000 Concession: a typological study. Dissertation, University of
Amsterdam.
Devriendt, Betty, Louis Goossens and Johan van der Auwera (eds.)
1996 Complex structures: A functionalist perspective. BerlinlNew York:
Mouton de Gruyter.
Dik, Simon C.
1989 The theory of Functional Grammar. Part 1: The structure of the
clause. Dordrecht: Foris.
Dik, Simon C., Kees Hengeveld, Elseline Vester and Co Vet
1990 The hierarchical structure of the clause and the typology of adverbial
satellites. In: Jan Nuyts, A. Machtelt Bolkestein and Co Vet (eds.),
25-70.
Dryer, Matthew S.
1987 Large linguistic areas and language sampling. Studies in Language
13: 257-292.
Giv6n, Talmy
1984 Syntax: a functional-typological introduction, 2 vols. Amsterdam!
Philadelphia: Benjamins.
Greenbaum, Sidney
1969 Studies in English Adverbial Usage. London: Longman.
Greenberg, Joseph
1978 Diachrony, synchrony and language universals. In: Joseph
Greenberg, Charles A. Ferguson and Edith A. Moravcsik (eds.),
Universals ofHuman Language, Vol. 1: Method and Theory, 61-92.
Stanford: Stanford University Press.
Grice, H.P.
1975 Logic and Conversation. In: Peter Cole and Jerry Morgan (eds),
Syntax and Semantics, Vol 3: Speech Acts, 41-58. New York:
Academic Press.
338 Mily Crevels
Gtinthner, Susanne.
1993 "...weil-man kann es ja wissenschaftlich untersuchen"-
Diskurspragmatische Aspekte der Wortstellung in WEIL-Satzen.
Linguistische Berichte 143/1993: 37-59.
Haiman, John and Sandra A. Thompson (eds)
1988 Clause combining in grammar and discourse.
AmsterdamlPhiladelphia: Benjamins.
Hengeveld, Kees
1989 Layers and operators in Functional Grammar. Journal ofLinguistics
25.1: 127-157.
1990 The hierarchical structure of utterances. In: Jan Nuyts, A. Machtelt
Bolkestein and Co Vet (eds.), 1-23.
1992 Non-verbal predication: theory, typology, diachrony. Berlin/New
York: Mouton de Gruyter.
1996 The internal structure of adverbial clauses. In: Betty Devriendt,
Louis Goossens and Johan van der Auwera (eds.), 119-147.
1997 Cohesion in Functional Grammar. In: John H. Connolly, Roel M.
Vismans, Christopher S. Butler and Richard A. Gatward (eds.),
Discourse and Pragmatics in Functional Grammar, 1-16.
Berlin/New York: Mouton de Gruyter.
1998 Semantic type and expression format: On the internal structure of
adverbial clauses. In: Johan van der Auwera (ed.), 335-419.
Konig, Ekkehard
1986 Conditionals, concessive conditionals and concessives: areas of
contrast, overlap and neutralization. In: Elizabeth Traugott, Alice Ter
Meulen, J." Snitzer Reilly and Charles A. Ferguson (eds.), On
conditionals, 229-246. Cambridge: Cambridge University Press.
1988 Concessive connectives and concessive sentences: Cross-linguistic
regularities and pragmatic principles. In: J. Hawkins (ed.),
Explaining Language Universals, 145-166. New York: Blackwell.
1994 Concessive clauses. In: R.E. Asher (ed.), The Encyclopedia of
Language and Linguistics, 679-681. Oxford: Pergamon Press.
Konig, Ekkehard and Johan van der Auwera
1988 Clause integration in German and Dutch: conditionals, concessive
conditionals and concessives. In: John Haiman and Sandra A.
Thompson (eds.), 100-133.
Lyons, John
1977 Semantics, 2 vols. Cambridge: Cambridge University Press.
Nuyts, Jan, A. Machtelt Bolkestein and Co Vet (eds)
1990 Layers and levels of representation in language theory.
AmsterdamIPhiladelphia: John Benjamins.
Quirk, Randolph, Sidney Greenbaum, Geoffrey Leech and Jan Svartvik
1985 A comprehensive grammar of the English language. London:
Longman.
Rijkhoff, Jan and Dik Bakker
1998 Language sampling. Linguistic Typology 2-3: 263-314.
Rijkhoff, Jan, Dik Bakker, Kees Hengeveld and Peter Kahrel
1993 A method of language sampling. Studies in Language 17.1: 169-203.
Concessives on different semantic levels 339
Sweetser, Eve E.
1990 From etymology to pragmatics. Metaphorical and cultural aspects of
semantic structure. Cambridge: Cambridge University Press.
van der Auwera, lohan (ed.)
1998 Adverbial constructions in the languages of Europe. Berlin/New
York: Mouton de Gruyter [with D6nall P. 6 Baoill].
Causal and concessive clauses:
Formal and semantic relations
1. Introduction
Hypothetical Factual
Harmony conditional causal
Dissonance concessive conditional concessive
(2) I would not marry you ifyou were THE LAST MAN ON EARTH.
(3) a. Even ifyou dislike museums - which I know you do - the Mu-
seum ofModem Art is worth a visit.
b. I'll speak to it though hell itselfshould gape and bid me hold
my peace. (Sh. Hamlet I.ii.244)
The first fact is the equivalence of sentences like the following, given a
certain intonation pattern and information structure: 2
(6) a. It is NOT the case that Brutus's love for Caesar was a reason
for not killing him.
b. Brutus killed Caesar, although he LOVED him.
What our examples also show is that the relevant relations of scope
go together with specific intonation patterns, which are ultimately indi-
cators of different information structures. Whether or not differential
scope is simply a consequence of information structure is a question that
Causal and concessive clauses 345
will be taken up again later. Note furthermore that the wide-scope nega-
tion in (5a) and (6a) is not an instance of the focusing not ... but nega-
tion. The distinction between a focusing, or "replacive", not ... but ne-
gation with a focused adverbial clause and a sentential negation that
does not relate to the adverbial clause is illustrated by the following
minimal pairs. There is a fall-rise on money in (7a), whereas the intona-
tion of (7b) is falling:
(7) a. / John is not unhappy because he lost a lot ofMONEY / , but ...
b. / John is NOT unhappy because he has lost a lot ofmoney. /
What is rejected in (5a), (6a) and (7b) is the causal connection be-
tween the two situations described in the clauses and, as a result, also
the content of the main clause. The subordinate clause is not affected by
the negation and can therefore be assumed to be presupposed. The fol-
lowing paraphrases of (7b) show this even more clearly:
Closely related to the first set of facts is the observation that the
paraphrase relations discussed above can also be found in the corre-
sponding paratactic constructions. In such examples the cause of a con-
sequence is identified in a separate sentence preceding the negated con-
sequence:
(10) a. John has lost a lot ofmoney. He has not become a poor man
because ofthat.
b. John has lost a lot ofmoney. He has not become a poor man in
spite ofthat.
346 Ekkehard Konig and Peter Siemund
(11) Johann hat viel Geld verloren. Deshalb ist er trotzdem kein armer
Mann geworden. / Trotzdem ist er deshalb kein armer Mann ge-
worden.
(12) John is not an Oxbridge man, though none the worse for that.
In addition to this causal use and to various others, for may also have
a concessive interpretation. In fact, the concessive interpretation is the
only option whenever for combines with all:
More often than not, the intuition that concessive relations are semanti-
cally the opposite of causal relations is simply mentioned rather than
explicated (cf Hermodsson 1973, 1994; Azar 1997). In what follows,
we will discuss some recent attempts at providing such an explication as
a preparatory step towards presenting our own explication.
The most recent contribution to the topic comes from Di Meola
(1998). The basic assumption guiding his proposal is that concessive
relations do not involve only one underlying causal relation, as assumed
in many previous studies, but that there are two such relations operating
348 Ekkehard Konig and Peter Siemund
The problem obviously consists in the fact that it is not easy to re-
cover a cause for the bad taste of the pizza. Although such a cause might
in fact exist, say decayed ingredients, under the most plausible interpre-
tation this sentence simply expresses that the pizza is overpriced for its
taste. 6 Although the symmetrical nature of Di Meola's analysis is very
appealing, examples like (19) make one wonder if it is at all the case
that concessive constructions encode a "hidden cause".
To avoid all these problems, Di Meola's analysis should perhaps be
restricted to what Sweetser calls linking at the content level (Azar's
1997 "indirect-rejection concessivity"). Such a move might exclude (18)
as potential counterexamples. But even for this subset of concessive
constructions, there are fundamental problems: the second cause (B 1) is
never encoded, nor does it seem to play any role as background assump-
tion for conversational inference, i. e. a sentence like (19b) can never
give rise to a conversational inference for which propositions like 'the
pizza is cheap/old ... ' are used as a premise. Moreover, its basic seman-
tic outlines are never contextually given. Above all, Di Meola has noth-
ing to say on any of the phenomena discussed in the preceding section.
Finally, there is also the problematic view, which has not been men-
tioned so far, that concessive constructions are somehow non-
cooperative, involving as they do "hidden causality". If the Gricean
maxims of cooperative conversation are an essential basis for any com-
350 Ekkehard Konig and Peter Siemund
(21) a. It is NOT the case that Brutus's love for Caesar was a reason
for not killing him.
b. Brutus killed Caesar, although he LOVED him.
(23) a. It is not the case that the road is getting wet although it is rain-
ing.
b. The road is getting wet because it is raining.
Causal and concessive clauses 351
Pasch (1992a, 1992b) rejects this analysis for several reasons. The
first reason is that the truth conditions for concessive constructions and
negated causal constructions do not come out identically, according to
her analysis. But this argument depends very much on the details of the
analysis. The perspective of Konig (1991) is different: our intuitions
about an opposition between causality and concessivity, which mani-
fests itself in the phenomena discussed in section 2, should be taken as a
basic problem for their semantic analysis. In other words, the ability to
explicate this relationship should be taken as a criterion of adequacy for
such analyses. Pasch's second reason is more to the point and impossi-
ble to refute. The relationship between 'because p, q' and 'although p,
q' is not a clear instance of duality, since one does not get the expected
paraphrases, except for the case where a wide-scope negation of a
causal construction clearly corresponds to a concessive construction
with the negative counterpart of the main clause. All the pairs discussed
in section 2 (e.g. (5), (9» are of this kind. But, there is simply no way
one can get from a negated concessive causal construction to a causal
one. The relevant paraphrase relations seem to hold only in one direc-
tion. Moreover, Pasch is right in pointing out that the internal negation
of a causal construction «because p) --, q) does not admit of a conces-
sive paraphrase «although -,p) q):
(24) a. Since it is so hot, there are not many people in the streets.
b. Although it is not hot, there are many people in the streets. =I:-
(24a)
Worse still, causal constructions that make sense lose this quality as
soon as the main clause is negated and therefore have no paraphrase in
terms of concessivity:
4. Towards an explication
(26) a. because p, q
b. although p, q
linguistic observations. At the end of her paper, however, she does have
some suggestions for the analysis of equivalent pairs like (Sa) and (Sb),
which we repeat here for convenience:
(28) NEG ([this house dispenses with air conditioning] CAUSE [it is less
comfortable])
Since the negation of a causal construction does not affect its pre-
suppositions, these go through unchanged (cf line (ii)). The negation
relating to the whole causal construction in line (i) can therefore only
relate to the assertive part of its meaning, i.e. the conjunction p & q (cf
line (iii)). Since we also assume that the causal clause represents pre-
supposed material, the negation can only affect the main clause, as is
Causal and concessive clauses 355
and (b) going on to make one's own (contrasting) point (cf Couper-
Kuhlen and Thompson 1998: 29).
