Causal Categories in Discourse and Cognition

Transcription

Causal Categories in Discourse and Cognition
Causal Categories
in Discourse and Cognition
Categories of subjectivity in Dutch causal
connectives: a usage-based analysis
Ninke Stukker, Ted Sanders and Arie Jlerhagen
1. Introduction 1
Edited by
Ted Sanders
Eve Sweetser
The meaning and the use of different types of causality markers has often been
described with reference to our conceptual understanding of causality (e.g.
Talmy 1988; Verhagen and Kemmer 1997; Wolff and Song 2003). We focus
on causal connectives -causality markers functioning at the discourse level of
linguistic structure, relating discourse segments into a coherent whole (cf. Hobbs
1979; Mann and Thompson 1988; Sanders, Spooren and Noordman 1993). In
this paper, we adopt the cognitive semantic view that causality markers specifically have a categorizing function: when selecting one of the options available
in a language, the speaker assigns the causal relations expressed to a specific
conceptual type of causality. The meaning and use of causal connectives has
recently been characterized in terms of the concept of "subjectivity" (e.g. Pander Maat and Sanders 2000, 2001; Pander Maat and Degand 2001; Pit 2003;
Degand and Pander Maat 2003; Stukker 2005; Verhagen 2005; see also the
contributions to the present volume). Thus, typical examples of the frequently
used Dutch connectives daardoor, daarom and dus marking "forward" Cl!usality
(cause precedes effect in order of presentation) can be characterized as follows:
(I)
Hetwas extreem koud. Daardoor waren de waterleidingen gesprongen.
'It was extremely cold. Daardoor the water pipes had burst.'
(2)
Het was extreem koud. Daarom :-tochten we een caje op.
'It was extremely cold. Daarom we entered a cafe.'
(3)
Het is onbewolkt. Dus het zal we! koud worden vandaag.
'The sky is clear. Dus it will probably be cold today.'
2009
Mouton de Gruyter
Berlin · New York
Daardoor is typically used for marking objective, "non-volitional" causal
processes occurring in observable reality (1 ). Daarom is typically used in contexts of objective, 'volitional" causality, in which an volitional action performed
in the real world ("entering a care" in (2]), is motivated by a situation presented
as the cause ("the extreme cold"). Dus is typically used for marking subjective,
l
This chapter goes back to Stukker (2005) -- notably chapter 4.
120
Ninke Stukker, Ted Sanders and Arie ~rhagen
epistemic causal relations, in which a causal relation is constructed on the illocutionary level, between a conclusion of the speaker, presented as the causal
effect ("it will be cold today" in [3]), and an argument fimctioning as the causal
antecedent ("the clear sky").
The idea that each one of the connectives is related to a specific conceptual
model of causality will be referred to in this paper as the "categorization hypothesis". This hypothesis v11as tested in several corpus studies (e.g. Pander Maat and
Sanders 1995; Pander Maat and Degand 2001; Pit 2003; Stukker 2005). These
studies revealed that, indeed, the majority of connectives' natural usage contexts
reflect these conceptual categories of causality more or less directly. In a minority of cases, however, the relation of connectives to their assumedly typical
causality category appears to he less straightforward. Under specific circum.stances, connectives are used in contexts which are taken to belong to other
connectives' causality categories. Well-knoWll examples are the usage of dus in
volitional causal relations as in (4), and the usage of daarom in epistemic causal
relations as in (5).
(4)
Het was extreem koud. Dus zochten we een cafe op.
'It was extremely cold. Dus we entered a care.'
(5)
llet is onbewolkt. Daarom zal het wei koud wonien vandaag.
'The sky is clear. Daarom it will probably be cold today.'
Should these findings he interpreted as evidence against the categorization
hypothesis? This is indeed what has been proposed by several of lhe studies
mentioned above. In this paper, we follow a different line of reasoning. We
re-interpret the findings from previous connective studies within a usage-based
framework. The "usage-based approach to language" assumes that variation is an
inherent characteristic oflangnage use, and seeks tu explain occurring patterns
ofvariation with reference to more general cognitive mechaniqms (cf. Langacker
!987; Bybee 1985, 2006, 2007; contributions to Barlow and Kemmer2000).ln
line with this framework, we propose that an interplay of conceptual and usage
factors can explain why the usage of Dutch causal connectives does not always
conform to abstract detinitions that seem to he quite straightforward otherwise.
Our hypothesis is that the apparent"counterexamples" are actually non-typical,
or: peripheral, members (referred to in this study as "non-prototypical usages":
NPU) ofthe very same conceptual category the connectives refer to in their more
typical usage contexts (referred to as "prototypical usages": PU). In this paper,
we focus on one specific factor c{JJJSing variation in connective use, namely: the
idea that language users categorize causal relations not on the basis of" objective
reality", but on the basis of their subjective construal of the situation (cf. Lan-
Categories ofsubjectivity in Dutch causal connectives: a usage-based analysis
121
gacker 1987; Verhagen 2000), and on one specific factor constraining variation,
namely: the prototypicality structure of semantic categories discussed above.
The approach proposed here has several advantages: not only can it adequately
describe the flexibility connectives show in language use, it can also explain their
flexibility a,q a consequence oftheir categori7..ation function and as a consequence
of more general cogoitive mechanisms governing natural language use.
In section2, we discuss previous analyses ofDotch causal connectives, focusing on the usage types considered "problematic" for the categorization analysis.
In section 3, we discuss theoretical assumptions underlying the usage-based perspective on language use that are relevant to our analys.is of causal connectives
in terms of prototypicality structure. Section 4 presents results of an analysis of
PU and NPU of daardoor, daarom and dus in a corpus of newspaper texts.
2. Patterns in cansal connective use
The meaning and use of causal connectives have been deserihed with reference
to the cognitively basic concepts of subjectivity and volitionality. The importance of the concept subjectivity in determining linguistic phenomena is widely
atte;ied (see for example Traugott 1989, 1995; Langacker 1990; Lyons 1995;
cf. discnssion in Sanders, Sanders and· Sweetser this volume). Notwithstanding
terminological differences (Pit 2003; De Smet and Verstraete 2006), theorists
agree that subjectivity is to a great exteut equivalent to "speaker involvement".
We detine "speaker involvement" as: referring to the degree to which the present
speaker is involved in the construal of the causal relation. 2 A causal relation is
"subjective" if for itq interpretation reference to the speaker is needed Conversely, a linguistic element counts as "objective" if speaker involvement is
absent in the interpretation of the causal relation. The results of the studies mentioned above suggest that prototypiCal usage contexts of daard<Jor, daarom and
dus can he related to different categories of"objective" and "subjective causality" detined in terms of"speaker's roles": the speaker acts as a "concluder", as
an volitional agent, or is absent Typical corpus examples of each category are
(6)-(8) 3 :
2
This characterization is based on the definition presented by Pander Maat & Degand
(2001: 214). The original definition is: "the degree to which the present speaker is
implicitly involved. .. " Since SI can also be signaled explictly (with modal elements,
perspective markers, etc.), we did not include the element "implicit".
3
Examples discussed are ( ~mless stated otherwise) fragments taken from the Dutch
daily newspaper Trouw (year of publication: 200 l ), part ofthe corpus analyLed by
Stukker (2005).
122
(6)
Ninke Stukker, Ted Sanders andArie ferhagen
De Boeing 747, het duurste vliegtuig dat rondvliegt, daalt steeds sneller
in waarde. De afgelopen jaren is het vermogen van vliegmaatschappijen
dllardoor met vele miljarden dollars verminderd.
'The Boeing 747, the most expensive plane in the air, is continuing to
diminish in value rapidly. Daardoor airlines' capital has decreased by
billions of dollars over the last few years.4•
(7)
(ln Denmark and in the Netherlands, carcasses of cows older than 30
months are tested for the cattle plague BSE). Andere Ianden zijn nag
niet klaar om elk voor de s/acht aangeboden rund te onderzoeken. Zij
vemietigen daarom op grote schaal dieren.
'Other countries are not yet ready for testing any individual hovine destined for consumption. Daarom they destroy animals on a large scale,'
(8)
(Dutch soldiers who served in Bosuia relate the high incidence of
leukemia among them to frequent exposure to impoverished uranium),
Maar de huidige hypothese wijt de leukemieiin aan een virus(. . . ). Het is
dus denkhaar dat de soldllten die nu leukemie hebben gekregen, slachtof
for zijn van iets anders dan verarmd uranium.
'But the current hypothesis attributes the leukemias to a vims. ( ... ) Dus
it is conceivable that the soldiers who suffer from leukemia now, are
victims of something else than impoverished uranium.'