Let us now take a closer, critical, look at such a discourse-oriented
approach. Couper-Kuhlen and Thompson (1998: 30), for example, char-
acterise the distinctive features of the concessive relation in interaction
by formulating the following Cardinal Concessive Schema (where X
and Yare understood to be potentially incompatible):
6. Conclusion
Our discussion was meant to show that we can only explain some inter-
esting phenomena in the formal coding of concessive relations across
languages if we succeed in providing a clear explication of the contra-
dictory relationship intuitively felt to oppose causal to concessive con-
structions. A few attempts at explication were presented and critically
examined. The basic outlines of such an explication, which we claimed
to be an indispensable criterion of adequacy for any analysis of conces-
sivity, seem to be fairly clear. It was also shown that there is no need to
develop different analyses for concessive constructions (i.e. complex
sentences with concessive adverbials) and interactive concessive
schema or adversative sentences. The analysis developed above also
subsumes the relevant concessive schema and the traditional (non-
interactive) "adversative" or "contrastive" sentences. What we need to
distinguish are, of course, different concessive sentences such as "di-
rect-rejection concessive" vs. "indirect-rejection (argumentative) con-
cessive" (cf Azar 1997) or concessive vs. adversative sentences. Unfor-
tunately, there is no unambiguous, generally accepted terminology in
this domain. What the discourse-oriented studies have shown is that the
interactive/non-interactive distinction needs to be made for all argumen-
tative or "indirect-rejection" concessives. Perhaps the term "adversa-
tive" should be restricted to all argumentative concessives (p but q) that
lack the interactive component, i. e. to such cases where two arguments
358 Ekkehard Konig and Peter Siemund
Notes
1. Note that we are not assuming that there will be many cases where the same
connective is used both in a concessive and in a causal interpretation. As is
pointed out in a typological study by Kortmann (1997: 202) this is very rarely
the case. What we do expect on the basis of the analysis developed below is
that a concessive connective will often be a negated counterpart of a causal
one. Such cases do indeed exist in a wide variety of languages: in view of, for
instance, may be regarded as a causal preposition and both regardless and
irrespective of can plausibly be analysed as the negated counterparts. On the
other hand, the relatedness between causal connectives and concessive ones
may only manifest itself in the use of the indicative mood (as in German) as
opposed to the subjunctive mood, which is used in concessive conditionals
and concessive clauses in French. Note also that negated correlative markers,
which are closely related to causal expressions, are frequently used as conces-
sive connectives (cf. English nevertheless, nonetheless).
2. A major part of the following discussion is based on Konig (1991).
3. Such cooccurrences of causal and concessive connectives seem to be only
possible in cases where we have epistemic or argumentative linking in the
sense of Sweetser (1990), rather than linking at the content level. A combina-
tion of causal and concessive adverbs always amounts to rejecting a conse-
quence or conclusion (held by an interlocutor) based on the preceding sen-
tence as a premise. Examples like the following therefore have an interactive
character, i.e. if both sentences are uttered by the same speaker, they suggest
that the second is a rejection of an opinion held by the hearer:
i. Karl hat mir sehr weh getan. Deswegen ist er aber trotzdem kein
schlechter Mensch.
'Karl really hurt me. But that does not mean he is a bad guy. '
ii. Ich habe zwei Wochen lang fast nichts gegessen. Deswegen hahe ich
trotzdem nicht abgenommen.
'I have not eaten anything for almost two weeks. 1 have not lost any
weight for all that.'
4. 'Masdar' is the term traditionally used for nominalisations of verbs or verb
phrases in Caucasian languages. Like the gerund in English, masdars have the
external syntax of noun phrases, but the internal syntax of verb phrases.
5. -matta is a derivational affix with a privative meaning: V-matta roughly
means 'without Ving'.
6. Note that it would not be possible to assume Pizza is cheap as Bl because this
simply contradicts AI. What is necessary is a stronger cause.
7. German zwar, like English true, derive from expressions encoding the notion
of 'truth' and the second components of obwohl and obschon are particles that
may be used for focusing positive polarity in declaratives ('verum focus'):
Karl hat das WOHL/SCHON zu verantworten.
Causal and concessive clauses 359
8. The difference between Table 2 and Table 3 is that the subordinated clause is
analysed as representing presupposed information and that the implicational
presupposition is formulated as a connection between general tendencies (P,
Q) rather than specific instantiations (p, q).
References
Azar, Moshe
1997 Concession relations as argumentation. Text 17: 3.
Bybee, Joan et al.
1994 The Evolution of Grammar. Chicago: The University of Chicago
Press.
Couper-Kuhlen, Elizabeth and Sandra A. Thompson
1998 On the concessive relation in conversational English. In: Fritz-
Wilhelm Neumann and Sabine Schtilting (eds.), Anglistentag 1998
Erfurt. Proceedings, 29-39. Trier: Wissenschaftlicher Verlag.
Di Meola, Claudio
1997 Der Ausdruck der Konzessivittit in der deutschen Gegenwartsspra-
che. Theorie und Beschreibung anhand eines Vergleichs mit dem
ltalienischen. (= LA 372.) Ttibingen: Niemeyer.
1998 Zur Definition·einer logisch-semantischen Kategorie: Konzessivitat
als 'versteckte Kausalitat'. Linguistische Berichte 175: 329-349.
Dowty, David R.
1979 Word Meaning and Montague Grammar. Dordrecht: Reidel.
Haspelmath, Martin
1993 A Grammar ofLezgian. Berlin: Mouton de Gruyter.
Hermodsson, Lars
1973 Inkonditionalsatze. Zur Semantik der sogenannten Konzessivsatze.
Studia Neophilologica 45: 298-305.
1994 Der Begriff "konzessiv". Terminologie und Analysen. Studia Ne-
ophilologica 66: 59-75.
Konig, Ekkehard
1988 Concessive connectives and concessive sentences: cross-linguistic
regularities and pragmatic principles. In: John Hawkins (ed.),
Explaining Language Universals, 145-166. Oxford: Blackwell.
1991 Concessive relations as the dual of causal relations. In: Dietmar
ZaefIerer (ed.), Semantic Universals and Universal Semantics, 190-
209. Dordrecht: Foris.
1994 Concessive clauses. In: R. E. Asher (ed.) The Encyclopaedia ofLan-
guage and Linguistics. Oxford: Pergamon Press.
Kortmann, Bernd
1997 Adverbial Subordination. A Typology and History ofAdverbial Sub-
ordinators Based on European Languages. (= Empirical Approaches
to Language Typology 18.) Berlin: Mouton de Gruyter.
360 Ekkehard Konig and Peter Siemund
Lobner, Sebastian
1990 Wahr neben Falsch. Duale Operatoren als die Quantoren natilrli-
cher Sprache. (= Linguistische Arbeiten 244.) Tubingen: Niemeyer.
Pasch, Renate
1992a Sind kausale und konzessive Konstruktionen Duale voneinander? (=
Arbeiten des Sonderforschungsbereichs 282 - Theorie des Lexikons,
Nr. 31.) Dusseldorf: Heinrich-Heine-Universitat.
1992b Kausale, konzessive und adversative Konnektive: Konnektive als
Mittel des Ausdrucks von Diskursprasuppositionen. Milnstersches
Logbuch zur Linguistik 1: 33 - 48.
Sweetser, Eve
1990 From Etymology to Pragmatics. Cambridge: Cambridge University
Press.
Thompson, Sandra A.
1987 'Concessive' as a discourse relation in expository written English.
In: Brian D. Joseph and Arnold M. Zwicky (eds.), A Festschrift for
lise Lehiste, 64-73. (Working Papers in Linguistics 35.) Columbus,
Ohio: Ohio State University.
Thompson, Sandra A. and William C. Mann
1987 A discourse view of concession in written English. In: Scott De-
Lancey and Russell Tomlin (eds.), Proceedings of the Second An-
nual Pacific Linguistics Conference, 435-447. Eugene, Oregon: Uni-
versity of Oregon.
Concession implies causality, though in some
other space
Arie Verhagen
This paper addresses the issue of the relationship between concessivity and
causality, in the sense that these concepts can and should be construed as
parallel, concessivity being in some sense a negative counterpart of
causality. It is argued that in order to avoid the risk of internal contradiction
and other conceptual problems, it is necessary to construe the relationship in
terms of partly but not completely similar 'points of view', or mental
spaces. While aiming for an explanation of obvious facts of interpretation
and usage, the discussion focuses on clarification of conceptual questions.
1. Introduction
In (3), the causal relation holds between the facts that John passed his
exams and that he worked hard (i.e. in the "content domain"). In (4), on
the other hand, it holds in the conceptualizer's mind, i.e. the proposition
that John worked hard is epistemically caused by an argument based on
the knowledge that he passed his exams. Postulating a systematic
conceptual distinction between content and epistemic domains for the
application of the meaning of modal verbs as well as connectives
provides a unified account for such phenomena, and thus has great
explanatory power.
At first sight, Sweetser's approach is further corroborated by the fact
that it also applies to connectives with some negative aspect of meaning,
i.e. concessives. However, precisely this application also produces a
conceptual dilemma, as we shall see.
On the one hand, cases like (5) and (6) seem to be clear negative
parallels to (3) and (4):
(5) John did not pass his exams although he worked hard.
(6) John did not work hard, although he passed his exams.
Intuitively, one could say that in (5) the real-world causal connection
between passing exams and working hard is in some sense negated, so
that this would be a case of concession in the content domain.
Analogously, what is negated in (6) seems to be the validity (in this
case) of the inference from the knowledge of John's passing his exams
364 Arie Verhagen
(7) John worked hard, but he did not pass his exams.
for the parallel between this pair on the one hand, and (3) and (4) on the
other.
My purpose in the remainder of this paper is to argue that the effects
previously analyzed in terms of domains can and should be derived
from the construction of mental spaces, with (especially) negative
semantic elements taken as instructions for setting up alternative mental
spaces, and that precisely this reconstruction avoids the dilemma just
noted. More exactly, an approach using mental spaces, as dynamically
constructed domains for the interpretation of linguistic elements, allows
us to simultaneously (a) account for the parallel between (3)/(4) and
(5)/(6), (b) claim that contrastive-as opposed to causal-relations
necessarily involve reference to some reasoning process, and (c)
account for the connection between concessivity and (negated)
causality.
3. Elements of a solution
Let me begin by illustrating the idea with the relatively simple example
(5), repeated here for convenience:
(5) John did not pass his exams although he worked hard.
However, understanding the concessivity of John did not pass his exams
although he worked hard involves understanding that such a mental
space is to be set up as a second space projected from the speaker's
actual mental space. By uttering although p the speaker instructs the
hearer to set up two similar but distinct mental spaces in which p is true;
in one (Spacel, with the same structure as in Figure 1, but now projected
from the speaker's own mental space) the causal inference "p therefore
Causality in some other space 367
q" is valid, while in the speaker's mental space (Spaceo), the truth of q is
denied so that this causal inference is not valid. Represented
graphically:
Spaceo
the topos is explicitly excluded from the speaker's own space. In any
case, the topos is neither necessarily entailed nor contradicted within
Spaceo, something that is very hard to capture without distinguishing
between the two spaces.
the latter two notions will tum out to be relevant for the discussion
below, but I think it is an immediate consequence of other features of
the analysis and does not require independent stipulation.
Third, Ducrot's concept of polyphony is intimately linked to his
theory of argumentativity in natural language. At this point I think there
is both a real difference between the mental space and the polyphony
approaches, and a point where Mental Space theory can profitably be
augmented, if we take 'argumentation' not so much in the sense of
verbal conflict, but more in the sense of 'trying to influence a person's
point of view', 'orienting one's addressee(s) towards some conclusion'.
I think this notion of argumentative 'orientation' (in my view more
important for linguistic analysis than the notion of argumentation in
itself) is very useful, and can help us to see how mental spaces work in
certain areas (such as scalar semantics, or certain phenomena of
language change). Still, for the purposes of the analysis of discourse as
such, the mental space terminology seems to me to involve the
minimum of necessary assumptions concerning the nature of discourse
interpretation. Analysts will agree, I assume, that any approach to
discourse interpretation has to take into account that people can manage
multiple viewpoints simultaneously, whereas the question if these are
'better' viewed as individual psychological phenomena or as dialogic,
social ones, requires more specific assumptions and evidence. This is
the reason why I prefer to use the mental space terminology here. But
let us return to the main theme now.
Content causality, e.g. stating John passed his exams and then going on
to say how this came about (. .. because he worked hard) allows for the
first clause to be taken as representing established knowledge, and the
second as the only news. But when the second segment is an argument
as in (4), then its being uttered would be beside the point, violating the
maxim of relevance, if the first clause was already taken to represent
undisputed knowledge at its being uttered: one does not provide support
for information that one knows will remain unchallenged.
Consequently, if a speaker provides an argument in the second clause
(i.e. produces a relation of epistemic causality), she projects a mental
Causality in some other space 371
space with a non-positive epistemic stance (unlike her own) towards the
proposition expressed in the first clause:
Space! Spaceo
By the same reasoning, the content reading of the causal relation in (3)
does not involve the construction of another mental space representing
another epistemic stance:
As has been observed frequently, the content reading provides only one
statement, e.g. an answer to a single question of the type "Why?" or
"How?", and thus in a sense presupposes the proposition that John
passed his exams; this is what I called the "undisputed information"
interpretation of the first clause above. Thus in a content reading of p
because q the speaker precisely does not project a mental space in
which the validity of p is not (completely) certain. 6 We can now
formulate what it is that makes epistemic causality parallel to contrast
(including, in particular, concessivity): the construction of two mental
spaces with distinct epistemic stances towards a proposition.
(6) John did not work hard, although he passed his exams.
Space. Spaceo
although
That is, in uttering (6): John did not work hard ["not p" in Spaceo],
although he passed his exams ["q" in Spaceo], in particular through the
use of although, the conceptualizer projects another mental space
(Spacel), containing the abductive inference that John must have been
working hard ["p" in Spacel] because he passed his exams ["q" in
Spacel], and then invalidates this inference by denying its conclusion.
This analysis is corroborated by the fact that (6) allows for the following
explication: "Actually, the inference that John must have been working
hard considering that he passed his exams, is invalid: it is both true that
he passed [=second clause], and that he did not work hard [=first
clause]". The applicability of such a paraphrase is predicted on the basis
of the present analysis: since the mental space projected immediately
from the conceptualizer's space contains an instance of epistemic
causality, this space projects yet another mental space as its background,
given our analysis of epistemic causality in the previous section.
Causality in some other space 373
I now want to show how the mental space approach also provides an
analysis of the connection between concessivity and negated causality,
i. e. the (partial) overlap between the wide-scope interpretation of (1)
and the narrow-scope interpretation of (2), repeated here for
convenIence:
Spaceo
It is now evident that the concessive interpretation of (1) entails that the
element because marks a causal relationship in another mental space
than the conceptualizer's. The relations between the conceptual
structure of Figure 5 and linguistic material in (1) are as indicated by
means of italicization in Figure 6:
Spaceo
Space! Spaceo
(10) ?11t is not the case that John failed his exams although he worked
hard.
(11) Beatie finds that although she has lost a lover she has gained her
freedom.
(12) They could not believe that although John had worked hard, he
failed his exams.
What seems to make this example so much better than (10) is the
factivity of the matrix predicate in (12); when I say X could not believe
that p, I am still committed to the truth of p myself, unlike when I say It
is not the case that p. Thus, in both (11), a non-negative sentence, and
(12), negative but factive, the embedded propositions are not
contradicted in the primary conceptualizer's mental space: in Spaceo,
there simply is no denial of "p although q". The problem with (10) on
this view is that it contains two instructions (not the case and although)
to set up mental spaces as alternatives to the conceptualizer's, and that
these cannot be properly related to each other.
It is interesting to note that with a non-factive matrix predicate the
interpretation of the entire sentence becomes unclear:
(13) ?They did not conclude/think that although John had worked
hard, he failed his exams.
When I say X did not think that p, my utterance may well be taken as an
indication that I am at least uncertain about p myself, unlike the
situation with factive predicates. Consequently, if the complement
clause contains a concessive one, this would be in the scope of the
'uncertainty operator' and the same problem of distinct spaces with a
non-positive epistemic stance would arise as in (10). Whatever the
details of the analysis of these cases, they help to show clearly that the
problem with the interpretation of a negation having a concessive clause
in its scope is a problem of manipulating points of view: when there are
several of them, it may be hard or even impossible to relate them to each
other in a consistent manner.
378 Arie Verhagen
6. Conclusion
Notes
1. There is quite a variety of labels for this concept, such as topos, (discourse)
presupposition, conventional implicature, defeasible implication. For some
analysts these are probably not the same, but the differences do not have to
concern us here. The term for this kind of "inference licensing rules" that I
will use myself later in this paper is "topos". It stems from Anscombre &
Ducrot's theory of argumentation in natural language, e.g. (1983, 1989). See
also Ducrot (1984, 1996).
2. Thus I am in full agreement here with Konig and Siemund's emendation of
Pasch's analysis (cf. their section 4). What they fail to note, however, is that it
is not only a matter of descriptive adequacy that the 'background assumption'
involves some kind of generalization, but actually one of conceptual
necessity. As observed with respect to (8) , this schema (as well as the right
hand column of Table 2 in Konig and Siemund (this volume» leads to a
straightforward contradiction, as it entails both q and -'q. In order to avoid
this, only the propositions p and q themselves should be allowed to represent
particulars ("singular statements" in terms of Popper (1972», thus allowing
for verification (cf. Popper 1972: 27, 33, and elsewhere); the background
Causality in some other space 379
assumption must be some general rule whose logical form does not allow for
verification. For similar reasons, the background assumption must be 'only'
general and not universal~ if it were truly universal, e.g. with q in the
presupposition taken as "Vx(j{x»", the use of a concessive, asserting p&q,
would again entail a contradiction (p and "'P). These points have been
elaborated in other contexts before (cf., for example, Lagerwerf (1998) and
some references cited there), and especially in work in the tradition of
Anscombre and Oucrot's theory of argumentation in language. As Oucrot
(1996: 144) suggests, it might in fact be much older: "When we say ['It's
warm, let's go for a walk'], we do admit that there might be exceptions but
that does not prevent the topos from being valid, which is the point this highly
famous formula attributed to Aristotle makes: 'exceptions make it possible to
uphold the rule in unforeseen cases' .".
3. Precise terminology may be important here. Notice that what is to be avoided
is the attribution of contradictory beliefs. It is, of course, no problem to
entertain contradictory propositions. In fact, Mental Spaces is an explication
of what is involved in doing just that.
4. It should perhaps be pointed out that this description is not a full semantic
analysis of a linguistic element such as although (though it does provide
essential conceptual structure that will have to enter into such an analysis).
For example, this description does not yet provide a basis for distinguishing
between although and but, whereas the functions of these words are not
identical.
5. Abductive reasoning is a sufficient but not a necessary condition for the
construction of multiple spaces: a deductive inference as in may also be
construed as an argument to take away doubt. In domain terms, and
informally: content causality may always be 'used as' epistemic causality, but
the reverse does not hold.
6. This is a point, by the way, where I disagree with a specific polyphonic
analysis. It is sometimes proposed that such a distribution of old-new-
information should be analyzed in terms of two voices, one asking the
question ("Why?" or "How?"), the other providing the answer. In my view,
such a move threatens to destroy the usefulness of the concept, since it would
actually make it quite hard again to represent the difference between content
and epistemic causality.
7. In these and the following mental space representations of concessives, I will
leave out the connectors between the topos-representations (one obligatory
and one optional) in the different spaces, in order to make the pictures
somewhat simpler to read.
References
Ducrot, Oswald
1984 Le dire et Ie dit. Paris: Editions Minuit.
1996 Slovenian Lectures/Conferences Slovenes. Argumentative Semantics/
Semantique argumentative. EditortEditeur Igor Z. Zagar. Ljubljana:
ISH Institut za humanisticne studije Ljubljana.
Fauconnier, Gilles
1994 Mental Spaces. Aspects of Meaning Construction in Natural
Language. Cambridge: Cambridge University Press.
1998 Mental spaces, language modalities, and conceptual integration. In:
Michael Tomasello (ed.), The New Psychology of Language.
Cognitive and Functional Approaches to Language Structure, 251-
279. LondonIMahwah, NJ: Lawrence Erlbaum Ass..
Fauconnier, Gilles, and Eve Sweetser (eds.)
1996 Spaces, Worlds, and Grammar. Chicago/London: The University of
Chicago Press.
Konig, Ekkehard
1991 Concessive relations as the dual of causal relations. In: Dietmar
ZaefIerer (ed.), Semantic Universals and Universal Semantics, 190-
209. Dordrecht: Foris.
1994 Concessive clauses. In: R.E. Asher (ed.), The Encyclopedia of
Language and Linguistics, 679-681. Oxford: Pergamon Press.
Lagerwerf, Luuk
1998 Causal Connectives have Presuppositions. Effects on Discourse
Structure and Coherence. Den Haag: Holland Academic Graphics.
Pasch, Renate
1992a Sind kausale und konzessive Konstruktionen Duale voneinander?
[=Arbeiten des Sonderforschungsbereich 282-Theorie des Lexikons,
Nr. 31]. Dusseldorf: Heinrich-Heine-Universitiit.
1992b Kausale, konzessive und adversative Konnektive: Konnektive als
Mittel des Ausdrucks von Diskursprasuppositionen. Munstersches
Logbuch zur Linguistik 1, 33-48.
1994 Konzessivittit von wenn-Konstruktionen. Tubingen: Narr.
Popper, Karl R.
1972 The Logic ofScientific Discovery. London: Hutchinson & Co. [First
English edition 1959]
Sweetser, Eve E.
1990 From Etymology to Pragmatics. Metaphorical and Cultural Aspects
ofSemantic Structure. Cambridge: Cambridge University Press.
Concessive patterns in conversation *
1. Introduction
(1) Joanne and Lenore are discussing a mutual friend who has
engaged in substance abuse.
expresses the fact that one gives a favorable receipt to one's opponent's real
or presumed arguments. By restricting his claims, by giving up certain
theses or arguments, a speaker can strengthen his position and make it
easier to defend, while at the same time he exhibits his sense of fair play
and his objectivity (488).
We will see that the way English conversationalists use concession does
in fact exhibit these very characteristics. More recently,
Antaki/Wetherell (1999) identify what they call 'show concession', a
pattern not identical but related to the ones described here, in which
speakers address some vulnerability in their starting claim and, by
making a show of conceding it, achieve the rhetorical effect of
fortifying their own position.
Building on these studies, we became fascinated with the ways in
which mundane conceding is accomplished by conversationalists. We
incorporate the traditional view of concession as involving a conflict of
expectations with our idea that the third move in a conceding format
involves a potentially contrasting point of view. We incorporate the
rhetorical perspective in our view of conceding as being situated (often)
in a larger context of argumentation.
In this paper we would like to explore how and why a speaker
concedes another's point in everyday interactions. The questions we
address include:
384 Elizabeth Couper-Kuhlen and Sandra A. Thompson
2. Database
The data for this study were taken from a variety of British and
American sources, including conversations from the Corpus of Spoken
American English (DC Santa Barbara)3 as well as our own collection of
conversations and radio call-in show data. We identify the source of
each of our examples below.
In carrying out this piece of research, we studied our transcripts for
all cases in which we felt, as native speakers, that conceding was being
done. For us, as suggested above, 'conceding' involves, first,
acknowledging another's point (though not necessarily 'giving in' or
'yielding' 4), and second, going on to make one's own «potentially)
contrasting) point. We also included in our collection instances in which
one or the other of these parts was missing but strongly projected. We
collected a database of approximately 200 examples and examined them
for recurrent interactional and grammatical features. In this paper we
will report on the most robust patterns.
X' and Y-are realized verbally, and are maximally distinctive with
respect to other possible hypotactic clause combinations. The pattern
identified in example (1) just above is an example of a Cardinal
Concessive and could be represented provisionally as:
A: X
B: X'
y
(where X and Yare understood by participants as potentially
contrasting)
(2) Charles and Steve, currently in NYC, are lamenting the advent of
Starbuck's coffeehouses, which Steve says will 'blowout' the
small privately owned places in cities like Minneapolis.
Charles claims in lines 1-2 that even. the privately owned coffeehouses
'here', i.e. in New York City, are so characterless. When Steve does not
immediately second this assessment, Charles downgrades it somewhat
by appending by and large (line 3). In the next tum, a conceding tum,
Steve acknowledges that Charles' assessment is partially true: he agrees
that privately owned places outside Greenwich Village are indeed
characterless (line 4). But he goes on to maintain that in another sense
Charles' claim is wrong: there are privately owned places in the Village
which are nice (line 6).
A prominent feature, then, of this example is the partial nature of the
conceding move, labeled X', at line 4. In other words, in his conceding
move, Steve sets up a contrast between two aspects of a single
proposition, which we will call 'partitioning'. Thus he acknowledges
that Charles is right about places being characterless, but only about the
places outside of the Village. But there's some in the Village that are
nice. In fact, one of our most striking findings was that speakers very
often accomplish conceding by acknowledging only part of what the
other speaker has said. In other words, one of the recurrent ways in
which a speaker concedes another's point is to acknowledge that part of
it is, or may be, true.
Again, we can characterize this interaction in terms of the Cardinal
Concessive schema as follows:
Concessive patterns in conversation 387
(3) Bee and Ava (on the phone) have been joking about each other's
knowledge of where the other is.
Aside from its interactional reality, the Cardinal Concessive schema has
two further advantages as an analytic tool: (i) it provides for maximal
flexibility with respect to the lexico-syntactic representation of its parts,
and (ii) it underspecifies the linkage between X' and Y. We expand on
each of these features below.