The differences hetwcen the causal relations represented above can he described as follows: Dus is typically used for marking epistemic causal relations5,
in which a causal relation is constructed hetween a conclusion of the speaker
presented as the causal effect ("it is conceivable that. .• " in [8]), and an argument fimctioning as the causal antecedent (the faet that the current hypothesis
attributes leukemia to a virus). In epi'>lemic causal relations, the speaker functions as the source of the causal relation (it is he who relates argument and
concinsjon). In other words: reference to the speaker is obligatory in order to
interpret the causal relation correctly, hence the causal relation is subjective.
4
5
We focus on lhe causal relations mainly from a conceptual perspective. Therefore,
lhe English glosses of our Dutch text material do not contain lireral translations;
neither did we attempt to reproduce specific syntactic characteristics of the Dutch
causal connectives (see for a discussion e.g. Evers-Vermeul, 2005).
The category of subjective causal relations also contains "speech act" relations,
in which the causal relation is constructed between a speech act and a proposition
functioning as a justification ofthat speech act (Sweetser 1990); an example would be
What are yau doing tonight? Because there is a good movie on). Speech act relations
hardly occurred in our corpus; therefore, they will not be taken into acconnt in this
paper.
Categories ofsubjectivity in Dutch causal connectives; a usage-based analysis
123
This type of speaker involvement is lacking in causal relatious typically
marked with daardoor and daarom. These types of relations can he seen as
inherenrly "objective".6 Fragments (6) and (7) both descrihe causal relations
hetween states of affairs in the observable world, having their source outside the
speaker. Yet, they differ Vlith respect to the concept of volitionality. The causal
relation in (7) describes an volitional act which is motivated by the situation
described in the first segment Volitionality is ahsent in fragment (6) where one
physical process induces another one, without intervention of a human heing. In
Dutch, volitional causality is typically expressed with daarom; non-volitional
causality is typically expressed with daanloor. The distinction hetween volitional and non-volitional causality is closely related to a distinction exhibited
by Dutch causal auxiliary verbs, causality markers operating at the clause level
of the linguistic structure (Stukker, Sanders and Verhagen 2008). The causal
verbs doen and Iaten distinghuish hetween physical, "inanimate" causal processes and causal processes in wbich animate heings are involved. Just like the
distinction objective-subjective, the distinction volitional-non-volitional is considered to he a cognitively important one; it reflects the conceptual model of
"Naive Dualism" (Verhagen and Kemmer 1997; see also D' Anrade 1987).
"Typicality", or "prototypicality" is defined in this study in terms of usage
frequency. The more frequent a usage-type occurs in natural language use, the
more prototypical it is taken to be (see section 3).ln the remninder of this paper, the connectives' prototypical usage contexts are referred to as "PU". As
we already noted in section 1, in a minority of cases, usage-contexts of daardoor, damum and dus do not conform to the patterns described above. Under
specific circumstances, co1111ectives are used in contexts which are taken to helong to other co1111ectives' causality categories. These non-prototypical usage
types (which we will call "NPU'') oeeur with each of the connectives under
investigation, in all of the causality categories discussed. By way of illustration, we discuss the use of dus in volitional causal contexts, assumed to he the
prototypical contexts of use for daarom. Consider (9) and (10):
6
Approaches disagree on the question whether volitional causality should count as inherently subjective (see for example Pander Maat & Degand 2001 ), or as inherently
objective (Stukker 2005), In this paper we assume that subjectivity is not an intrinsic
characteristic of volitional causality; the fact thal some 4 'subject of consciousness"
necessarily plays a role opens the way to introduce elements of subjectivity in such
wntexts. The souree of subjectivity is then transferred from the speaker to an·
other "subject of consciousness" ("Perspectivization": J. Sanders & Spooren 1997;
specifically with reference to causal relations, see Sanders, Sanders & Sweetser this
volume).
124
(9)
Ninke Stuklrer, Ted Sanders and Arie W?rhagen
(Bystanders ntsh to help outatt.'le Volendam pub fire.) "Ikwoon vlakbij,
dus ik ben brandwondencreme gaan halen."
"'I live nearby dus I ran to get burn ointment.'"
(10)
(Letter to the editor discussing the consequences oflowering the age
limit for child adoption in the Netherlanda.) Het is allemaal heel goed
te begrijpen dat de realistische adoptieouders, na de wachttijd van vele
jaren, hun kindje zo snel mogelijk willen hebben. Dus kiezen ze voor
een kintije uit China of een ander "snel" land.