The Cardinal Concessive schema allows for its constituent parts X, X'
and Y to be realized by linguistic units of varying size: words, phrases,
clauses or whole chunks of discourse-all possibilities which are in fact
attested in our database. In examples (1)-(3), all three parts of the
Cardinal Concessive schema were phrases or clauses. But in the
following case, X' is realized by a single word:
In this excerpt Marie expresses the point that at least some of the
participants in the demonstration simply wanted to make their opinion
known that the bombing had started too early, before efforts at
negotiation had been exhausted, and that they should not be disparaged
because of this. In line 8 she appends a reference to an earlier remark by
the moderator that freedom of expression, as long as it is peaceful, is a
fundamental American right. Leo, the moderator, initially agrees with
this (yeah) (line 10), but then proceeds to evaluate what happened as a
shame (line 11), an assessment which is potentially incompatible with
the positive light in which the basic right to freedom of expression has
been presented. Yet his use of the Cardinal Concessive schema allows
him to come across as claiming that both points hold.
In this example the X' move has minimal size, being realized by the
single word yeah. 5 However it can be more substantial, as we saw in (1),
where Joanne concedes by saying as much as he's abused his liver. In
the following case the conceding move is considerably more substantial:
(5) On this call to a British radio phone-in program, Mrs. Etchalls and
the moderator Dick Hatch are comparing the advantages and
disadvantages of a penal system in which people who are
convicted of felonies wear electronic bleepers rather than go to
prIson.
in our corpus. For instance, the relation between X' and Y may not be
marked by any morphological markers at all, as was the case in example
(2) above. Here Steve merely juxtaposes the two claims outside of the
Village they are (so characterless) and there's some in the Village that
are nice. Yet X' and Y are linked by a parallelism between the two
prepositional phrases and contrastive stress on in in the second one.
As was illustrated in (4), speakers may link Y to X' with the
coordinating conjunction but. And, as the following example
demonstrates, speakers may also have recourse to subordinating
conjunctions in constructing Concessive patterns:
point: you have to recognize ... that there are a great many people ..
whose lives are built round ... their pets (lines 13-16). The conclusion to
be drawn from this remark is presumably that owners are justified in
being concerned about the well-being of their pets. Notice that Dick
Hatch's counter-point Y is linked to his prior X' by the subordinator
except that (line 12).
Thus a variety of means of explicitly linking X' and Y-as well as
the absence of any explicit link-are attested in our corpus. 6 The
Cardinal Concessive schema does justice to this fact by underspecifying
how its constitutive parts are conjoined.
In this section we have tried to show that approaching our data in
terms of a Cardinal Concessive schema has several advantages,
allowing us to show what all of our examples have in common in terms
of what the interactants are doing and how they understand what they
are doing. We have seen that acknowledging a previous speaker's point
to have been (partially) right and then going on to make one's own
contrasting point is a recurrent format with which speakers counter each
others' claims. In the next section, we will examine the two most
frequent variations on the Cardinal Concessive schema.
(7) Prior to this excerpt Charles has talked about a French friend
Didier, who he describes as 'a musician and kind of cabaret
artist' . The conversation then shifts to an occasion on which
Charles heard Didier play some of his own songs and to another
occasion on which Karen, a present co-participant, heard Didier's
manager Ray sing. The topic of talk now reverts to Didier.
(8) Two friends Ann and Barb are talking here about a mutual friend
of theirs who has become seriously ill and given up on life. Barb
has just asked rhetorically "What about the people who care about
her and the people who are around?"
9 and CASS:,
and people ALSO,
11 who've never been CLOSE friends of hers,
12 (0.3) but who'd BEND over backwards (0.2)
for this woman,
When Ann claims that their friend doesn't have any people around her
anymore (line 1), Barb contradicts this with a Y in her next tum: but she
does (line 2). Next Barb reasserts her counter-claim with she does still
have a lot ofpeople who care (line 3). This formulation re-specifies the
predicate in question as having people around her who care. Yet Barb
goes on to acknowledge, with an X', that in one sense Ann's point
holds: she doesn't have anybody she 'II let get close to her (line 4). This
acknowledgement, like the one in example (7), serves to weaken the
force of the prior contradictory move by suggesting that in one way the
interlocutor's statement was right. Yet in contrast to (7), the conceded
part, X', does not remain in end position: Barb now reiterates her initial
counterclaim (Y): but she's got a lot ofpeople who care (line 5). This in
effect expands the sequence by turning X' into a conceding move
which-as in the Cardinal Concessive-is followed up by a counter-
claim Y. Moreover, Barb now proceeds to construct a list of people who
care, thereby underlining the validity of her counterclaim.
What Barb does here can be thought of as first disagreeing, and then
retracting somewhat by weakly (partially) agreeing. In fact, in this
variation on the Cardinal Concessive, where speaker B's conceding
move X' follows the contrasting claim Y, we often have the vernacular
sense that speaker B is 'backing down'.
A second way in which the sequential implicativeness of an X' in
end position can be avoided is to downplay it lexically and prosodically.
This is frequently done by using the unstressed particle though to
represent X'. For instance:
(9) In this excerpt Laura has been talking about her frequent
backaches. Ruth has reported her doctor's claim that everybody
gets backaches because the backbone is 'built for an animal to
walk on all fours' .
1 Mom: ACTUALLY,
2 X that HELPS,
3 X (-) when you have a backache.
4 Laura: really?
5 X Mom: (--) is to -
6 X (-) to get down on your hands and KNEES,
396 Elizabeth Couper-Kuhlen and Sandra A. Thompson
(10) Harold, Miles, Pete and Jamie are gossiping about a neighbor who
has recently produced a number of 'brat' kids.
1 X Harold: does she even have a b- a MAN?
2 X I guess she MUST.
3 Miles: (0.5) does she have a WHAT?
4 Jamie: [a MA.N.
5 Harold: [a MA.N.
6 Pete: ( .hh)
7 x' Jamie: she has SOME [kind of a-
8 x' Miles: [at least TEMPORARILY,
9 Pete: [yeah,
10 Harold: [yeah.
11 Jamie: heh heh (.h) [at ONE time
12 x' Harold: [for about FIVE minutes,
13 x' PROBABLY.
14 Miles: heh
15 Pete: heh
At the beginning of this excerpt Harold asks a question: Does she even
have a man? (line 1), which he then proceeds to answer himself: I guess
she must (line 2). This establishes the point 'She must have a man',
which we are calling X. Following a short side sequence for repair (lines
3-5), Jamie and Miles now each produce, in partial overlap, a conceding
move with respect to this point. Jamie's she has some kind ofa- (line 7)
has strong stress on some, Miles' at least temporarily (line 8) has strong
stress on temporarily.
What we see here, then, is Jamie and Miles each partitioning
Harold's claim X in a different way. In Jamie's she has some kind of a-
(line 7), the stress on some sets up a contrast between 'some kind of
man' vs. 'the real kind of man'. In Miles' at least temporarily (line 8),
398 Elizabeth Couper-Kuhlen and Sandra A. Thompson
6. Projectability of Y
As we have noted, there are two mechanisms which provide for the
projectability of Y and render its actual production in interaction
negotiable, both of which are in evidence in (10). One is semantic and
the other prosodic; we will briefly examine each in this section.
X, for instance with respect to size or amount (little vs. a lot), type (fake
vs. real), time (now vs. then), location (inside vs. outside), frequency
(sometimes vs. often), chance of occurrence (likely vs. unlikely), out-
ward appearance (looking or seeming vs. really being), etc. In (10), for
example, as we have shown, two partitions are in evidence, one
concerning the type of man the neighbor has, the other concerning the
period of time during which she has had a man. The concessive move
involves confirming one aspect of the partition, e.g. one type of man or
length of time, but disconfirming the other. Since the contrast estab-
lished is binary, very often evoking one part naturally calls to mind the
other. Moreover, the affirmation of one part implies the denial of the
other (and vice versa). Thus the implicational relations in (10) could be
more realistically captured if the X' moves were represented as [X' +Y-
implied]:
In these cases, then, one type of man the neighbor has or one time
period of her having a man is affirmed; the other type and time period
are by implication disconfirmed.
Our analysis of the way partitioning allows Y to be projected but not
made explicit also gives us a means to schematically enrich our analysis
of the conceding that Steve does in example (2), repeated in short form
here:
What we see here, then, is that partitioning often projects what the
speaker's next move, the (potentially contrasting) counter-claim Y, will
be. Sometimes Y is then offered explicitly, as we see in Steve's
comparison between outside of the Village and in the Village in example
(2). But sometimes Y is left inexplicit, often with humorous or satirical
overtones, as we saw in Jamie's she has some kind of a- and Miles' at
least temporarily in example (10) about the neighbor's relationships.
Example (10) thus illustrates the way in which this partitioning strategy,
precisely because it is so frequently employed in doing conceding, can
be used to project a counter-claim Y which then does not need to be
made explicit.
Concessive patterns in conversation 401
1 Wendy:
[but don't make FUN of me. ]
2 X Marci:
[you look pretty PROUD of your]self
Wendy,
3 X'+Y-impl Wendy: WE:LL,
4 X'+Y-impl I'm a LITTLE proud, =
5 Y-hypoth =except if you [think it's STUPID, heh
6 Kevin: [ha
402 Elizabeth Couper-Kuhlen and Sandra A. Thompson
In this excerpt Wendy repeats her playful plea don't make fun of me
(line 1), while Marci suggests in overlap that Wendy actually looks
rather proud of her clue (line 2). In response, Wendy acknowledges that
Marci's assessment is partly true-she does feel a little proud-but she
goes on to claim that should the others find her clue stupid, then
something potentially incompatible will hold, namely she will feel bad.
(As indicated, Y is expressed hypothetically here.) Partitioning is
signaled in this instance not only by contrastive accent placement, with
the main accent on little rather than proud, but also by the fall-rise tone,
which begins with a pitch step-up on lit- and is spread over the rest of
the tail of the intonation phrase. The falling movement on little is quite
marked, while the movement on proud is only minimally rising.
The projecting prosodic devices discussed in this section are not used
consistently enough in our materials to make any strong claims about
,concessive intonation'. We have many cases of concessive patterns
which lack this kind of marking. Yet when contrastive accent placement
and/or fall-rise tone are found,. they often cue partitioning and are thus
hints that Concession may be involved. Furthermore, when we find an
example in which Y is missing, we always find either or both projecting
prosodic devices and partitioning.
7. Functional considerations
(12) Connie, Vera and her boyfriend Keith are talking about Vera's
father, who is suffering from the aftermath of a knee operation. In
prior conversation it has emerged that Vera's mother has led her
to question whether the operation was necessary in the first place.
Just prior to this excerpt Connie has pointed out that Vera's father
was in considerable pain before the operation.
12 Kei: eh h[eh,
13 X'Con: [well I mean pr - maybe that too,
14 Y but she just can't HANDLE (0.2) [people
close to her being in PAIN.
15 Ver: [reality.
a backache ... is to get down on your hands and knees and ... crawl
around than would be 'no it doesn't', because it opens up a way of
looking at the matter which leaves some room for the other to be
(partially) right.
8. Conclusion
Notes
* We are grateful to Susanna Cumming and Gene Lerner for valuable
discussion of the issues brought up in this paper and especially to Dagmar
Barth, Cecilia Ford, Christine GoW and Susanne Gfinthner for reading and
commenting on an earlier version. Any remaining errors are our own
responsibility. Both authors contributed equally to the conceptualization and
writing of this paper.
1. Unless otherwise indicated, all of the excerpts given as examples in this paper
stem from face-to-face interaction. We have standardized transcriptions to
facilitate reading. For notational conventions, please see the appendix.
2. This notion was implicit in the 'Concessive' relation which Mann and
Thompson identified in written texts (see below).
3. We are grateful to John W. Du Bois, director of the CSAE, for his efforts in
creating the CSAE and making its contents available to us, and to Steve
Albert for sharing his data with us.
4. See, however, Kotthoff 1993 and Vuchinich 1984, 1990 for work in which
concession is understood this way.
5. A minimal token like yeah may of course be merely signalling recipiency
rather than conceding a point. To distinguish between the recipient-signal use
and the conceding use, we suggest that it may be helpful to consider the
prosodic realization of the minimal token. If, as in (4), the token forms a full-
fledged intonation phrase of its own and is not produced with reduced
prosody or in overlap, there are grounds for regarding it as a conversational
move in its own right, which in the context under discussion would have
conceding force. At the same time, as Susanne Gfinthner has suggested to us,
the yeah but format can be seen as a conventionalization of the Cardinal
Concessive pattern, which in many instances of its use has lost all or most of
its concessive force.
6. It is also possible for X' and Y to be expressed hypothetically and linked by
if... then. For instance in one of the Gulf War calls, the moderator claims one
bomb is a big difference from making a hundred or a thousand. and nobody
said he'd make more than one or two (the reference is to Saddam Hussein),
whereupon his interlocutor replies well even then the damage with just one
bomb is quite a bit. We propose that this sequence can be analyzed as:
Concessive patterns in conversation 407
References
Altenberg, Bengt
1986 Contrastive linking in spoken and written English. In: Gunnel Tottie
and Ingegerd Backlund (eds.), English in speech and writing, 13-40.
Uppsala: Academiae Upsaliensis.
Antaki, Charles, and Wetherell, Margaret
1999 Show concessions. Discourse Studies 1: 7-27.
Azar, Moshe
1997 Concession relations as argumentation. Text 17.3: 301-316.
Couper-Kuhlen, Elizabeth
1986 An introduction to English prosody. Ttibingen: Max Niemeyer and
London: Edward Arnold.
Couper-Kuhlen, Elizabeth and Sandra A. Thompson
1999 On the concessive relation in conversational English. In: Fritz-
Wilhelm Neumann and Sabine Schuelting (eds.), Anglistentag 1998
Erfurt: Proceedings, 29-39. Trier: Wissenschaftlicher Verlag.
Gtintlrnner, Susanne
1999 Entwickelt sich der Konzessivkonnektor obwohl zum
Diskursmarker? Grammatikalisierungstendenzen im gesprochenen
Deutsch. Arbeitspapier Nr. 95, Fachgruppe Sprachwissenschaft,
Universitat Konstanz.
Hopper, Paul 1. and Elizabeth C. Traugott
1993 Grammaticalization. Cambridge: Cambridge University Press.
408 Elizabeth Couper-Kuhlen and Sandra A. Thompson
Karantzola, Eleni
1995 Let's talk about concession: the case of the Modern Greek particle
as. Journal ofPragmatics 24.1/2: 55-75.
Konig, Ekkehard
1985a On the history of concessive connectives in English. Diachronic and
synchronic evidence. Lingua 66: 1-19.
1985b Where do concessives come from? In: 1. Fisiak (ed.), Historical
semantics, historical wordformation, 263-282. Berlin: Mouton.
1986 Conditionals, concessive conditionals and concessives: areas of
contrast, overlap and neutralization. In: Elizabeth Traugott, Alice ter
Meulen, Judy Reilly, and Charles Ferguson (eds.), On conditionals,
229-246. Cambridge: Cambridge University Press.
1988 Concessive connectives and concessive sentences: cross-linguistic
regularities and pragmatic principles. In: John A. Hawkins (ed.),
Explaining language universals, 145-166. Oxford: Blackwell.
Kotthoff, Helga
1993 Disagreement and concession in disputes: on the context sensitivity
of preference structures. Language in Society 22: 193-216.
Lerner, Gene H.
1991 On the syntax of sentences-in-progress. Language in Society 20:
441-458.
1996 On the 'semi-permeable' character of grammatical units in
conversation: conditional entry into the turn space of another
speaker. In: Elinor Ochs, Emanuel Schegloff, and Sandra A.
Thompson (eds.), Interaction and grammar, 238-271. Cambridge:
Cambridge University Press.
Longacre, Robert E.
1983 The grammar ofdiscourse. New York: Plenum Press.
Mann, William C. and Sandra A. Thompson
1987a Rhetorical Structure Theory: a theory of Text Organization, prepared
for Livia Polanyi (ed.), Discourse Structure, Ablex. Also available as
Information Sciences Institute Research Report 87-190, 4676
Admiralty Way, Marina del Rey, CA 90292-6695.
1987b A discourse view of concession in written English. In: Scott
DeLancey and Russell Tomlin (eds.), Proceedings of the Second
Annual Meeting of the Pacific Linguistics Conference, 435-447.
Eugene: University of Oregon Linguistics Department.
1988 Rhetorical Structure Theory: towards a functional theory of text
organization. Text 8.3: 243-281.
1992 Relational discourse structure: a comparison of approaches to
structuring text by 'contrast'. In: Shin Ja Hwang and William R.
Merrifield (eds.), Language in context: Essays for Robert E.
Longacre, 19-45. Dallas: SIL.
Martin, J. R.
1992 English text: system and structure. Amsterdam: Benjamins.
Matthiessen, Christian
1995 Lexicogrammatical cartography: English systems. Tokyo:
International Language Sciences Publishers.
Concessive patterns in conversation 409
Vuchinich, S.
1990 The sequential organization of closing in verbal family conflict. In:
A. D. Grimshaw (ed.), Conflict talk. Sociolinguistic investigations of
arguments in conversations, 118-138. Cambridge: Cambridge
University Press.
"that's true, although not really, but still":
Expressing concession in spoken English
Dagmar Barth
This contribution argues that despite what is claimed in the research litera-
ture concession in speech is not rare, though the means used to realize it in
this mode differ from those in writing. After an introduction to the struc-
tural realization of concessive relations in spoken English discourse, two of
the constructions in question will be examined in more detail: those that
occur most frequently, concessive but-constructions, and those that are tra-
ditionally associated with concession, a/though-constructions. The paper
will briefly list the reasons for the preference of paratactic concessive but-
constructions in spoken discourse and then concentrate on concessive a/-
though-constructions. It identifies several discourse functions and relates
them to differences in the realization of the concessive schema in compari-
son to concessive but-constructions.
1. Introduction
X 01 Mora: .hhh but you dOn't get a very good rAte ofexchAnge at
those () [those () banks.
X' Wally: [nO too trUE you don't; nol
Y 05 but it wi il it's bEtter than () [doing it at the bA: r;
Betsy: [but whEn do you change
YOUR money;
The Cardinal Concessive schema is not the only way to accomplish con-
cession. There can be variation with respect to the sequence of the ele-
ments of the concessive schema, the number of participants involved in
producing it and the kind of marking. This paragraph illustrates these
variations and gives a rough indication as to their frequency. For this pur-
pose 8.5 hours of mainly spontaneous spoken, public and private British
and American English data have been searched thoroughly for all realiza-
tions of the concessive relation.
In one dyadic variation, the sequence of X' and Y is reversed. This is the
case in about 34% of the examples in our corpus. Example (2) illustrates
such a case:
[23 lines omitted in which Bush defends himselfby listing women who
hold influential positions in his government]
X Sus:
why do you discriminate against women and ethnic mi-
norities in key positions
Y Bush: I don't, look at the women I have in my staff
X' yes, there is a white man, Jim Baker
Y' but there are so many women that he hardly counts
Apart from the order of X' and Y, variation can also concern the number
of participants involved in the expression of the concessive relation.
Whereas the Cardinal Concessive schema is dialogic, in monologic vari-
ants one single speaker produces a concessive relation. Here pseudo-
dyadic and monadic variations can be distinguished. In the pseudo-dyadic
variation the speaker produces an entire Cardinal Concessive on her own,
enacting the roles of both A and B, as in (3):
01
TD: there's a lEsson in thatfor All ofus.
Don: uh hm. ()
yEAh it's very EAsy.
X' now you cAn't dress rEAl FLASHy, ()
05 by dOing it that wA~·
Y bU:t () you know i'd rAther be () able to sort ofmA:tch;
than to be realflAshy.
[you knOw what i mean?
10 TD: [uh hm,
Don: so: - ()
Expressing concession in spoken English 417
What may also be surprising in the light of earlier research is that in none
of the examples given so far is the concessive relation marked by al-
though or even though. Instead it is marked by but.
Table 1 presents a frequency list for the marking of concessive rela-
tions in our corpus of spoken data. For comparison, it is complemented
by one for concessive markers in written data, which is taken from Ru-
dolph (1996).6
One does not have to look very far to account for the overwhelming pre-
dominance of but-constructions in expressing concessivity in speaking.
As paratactic constructions, they facilitate on-line production. But they
also provide an opportunity for face work by leaving the speaker room to
manoeuvre and by attending to the recipient's need for politeness.
Interestingly, with the exception of the third, these functions are of a dis-
course-organizational nature. This is paralleled by the fact that in some
although-constructions a topic-organizing function, such as topic shift or
topic insertion, can also be detected. Further evidence for these "functions
on a different level" is that they all occur in overlap, either with the con-
ceding function or, partly, with each other. Therefore, the percentages in
Table 3, the table which illustrates the frequency of these functions in the
examples investigated, exceed 100%.
Expressing concession in spoken English 421
Following his statement in lines 1-5 that Blakemore is now living on bor-
rowed time, giving him extra energy to accomplish his job so success-
fully, Clair seems to have realized that his claim was slightlyoverdrama-
tized. In lines 6-9 he downplays it by referring to a fact which is familiar
to him as a well-informed journalist, namely that that particular problem
was eased. He thus restricts the validity of X, Blakemore's suffering be-
ing the reason for his success today.
Although this example may be interpreted as a correction at first sight,
it is important to note that it is still restrictive: Clair only points to an im-
provement in Blakemore's condition; he does not claim that there has
never been any problem. Thus, in contrast to a correction, where the va-
lidity of the previous claim is annulled and which can therefore not be
considered concessive (Giinthner, this volume), in this restrictive example
it is still valid: Blakemore's earlier suffering could still be one of a num-
ber of reasons for his success today, even though it may not be the only
one as Clair's initial statement suggests. In contrast to German, where the
corrective function of obwohl 'although' is already so established that
there are purely corrective examples (Giinthner, this volume), in English
there is still a certain note of concession-restriction present in almost all
potentially corrective examples. Nevertheless, assigning X and Y in these
cases is extremely difficult: although usually provides a hint as to which
part of the structure is X'. In this case, however, what follows although is
more important for the rest of the conversation, i.e. it is Y rather than X'.
This underlines the status of these restrictive examples as a link between
concession and correction (Giinthner, this volume).
Restricting the validity of the previous claim is a strategy often em-
ployed when the speaker expects an unfavourable response, such as dis-
agreement (Pomerantz 1984), to her utterance. Clair also uses it in order
to be agreed with by his interlocutor. That Blakemore was in fact likely to
disagree with Clair's initially overstated claim is indicated by his collabo-
ration in producing the restriction before Clair has even finished his tum
(lines 9-10).
X Larry: the reform measure you got through is not really good
enough
X' Ann: well, it is not good enough as it is right now
X'a although we heard that they have worked on strengthen-
ing the lobbyist gift ban
y but he is at least putting these issues on the table
In this example, Ann first partly agrees with Larry's criticism that the
reform measure they have achieved is not good enough as it is right now
(lines 7-9). The stress on now indicates that at some other point in time it
may be better. She then points out that the situation is, in fact, not as bad
as her initial statement might suggest, since they have achieved partial
success in working on a lobbyist gift ban (X' a, lines 10-13). The informa-
424 Dagmar Barth
tion given in X' a is less directly related to the line of argumentation in that
it gives additional background information, whose relevance in the pre-
sent context is signalled by although. 13 At the same time it restricts the
validity of X' and thus in a way already foreshadows an upcoming coun-
tering Y. Ann then, in fact, returns to the original concessive but-
construction and completes it in lines 14-15.
01 Bern: although the russians dId declare war on japan very very
lAte,
a:nd they dId go into manchUria;
Y but/h we were concErned, ()
05 about their getting tOO strong a fOOthold in Asia.
and perhAps; ()
X' uh althOUgh uh mister trUman never sAId that,
Y' uh uh that might have been an (Underlying)/
10 anOther one ofthe rEAsons, ()
whY the bomb was drOpped.
Bernie is of the opinion that the bomb may also have been dropped be-
cause the United States was concerned about their [Russia's - D.B.]
getting too strong a foothold in Asia (Y, lines 4-7,9-12). Nevertheless,
he is aware of an objection which other people calling in on his show
could possibly make, namely that no one ever mentioned this as a possi-
Expressing concession in spoken English 425
ble reason for the dropping of the bomb. He orients towards this by con-
ceding that this is not a reason given officially: Mister Truman never said
that (X', line 8), and thus forestalls such possible objections. The conces-
sive schema then looks as follows:
As the objection is only a possible one, i.e. not made explicit on the sur-
face, there is no X in the concessive schema.
In a sense this function was also illustrated in (6) above in that the ad-
ditional information, which depicts the situation more accurately than in
Ann's initial reaction to Larry's claim, could also be interpreted as fore-
stalling a possible objection from her collaborators in Common Cause.
We will now see that these discourse functions nicely match the way
in which although-constructions realize the concessive schema. For this
purpose they were compared with the most frequent realization of con-
cessive patterns in our corpus: concessive but-constructions.
Four types of differences were found between these two concessive con-
structions. These include the number of speakers involved in producing
the relation (5.1.), the realization of the elements of the concessive pat-
tern, in particular X and X' (5.2.), the sequence of the concessive ele-
ments X' and Y (5.3.) and the distribution according to text type (5.4.).
Tasks such as those just mentioned also explain why concessive al-
though-constructions frequently introduce a 'new' X': X cannot be taken
up as, again, with additional information and possible objections it is not
(yet) realized on the surface. With the majority of the concessive but-
constructions speaker B takes up X in X' (cf Table 6 and example 2).
The concept of text type (Werlich 1976: 39-41) has been adopted for the
present study because it was assumed that the constructions are more
sensitive to their immediate context than to the nature of the speech event
as a whole (text sort), i.e. that the heat of a discussion in a press confer-
ence is more likely to be reflected in the way a politician tries to defend
herself, for instance, than in the otherwise emotional neutrality character-
istic of this genre.
Werlich distinguishes five text types: descriptive, narrative, exposi-
tory, instructive and argumentative. A text can and usually does consist
of more than one text type. For instance, whereas on the whole the tele-
vised presidential debate can be described as predominantly argumenta-
tive, we can undoubtedly also find clearly descriptive, narrative and ex-
pository chunks in it. 16
Table 8 shows that both constructions prefer descriptive-expository
environments (cf examples (1) and (8». However, there are also signifi-
cant differences with respect to their distribution over the text types.
Whereas concessive but-constructions also occur quite frequently in ar-
gumentative environments (Konig 1988) (cf Table 8 and example (2»,
this context is rather rare for concessive although-constructions. Instead,
the latter occur more frequently in narrative-descriptive parts of texts (cf
430 Dagmar Barth
6. Summary
Notes
1. I would like to thank the editors and their collaborators for valuable com-
ments on earlier versions of this paper. Special thanks is due to Elizabeth
Couper-Kuhlen, Christine Gohl and Susanne Giinthner.
2. More recently, even though has also been taken into consideration as a proto-
typical marker for concessivity (cf. Konig 1994).
3. These numbers indicate the exact location of the example in the corpus: the
initial digits specify the tape, the following digits the time of occurrence.
4. For transcription conventions cf. appendix.
5. This polyfunctionality can be considered an indication of the relationship
between but as a clausal conjunction and a discourse marker (cf. Schiffrin
1987).
6. The approach still needs to be systematically applied to written data.
7. Rudolph claims that the order of the items roughly reflects their frequency in
writing; in particular she refers to though and although as "the main and
most frequent connectives..." (Rudolph 1996: 6), not giving actual numbers,
however. Apart from frequency she employs ordering criteria such as seman-
tic peculiarity and diachronic development. Concessive conditionals have
been neglected in Table 1, except for even ifwhich can be both concessive
and concessive conditional (cf. Kjellmer 1989).
8. This number includes all constructions which can be categorized as conces-
sive according to Couper-Kuhlen and Thompson's criteria.
9. In her survey of the literature, however, Rudolph mentions Kaufmann's
(1974) observation that in spoken language "...the concessive thought is ex-
pressed through a construction with the very frequent aber." (Rudolph 1996:
201) (cf. also Konig 1991: 632).
10. For the notion of face cf. Goffman (1967, 1971); for the notion of politeness
cf. Brown and Levinson (1987). A concise summary is given in Foley
(1997).
11. This excluded examples occurring in texts written to be spoken, such as
advertisements, news and radio features.
12. Although the actual analysis only included this admittedly small number of
although-constructions, my claims are supported by the examples which a
search in the spoken part of the Bank of English on-line sample corpus
yielded (British English, approximately 9 million words).
13. In addition, the prosodic realization of the although-part underlines its status
as an insertion, in that lines 10-13 are produced at a higher speech rate and
with a lower pitch.
14. The fact that but can also be used as a concessive marker in monologic con-
cessive relations once again proves that it is a legitimate grammatical means
for producing concessive constructions, on a par with although and even
though as in traditional accounts.
15. The examples have been archived, classified and statistically analyzed by
means of the Systemic Functional Coder developed by Michael O'Donnell at
Edinburgh University (UK). In the following tables, statistically significant
434 Dagmar Barth
R~ferences
Altenberg, Bengt
1986 Contrastive linking in spoken and written English. In: Gunnel Tot-
tie and Ingegerd Backlund (eds.), English in speech and writing. A
Symposium, 13-40. (Acta Universitatis Upsaliensis, Studia Anglis-
tica Upsaliensis 60.) Uppsala, Stockholm: Almquist and Wiksell.
Beaman, Karen
1984 Coordination and subordination revisited: Syntactic complexity in
spoken and written narrative discourse. In: Deborah Tannen (ed.),
Coherence in spoken and written discourse, 45-80. Norwood, New
Jersey: Ablex.
Brown, Penelope and Stephen Levinson
1987 Politeness: Some universals in language usage. Cambridge: Cam-
bridge University Press.
Brunet, Jacqueline
1995 La subordination: chronique d'un declin annonce. In: Hanne L.
Andersen and Gunvar Skytte (eds.), La subordination dans les lan-
gues romanes, 57-68. (=Etudes Romanues 34.) Institut d'Etudes
Romanes, Universite de Copenhague.
Chafe, Wallace
1988 Linking intonation units in spoken English. In: John Haiman and
Sandra A. Thompson (eds.), Clause combining in grammar and dis-
course, 1-27. Amsterdam/Philadelphia: Benjamins.
Couper-Kuhlen, Elizabeth and Sandra A. Thompson
1999 On the concessive relation in conversational English. In: Fritz-
Wilhelm Neumann and Sabine Schiilting (eds.), Anglistentag Erfurt
1998. Trier: Wissenschaftlicher Verlag.
Di Meola, Claudio
1997 Der Ausdruck der Konzessivitat in der deutschen Gegenwartsspra-
che: Theorie und Beschreibung anhand eines Vergleichs mit dem
ltalienischen. Tiibingen: Niemeyer.
Foley, William A.
1997 Anthropological linguistics. An introduction. Oxford: Blackwell.
Expressing concession in spoken English 435
Ford, Cecilia
1993 Grammar in interaction: Adverbial clauses in American English
conversations. Cambridge: Cambridge University Press.
GofIman, Erving
1967 Interaction ritual: Essays on face-to-face behavior. New York:
Doubleday.
1971 Relations in public: Microstudies of the public order. New York:
Harper and Row.
Gohl, Christine and Susanne Gunthner
1999 Grammatikalisierung von weil als Diskursmarker in der gesproche-
nen Sprache. Zeitschriftfur Sprachwissenschaft 18: 39-75.
Grote, Brigitte, Nils Lenke and Manfred Stede
1997 Ma(r)king concessions in English and German. Discourse Proc-
esses 24: 87-117.
Haiman, John and Sandra A. Thompson (eds.)
1988 Clause combining in grammar and discourse. Amster-
damIPhiladelphia: Benjamins.
Halliday, M.A.K.
1985 An introduction to functional grammar. London: Arnold.
Hermodsson, Lars
1978 Semantische Strukturen der Satzgefuge im kausalen und konditiona-
len Bereich. Uppsala: Almqvist and Wiksell.
Hopper, Paul
1987 Emergent grammar. In: General session and parasession on gram-
mar and cognition. Edited by Berkeley Linguistics Society, Febr. 1-
16, 1987, 139-157.
Kaufmann, Gerhard
1974 Zur "konzessiven" Beziehung. Zielsprache Deutsch 5: 1-21.
Kjellmer, Goran
1989 Even lfand even though. English Studies 3: 256-269.
Koerfer, Armin
1979 Zur konversationellen Funktion vonja aber. Am Beispiel uiliversi-
tarer Diskurse. In: Harald Weydt (ed.), Die Partikeln der deutschen
Sprache, 14-29. BerlinlNew York: Mouton de Gruyter.
Konig, Ekkehard
1985a On the history of concessive connectives in English. Diachronic and
synchronic evidence. Lingua 66: 1-19.
1985b Where do concessives come from? On the development of conces-
sive connectives. Jacek Fisiak (ed.), Historical semantics, 263-282.
BerlinlNew York: Mouton de Gruyter.
1988 Concessive connectives and concessive sentences: Cross-linguistic
regularities and pragmatic principles. In: John A. Hawkins and
SandraA. Thompson (eds.), Explaining language universals, 145-
166. New York: Blackwell.
436 Dagmar Barth
Sacks, Harvey
1987 On the preference for agreement and contiguity in sequences in
conversation. In: Graham Button and John R.E. Lee (eds.), Talk and
social organization, 54-69. Clevedon: Multilingual Matters.
Schegloff, Emanuel, Gail Jefferson and Harvey Sacks
1977 The preference for self-correction in the organization of repair in
conversation. Language, 53: 361-382.
Schiffrin, Deborah
1987 Discourse markers. Cambridge: Cambridge University Press.
Stojanova-Jovceva, Stanka
1979 Untersuchungen zur Stilistik der Nebensatze in der deutschen Ge-
genwartssprache. Beitrage zur Geschichte der deutschen Sprache
und Literatur 100: 40-178.
Werlich, Egon
1976 A text grammar ofEnglish. Heidelberg: QueUe and Meyer.
Weydt, Harald (ed.)
1979 Die Partikeln der deutschen Sprache. BerlinlNew York: Mouton de
Gruyter.
From concessive connector to discourse marker:
The use of obwohl in everyday German
• • *
Interaction
Susanne Gunthner
1. Introduction
This paper deals with the relation between grammar and discourse, and
thus, with the use and functions of syntactic constructions in everyday
conversation (Schegloff/OchsfIhompson 1996; Schlobinski 1997). In
analyzing obwohl-constructions in everyday colloquial German inter-
action, this study aims at uncovering how grammatical resources are
exploited by interactants to fulfill communicative functions. 1
In spoken colloquial German, speakers frequently make use of
syntactic constructions treated as ungrammatical by traditional German
grammar: the use of obwohl 'although' and weil 'because' with main-
clause syntax (Eroms 1980; Gaumann 1983; Kuper 1991; Schlobinski
1992; Giinthner 1993, 1996; Keller 1993; Wegener 1993; Pasch 1997;
Uhmann 1998). The increasing use of main-clause syntax in seemingly
"subordinate" concessive and causal clauses has led to heavy criticism
among self-announced "caretakers" (Sprachpjleger) of the German
language, and has even resulted in the foundation of an "advocacy group"
2
(Aktionsgemeinschaft) to save German subordinate-clause order.
440 Susanne Gunthner
The obwohl-clause in (1) shows final positioning of the finite verb war
'was'. During the last ten to fifteen years, however, in spoken colloquial
German as well as in certain written genres that reproduce colloquial
language (e.g. in interviews, dialogues in advertising, and in hybrid
genres like e-mail communication and chat talk) speakers frequently use
obwohl-clauses with verb-second position. 5 The following example is
taken from a conversation between Lisa and Mara. Lisa has just
mentioned that she and her boy-friend have decided to go on vacation on
the Maldive Islands this year. In her reply Mara says she wishes she
could go too:
As Quirk et ale (1985: 1098) state, "Concessive clauses indicate that the
situation in the matrix clause is contrary to expectation in the light of what
is said in the concessive clause". Klara's utterance ich bin dann- bin
dann doch Mitglied geworden, 'then 1- then I became a member after all'
is presented as contrary to the expectation raised by the concessive clause
obwohl ich eigentlich nie in ne Partei (r)eintreten wollte 'although I
actually never wanted to join a political party' . The concessive
construction thus depends-as Konig/van der Auwera (1988: 107)
indicate-on a "relationship of 'normal incompatibility' or dissonance
between the two component propositions"; i.e. concessive constructions
license the inference that the two sentences asserted to be true "are
instances of situations that do not normally go together" (Konig 1994:
8
681). This can be stated as presupposing a conditional:
If p', then normally -, q'.
Accordingly, the discourse presupposition in the example (3) is: if
someone has never wanted to join a political party, then he/she normally
does not become a member.
Logically, concessive clauses entail both their component clauses, i.e.,
both "p" (she never wanted to join a political party) and "q" (she became
a member after all) are treated as valid (Konig 1991; 1994).9
MannfThompson (1992: 39), who analyze concessive relations within
the framework of Rhetorical Structure Theory, argue that speakers make
use of concessive relations
Applied to our example, Klara presents the information in the nucleus ich
bin dann- bin dann doch mitglied geworden, 'then 1- then I became a
member after all' as valid, despite the seemingly contradictory
information in the satellite: ich wollte eigentlich nie in ne partei
(r)eintreten 'although I actually never wanted to join a political party'. In
so doing, she removes the statement in the satellite as an obstacle, and
From concessive connector to discourse marker 443
(4) MANUSCRIPT
22Anna: obwohl das noch iiberarbeitungsbediirftig isch
WOLLT sie 's mitnehm.
Also in the case of (4) the speaker (Anna) promotes the validity of the
information in the nucleus (sie wollt es mitnehm; 'she wanted to take it
along' in spite of apparent contrary information in the satellite.
Prosodically, the obwohl-clause and the following main clause form one
intonational contour and are thus realized as a single intonational
trajectory.
Concessive obwohl-clauses can also be inserted into the main clause.
The following example is part of an episode in which Urs talks about his
political activism as a student:
With this complex sentence, Drs asserts two facts (he was a member of
the Junge Union and he carried the Mao Bible with him) against the
background of the assumption that these two do not normally go together:
if you are a member of a conservative party, you do not carry around a
Mao Bible. The satellite status of the inserted parenthetical obwohl-clause
444 Susanne Gunthner
20 (0.5)
21Mia: obwohl () ich komm doch besser bei dir vorbei.
22 des isch glaub =ich gschickter.
Hans first rejects the offer and states his dislike for green tea (line 45).
After a short pause, he changes his mind and asks for a half a cup (line
47). Here, the obwohl-clause corrects the validity of the preceding speech
act (Hans' refusal). Not only do both clauses have their own prosodic
contours, they also have their own illocutionary forces. In the next
example (8), the obwohl-utterance corrects a preceding question:
(8) PARTY
64Kati: kommt der eigentlich au?
65 (-)
66Kati: [0 obwohl) () s=isch mir eigentlich ega:/. 0
67Lars: [[glaub schon))
Examples such as (7) and (8) clearly reveal that the obwohl-utterances
are no longer within the scope of the illocutionary force of their main
clauses (Lehmann 1991: 526).
The correction introduced by obwohl may involve various levels of the
preceding statement: whereas in the case of (2), (6) and (7) the correction
was related to an explicitly stated utterance, the following example (9)
shows that obwohl can also be used to initiate the correction of a possible
implication. Ulla is talking about a neighbor who has suddenly fallen
seriously ill, and comes to the conclusion that one should consider oneself
lucky to have enjoyed good health so far.
448 Susanne Gunthner
(9) ILLNESS
36Ulla: do: kann man bis=jetzt=no=eigentlich (-)
37 Tal. Tal. Tal () no ganz FROH sei. gell?
38 (0.5)
39 OBWOHL man weij3 jo gar net was in oim SCHLUMMERT
40 (1.5)
41 vielleicht sen mir au scho bald mol DO. GWESE.
43Rolf: des weij3 mer halt [nie:.]
44Ulla: [hajo]
45Rolf: do steckt mer halt net drinne.
In lines 36-37 VIla states that up to now one can be glad, inferring that
she has been healthy so far. However, after a short pause, the obwohl-
clause introduces a sudden change in her perspective: obwohl man weij3
jo gar net was in oim SCHLUMMERT 'although you never know what's
slumbering inside of you' and thus corrects the suggested inference. The
correction does not affect the factual content of the preceding statement,
but a possible conclusion which may be drawn from that utterance. Thus,
corrective obwohl-constructions may also operate on the level of possible
conclusion, i.e., what Sweetser (1990) calls the "epistemic domain".
Syntactic non-integration (do: kann man bis=jetzt=no==eigentlich (-)
Tal. TOl. TOl () no ganz FROH sei. gell? (0.5) OBWOHL man weij3jo
gar net was in oim SCHLUMMERT.) is again echoed by prosodic
discontinuity: the two clauses have their own intonation contours and are
separated by a pause of 0.5 seconds. The obwohl is marked by a local
increase in volume and a rise in pitch.
Corrections using obwohl can also relate to a term used in a preceding
utterance. 18 In (10) Elli talks about a relative who plays various
instruments:
From concessive connector to discourse marker 449
(10) VIOLA
46Elli: der spielt do- doch au (soviel=[i=weifi)] CELLO.
47Hans: [de karl]
48Hans: alles MDGLICHE [(spielt=der).}
49Elli: [obwohl} (-) des isch glaub=i
50E/li: a BRATSCHE (was er noch spielt.)
75Lisa: [echt?]
75Anja: ich hab den (hi) echt kaum wieder(hi)erkannt.
Consider (12): Lina and Ira are discussing Ira's plans to defend her
doctoral thesis once she will have started her teacher training program:
Concessive Corrective
obwohl obwohl
(q, obwohl p) (q, obwohl p)
dialogical use:
disagreement
From concessive connector to discourse marker 453
Formal characteristics:
position of the
obwohl syntagma initial and final final only
concessive obwohl the speaker asserts "two facts against the background
of an assumption that situations of these types are incompatible and do
not normally go together" (Konig 1994; and Konig in this volume), in the
case of corrective obwoh/ the speaker asserts a dissonance or in-
compatibility between her/his upcoming utterance and her/his preceding
one. The element of dissonance is transferred to the discourse level. Thus,
the semantics of concessive obwoh/ is still partly inherent in the corrective
use of obwohl: it also expresses a dissonance or incompatibility. It loses
the conceding element, however. Thus, there is a loss in the semantic
complexity of the source item.
These two uses of obwoh/ (as a concessive subordinator and a
corrective marker) can be seen as two ends of a scale, which has
intermediate gradations. One such intermediate option is the restrictive
use of obwohl. This use does not share the typical characteristics of a
concessive relation (it does not presuppose "if p, normally"" q"), nor does
it cancel the validity of the preceding utterance (as the corrective obwohl
does). Instead, it restricts the preceding statement or a conclusion which
can be drawn from the preceding utterance. 24 Restrictive uses of obwohl
show verb-final as well as verb-second position.
The following transcript shows a restrictive obwoh/-construction with
final ordering of the finite verb. Elli and Fritz are talking about Greta and
her husband Hans, who "adores" her:
else". However, as with concessive relations, both "p" and "q" are
entailed and treated as valid; i.e., "p" (obwohl==s scho au grenze
{fiir==en] gebe hot., 'although he has reached his limits at times' as well
as "q" (der der hot doch viel fiir die [gmacht.], 'he he's done a lot for
her.' are logically true.
In our example (14), syntactic integration (via verb-final positioning in
the obwohl-construction) marks a close connection between the two
clauses. The prosodic realization, however, indicates that the two clauses
are produced as distinct units: the utterance in line 34 carries a fmal
intonation contour which marks completeness. Syntactic non-integration
(via verb-second positioning in the obwohl-construction) would be
possible without any change in meaning: der der hot doch vie:: 1fiir die
gmacht. obwohl s==hot scho au grenze fiir==en gebe. And indeed, in
colloquial German, speakers also use main clause syntax with this kind of
restrictive obwohl-construction.
In (15), Karl is talking about the composer Penderecki and his musical
oeuvre. In line 4 he produces an obwohl-construction in order to restrict
the validity of his preceding statement:
(15) PENDERECKI
lKarl: nein==nein ()
2 der is (sicher) nicht hi SCHRILL. (-)
3 was sicherlich der fall ist,
4 <<all> obwohl ich kenn nicht sein gesamtwerk, >
5 seine Musik ist zeitgenossisch aber nicht dogmatisch.
it is not surprising that the linear order is such that the obwohl-clause is
generally post-posed: having stated the "matrix", the speaker sees the
necessity of restricting the validity of this statement or of a possible
conclusion which may be drawn from it.
One should add, however, that the distinctions between the different
uses of obwohl are not clear-cut. Sometimes it remains unclear whether
an obwohl-utterance totally cancels or only restricts a preceding
utterance. Thus, we arrive at a continuum ranging from the concessive
subordinator obwohl via the restrictive use to obwohl as a correction
marker.
Due to its possible main clause syntax (and resulting main clause
phenomena), its prosodic independence and its loss of freedom in the
linear order, obwohl in its restrictive use can no longer be classified as a
subordinate conjunction. 26 Furthermore, due to the fact that restrictive
obwohl-constructions no longer presuppose "If p, then normally --, q",
they no longer represent concessive relations in the traditional sense. The
corrective use of obwohl is even further away from its original use as a
concessive subordinator: it does not show features of subordination nor
does it possess the central semantic characteristic of concessives; i.e. the
entailment of "p" and "q". Finally, its scope is extended to the level of
connecting utterances and units of discourse.
Given that corrective obwohl has lost syntactic, semantic, pragmatic
and discourse-functional properties which are characteristic for
concessive subordinators,27 the question arises: has the concessive
From concessive connector to discourse marker 457
...there are many words and phrases in English, and no doubt most
languages, that indicate the relationship between an utterance and the prior
discourse. Examples are utterance-initial usages of but, therefore, in
conclusion, to the contrary, still, however, anyway, well, besides, actually,
all in all, so, after all, and so on. It is generally conceded that such words
have at least a component of meaning that resists truth-conditional
treatment. .. What they seem to do is indicate, often in very complex ways,
just how the utterance that contains them is a response to, or a continuation
of, some portion of the prior discourse.
... an expression which signals the relationship of the basic message to the
foregoing discourse. [...] They (Le. "discourse markers"; S.G.) provide
instructions to the addressee on how the utterance to which the discourse
marker is attached is to be interpreted.
1998: 134). The examples Crevels provides show that in its textual use,
Spanish aunque comes close to the restrictive obwohl: "the concessive
clause is not meant as a plain comment on the felicity of the preceding
assertion, but rather modifies the preceding text unit by excluding certain
interpretation options which have become available to the hearer in the
course of it. (... ) Rectifying concessive clauses are even more loosely
linked to a main clause: They can only follow the main clause and
typically exhibit main clause order in languages like German and Dutch"
(Crevels 1998: 141).
Onodera's (1995) analysis of the development of the connectors demo
and dakedo in Japanese illustrates a similar process from concessive
conjunctions to discourse markers. Demo and dakedo develop from
clause-final concessive subordinators, indicating referential contrast to
clause-initial discourse markers used to signal contrastive actions. Parallel
to this semantic, syntactic and pragmatic change "the scope of the
connecting function is enlarged" from the sentence to the discourse level,
and the pragmatic functions are expanded (Onodera 1995: 422ff).
These studies show that the use of a traditionally concessive sub-
ordinator as a contrastive discourse marker is not restricted to German,
but that various languages show similar developments. 32 Still, there are
differences concerning the extent of the contrast: in English, paratactic
although may be used to restrict a preceding utterance, but "q" is still
treated as valid. In German and Japanese, however, those contrastive
markers originating from concessive connectors (obwohl, demo and
dakedo) can be used to cancel a preceding utterance or communicative
action.
7. Conclusion
Notes
32. Cf. also Konig's (1994: 681) discussion of rectifying concessive clauses. He
mentions the use of French encore que for this subtype of concessive
constructions. Even in Latin the concessive connectors quamquam and etsi
(although) were sometimes used to restrict or correct the preceding sentences.
Cf. Kuhner/Stegmann's (1962) Latin grammar.
33. Cf. Traugott (1995a~ b), Auer (1996), Gtinthner (1999) and Gohl/Gtinthner
(1999).
References
Auer, Peter
1996 The pre-front field position in spoken German and its relevance as a
grammaticalization position. Pragmatics 6 (3): 295-322.
Brinton, Laurel J.
1996 Pragmatic Markers in English. BerlinlNew York: de Gruyter.
Brown, Penelope and Stephen C. Levinson
1978 Universals in language usage: Politeness phenomena. In: Esther
Goody (ed.), Questions and Politeness, 56-311. Cambridge:
Cambridge University Press.
Crevels, Mily
1998 Concession in Spanish. In: Hannay, Mike and A. Machelt
Bolkestein (eds.), Functional Grammar and Verbal Interaction,
129-148. Amsterdam: Benjamins.
Di Meola, Claudio
1997 Der Ausdruck der Konzessivittit in der Deutschen Gegenwarts-
sprache. Tubingen: Niemeyer.
Eisenberg, Peter
1993 Der Kausalsatz ist Dicht zu retten. Praxis Deutsch 118: 10-11.
Eroms, Hans-Werner
1980 Funktionskonstanz und Systemstabilisierung bei den begrundenden
Konjunktionen im Deutschen. Sprachwissenschaft 5: 73 -115.
Fraser, Bruce
1990 An approach to discourse markers. Journal of Pragmatics 14: 383-
395.
1996 Pragmatic markers. Pragmatics 6 (2): 167-190.
1998 Contrastive· discourse markers in English. In: Andreas Jucker and
Yael Ziv (eds.), Discourse Markers. Descriptions and Theory, 301-
326. AmsterdamIPhiladelphia: Benjamins.
Gaumann, Ulrike
1983 'Weil die machen jetzt bald zu '. Angabe- und Junktivsatz in der
Deutschen Gegenwartssprache. Goppingen: Kummerle.
Giv6n, Talmy
1990 Syntax. A Functional-Typological Introduction. Amsterdam:
Benjamins.
From concessive connector to discourse marker 465
Keller, Rudi
1993 Das epistemische WElL. Bedeutungswandel einer Konjunktion. H.
J. Heringer and G. Stotzel (eds.), Sprachgeschichte und
Sprachkritik, 219-247. BerlinINew York: de Gruyter.
Konig, Ekkehard
1991 Konzessive Konjunktionen. In: Arnim von Stechow and Dieter
Wunderlich (eds.), Semantik. Ein internationales Handbuch der
Zeitgenossischen Forschung, 631-639. Berlin: de Gruyter.
1994 Concessive Clauses. In: R. E. Asher (ed.), The Encyclopedia of
Language and Linguistics, 679-681. Oxford: Pergamon.
Konig, Ekkehard and Jan Van der Auwera
1988 Clause integration in German and Dutch conditionals, concessive
conditionals and concessives. In: John Haiman and Sandra A.
Thompson (eds.), Clause Combining in Grammar and Discourse,
101-133. Amsterdam: Benjamins.
Kortmann, Bernd
1998 Adverbial subordinators in the languages of Europe. In: Johan van
der Auwera (ed.), Adverbial Constructions in the Languages of
Europe, 457- 562. BerlinlNew York: Mouton de Gruyter.
Kuhner, Raphael and Stegmann Carl
1962 Ausfuhrliche Grammatik der Lateinischen Sprache. Darmstadt:
Wissenschaftliche Buchgesellschaft.
Kuper, Christoph
1991 Geht die Nebensatzstellung im Deutschen verloren? Zur prag-
matischen Funktion der Wortstellung in Haupt- und Nebensatzen.
Deutsche Sprache 19: 133-158.
Lehmann, Christian
1991 Grammaticalization and related changes in contemporary German.
In: Elizabeth C. Traugott and Bernd Heine (eds.), Approaches to
Grammaticalization, 493-535. Amsterdam: Benjamins.
Lerch, Eugen
1929 Historische Franzosische Syntax. Leipzig: O. R. Reisland.
Levinson, Stephen C.
1983 Pragmatics. Cambridge: Cambridge University Press.
Mann, William C. and Sandra A. Thompson
1992 Relational discourse structure: A comparison of approaches to
structuring text by 'contrast'. In: S. Hwang and W. R. Merrifield
(eds.), Language in Context: Essaysfor Robert E. Longacre, 19-46.
Arlington: University of Texas Press.
Metrich, Rene
1980 Zur Syntax und Semantik von 'obwohl' und 'wenn auch'. Paris:
Centre Universitaire du Grand Palais.
Onodera, Noriko O.
1995 Diachronic analysis of Japanese discourse markers. In: Andreas
Jucker (ed.), Historical Pragmatics. Pragmatic Developments in the
History ofEnglish, 393-437. AmsterdamlPhiladelphia: Benjamins.
From concessive connector to discourse marker 467
Pasch, Renate
1997 Weil mit Hauptsatz-Kuckucksei im Denn-Nest. Deutsche Sprache
3: 252-271.
Pomerantz, Anita
1984 Agreeing and disagreeing with assessments: some features of
preferred/dispreferred tum shapes. In: John M. Atkinson and John
Heritage (eds.), Structures ofSocial Action: Studies in Conversation
Analysis, 57-101. Cambridge: Cambridge University Press.
Quirk, Randolph, Sidney Greenbaum, Geoffrey Leech and Jan Svartvik
1985 A Comprehensive Grammar of the English Language. London:
Longman.
Rudolph, Elisabeth
1996 Contrast. Adversative and Concessive Relations and their
Expressions in English, German, Spanish, Portugese on Sentence
and Text Level. Berlin/New York: de Gruyter.
Schegloff, Emanuel A., Gail Jefferson and Harvey Sacks
1977 The preference for self-correction in the organization of repair in
conversation. Language 53 (2): 361-382.
Schegloff, Emanuel A., Elinor Ochs and Sandra Thompson (eds.)
1996 Interaction and Grammar. Cambridge: Cambridge University Press.
Schiffrin, Deborah
1987 Discourse Markers. Cambridge: Cambridge University Press.
Schlobinski, Peter
1992 Funktionale Grammatik und Sprachbeschreibung. Eine Unter-
suchung zum Gesprochenen Deutsch sowie zum Chinesischen.
Opladen: Westdeutscher Verlag.
Schlobinski, Peter (ed.)
1997 Syntax des Gesprochenen Deutsch. Opladen: Westdeutscher Verlag.
Selting, Margret, Peter Auer, Birgit Barden, Elizabeth Couper-Kuhlen, Susanne
Giinthner, Uta Quasthoff and Peter Schlobinski
1998 Gesprachsanalytisches Transkriptionssystem (GAT). Linguistische
Berichte 173: 91-122.
Sweetser, Eve E.
1990 From Etymology to Pragmatics. Metaphorical and Cultural Aspects
ofSemantic Structure. Cambridge: Cambridge University Press.
Tabor, Whitney and Elizabeth C. Traugott
1998 Structural scope expansion and grammaticalization. In: Anna
Giacalone-Ramat and Paul Hopper (eds.), The Limits of
Grammaticalization, 229-272. Amsterdam: Benjamins.
Traugott, Elizabeth C.
1995a The Role of the Development of Discourse Markers in a Theory of
Grammaticalization. Paper presented at the 12th Inter~ational
Conference on Historical Linguistics. University of Manchester.
1995b Subjectification in grammaticalization. In: Dieter Stein and Susan
Wright (eds.), Language, Subjectivity and Subjectivisation, 31- 54.
Cambridge: Cambridge University Press.
468 Susanne Gunthner
Uhmann, Susanne
1992 Contextualizing relevance: On some forms and functions of speech
rate changes in everyday conversation. In: Peter Auer and Aldo di
Luzio (eds.), The Contextualization of Language, 297-336.
Amsterdam: Benjamins.
1998 Verbstellungsvarianten in weil-Satzen: Lexikalische Differenzie-
rung mit grammatischen Folgen. Zeitschrift fur Sprachwissenschaft
17.1: 92-139.
Wegener, Heide
1993 weil-das hat schon seinen Grund. Zur Verbstellung in
Kausalsatzen mit WElL im gegenwartigen Deutsch. Deutsche
Sprache 4: 289-305.
Index
288, 290, 341, 343, 348-353, 119-121, 123, 126, 131, 181,
355, 361-369, 371-374, 376, 236, 239f., 243-245, 300, 313,
378, 381-387, 389f., 393-400, 318, 320, 323, 326, 329, 349,
403-405, 411-414, 417-420, 363,370
422, 425, 430f., 439, 452, 454, contrast 2, 136, 225f., 244f., 247-
458 249, 258f., 283-285, 288-294,
Cardinal Concessive Schema 296-301, 303-306, 364, 382,
356f., 384-392, 394-398, 459f.
402-405,442-446,427-431 conversation 84, 89, 92, 96, 111,
concessives 5, 178, 251f., 313f., 173, 193, 222, 283, 286, 291-
318-320, 336, 342f., 357, 362f., 293, 296, 349, 381, 384, 388,
365,370,373,445,456 393, 401-403, 405f., 413, 422,
conditional clause 439-441
pre- vs. postposed 146, 148, conversation analysis 84, 89, 92,
154f., 157, 159, 173, 179f., 173,213,283, 292f.
181, 183-194, 196, 322 corpus analysis 5, 70, 77f., 146,
conditional perfection 149 164,173, 178f., 194f., 259, 417,
conditionals 5, 114-116, 118f., 122, 420
125f., 132, 134f., 139, 143, 147, correction 147, 264-267, 302, 305,
154, 156-159, 163, 165-168, 404, 422, 439, 441, 444-448,
174f., 178, 184, 205-208, 210, 450,452,456, 459f.
213,223-226,230, 342f., 369 corrective marker 454, 458f., 461
complex conditional connective
143
content 116, 139 desirability 205f., 208f., 212, 215f.,
counterfactual 2f., 12f., 139, 226f.,229,231
152f., 156, 166, 205-209, disagreement 5, 87, 93, 103, 278,
211-218, 221-231, 342 283, 286, 288f., 292f., 297, 355,
epistemic 116, 122, 135, 156 382, 403f., 405, 419, 422, 439,
speech-act 118, 126, 139, 157f., 441, 450-453, 460
168 discourse function 146, 184, 225f.,
conjunction 1-3, 6, 46, 48, 52, 59, 411f., 420f., 427, 431f., 439,
83f., 88f., 111-113, 118-123, 441,452
125-127, 132, 135-139, 147, discourse grammar 461
161, 175, 177, 195, 198, 209, discourse marker 5f., 8, 257f., 275,
218, 236-253, 259f., 289, 294, 439, 441, 457-460
300, 313, 315f., 318, 341, 346, division of labour between grammar
352, 354, 363f., 374, 391, 4Q2, and pragmatics 235, 246f., 249,
456- 460 253
connective If., 5f., 37, 57-63, 68, duality 21, 66, 350f., 355, 362, 376
7Of. , 75, 77f., 143-146, 149, Dutch 3, 6, 46, 52, 57-61, 63f., 66,
151, 162f., 167f., 257f., 267, 70f., 77f., 317, 319, 347, 375,
272, 300, 317, 341, 344, 346f., 460
352f., 355, 359, 363, 370, 375, dyadic 381f., 403, 414-416, 427
383,418
content domain/level/relation 36-38,
40, 43, 51f., 54, 59f., 63, 114,
Index 471
Christine Gohl
Elizabeth Couper-Kuhlen Sprachwissenschaft
Sprachwissenschaft (Anglistik) Universitat Konstanz
Universitat Konstanz Postfach 5560
Postfach 5560 0-78434 Konstanz
0-78434 Konstanz Germany
Germany Christine. GoW@uni-konstanz.de
elizabeth.couper@uni-konstanz.de
Susanne Giinthner
Mily Crevels Sprachwissenschaft
Dept. of Comparative Linguistics Universitat Konstanz
Leiden University Postfach 5560
PO Box 9515 0-78434 Konstanz
NL-2300 RA LEIDEN Germany
The Netherlands Susanne. Guenthner@uni-konstanz.de
crevels@rullet.leidenuniv.nl
474 Authors' addresses
Ewald Lang
Zentrum flir Allgemeine Ted Sanders
Sprachwissenschaft Dept. of Dutch/Centre for Language and
Jagerstr. 10/11 Communication
0-10117 Berlin Utrecht University
Germany Trans 10
Lang@zas.gwz-berlin.de NL-3512 JK Utrecht
The Netherlands
Ted. Sanders@let.uu.nl
Paul Georg Meyer
Institut flir Anglistik III
RWTHAachen Scott Schwenter
Karmanstr. 17-19 Dept. of Spanish and Portuguese
0-52062 Aachen The Ohio State University
Germany 266 Cunz Hall
paul.meyer@anglistik3.rwth-aachen.de 1841 Millikin Road
Columbus, OR 43210-1229
USA
Estrella Montolio schwenter.1@osu.edu
Facultat de Filologia
Dept. de Filologia Hispanica (Lengua
Espanola) Peter Siemund
Universitat de Barcelona Institut flir Englische Philologie
Gran Via de les Corts Catalanes, 585 Freie Universitat Berlin
08007 Barcelona Gosslerstrasse 2-4
Spain 0-14195 Berlin
montolio@fil.ub.es Germany
siemund@zedat.fu-berlin.de
Authors' addresses 475
Arie Verhagen
Eve Sweetser Dutch Linguistics
Dept. of Linguistics Leiden University
University of California at Berkeley P.O. Box 9515
1203 Dwinelle Hall 2300 RA Leiden
Berkeley, CA 94720-2650 The Netherlands
USA Arie. Verhagen@let.LeidenUniv.nl
sweetser@cogsci.berkeley.edu
Topics in English Linguistics
Edited by Bernd Kortmann and
Elizabeth Closs Traugott
Mouton de Gruyter . Berlin· New York