'It is only natural that realistic adoptive parents, atler having waited for
many years, want to have their child as soon as possible. Dus they opt
for a child from China or from another "quick" country.'
NPU of dus and daarom exhibit systematic patterning. Pander Maat and
Sanders (2000) and Pander Maat and Degand (200 I) observed that volitional·
causal relations marked with dus show a higher degree of subjectivity than
those marked with daarom. Tbe SOC of dus-marked contexts referentially coincide with the speaker (see [9]) more often than the daarom-marked contexts
do (see for example [7]; Pander Maat and Sanders 2000: 73-74; Pander Maat
and Degand 2001: 239-240). In addition, volitional contexts marked with dus or
daarom differ in terms of perspective configuration (Pander Maat and Sanders
2000: 71-3; for backward causal connectives, see Pit 2003; see also the chapter
by Sanders et al. in the present volume). Volitional causal relations marked with
dus more otlen contain "continuous speaker perspective" than those marked
with daarom. Both (9) and (10) contain continnous speaker perspective. In the
former, the speaker is continuously referred to by "I". In the latter fragment, the·
causal relation is reported from the perspective of "they", who functions as an
embedded speaker, from whose perspective the causal relation is reported. Lin~
guistic signals construing this interpretation are the occurrence of perspective
markers in the cattse-segment (mental state verb "want" and evaluative element "as soon as possible", both grannnatically attJ.ibuted to the prinaary SOC
"they" 7). Since this actor perspective is not explicitly "blocked" in the second
segment, we assume that the perspective is continued. In fragment (7), which
is representative of volitional causal relations marked with daarom, explicit
indications of perspectivization are lacking.
In the remainder ofthis paper, \ve Vlrill argue thatNPU ofa specific connective
are adequately analyzed with reference to the causal category the connective is
taken to refer to in its PU. As we will argue, a recurring pattern seems to be that
7
The tllctthat in this specific fragment the perspective of"tbey" is embedded underthe
speaker's own evaluation does not make a difference for the analysis presented here.
Categories ofsubjectivity in Dutch causal connectives: a usage-based analysis
125
non-typically used connectives seem to import specific elements of their PU into
the foreign context, creating ambiguity in terms of causality category. In order
to characterize the contrast between volitional causal relations marked with dus
rather than with daarom, for example, reference to dus' more typical context
of use (epistemic cattsality) is essential, no matter how subtle these differences
are, or with what linguistic indications "speaker involvement" is construed.
The inverse pattern seems to occur in epistemic causal relations marked with
daarom. These contexts systematically contain less subjective elements than
those marked with dus (Pander Maat and Sanders 2000; Pander Maat and Degand
200 I), which cao be interpreted as an indication that daarom 's epistemic NPU
are in tact conceptually related to its PU of descriptive (content, objective)
volitional causality. Previous studies did not analyze NPU within the objective
causal domain (of daardoor and daarom), but we will see below that a similar
pattern holds in these cases.
3. A usage-based interpretation of patterns in causal
connective use
Before we start analyzing the usage of daardoor, daarom and dus in terms of
PU and NPU (see section 4), a crucial question is: How is the occurrence of
NPU to be accounted for at all?
We propose to do so with reference to general mechanisms of langusge use.
'The relation between linguistic knowledge and langusge nse has been studied
within the "usage-based approach" to langusge (Langacker 1987, 2000; Bybee
1985, 2006; for an overview sec Barlow and Kemmer2000 and contributions to
that volume). The usage-based approach assumes that variation is an inherent
characteristic oflangusge use, and it seeks to explain occurring patterns of variation and stability with reference to more general coguitive mechanisms. In this
section, we discuss a number of usage mechanisms which, we believe, mediate
the relation between the mental representation of the conventional meaning and
function of daardoor, daamm and dus, and their occurrence in language use.
The first issue \ve w.mt to address is: Under the assumption that the meaning
of daardoor, daarom and dus can adequately be described with reference to
clearly delineated, well-defined conceptual categories of causality- how can we
explain the variation occurring in contexts of natorallanguage use? We specifically focus on "subjective coru;trual" as a mechanism tJnderlying linguistic
categorization (3.1) Tite seeond issue is: in view of this variation, how can we
maintain the idea that daardoor, daarom and dus bave a constant meaning? We
focns on prototypicality structure as a mechanism allowing for flexibility while
126
Ninke Stukl<er, Ted Sanders and Arie Verhagen
maintaining the stability required for language to function as a more or less
stable system of conventional symbols (3.2). Our claim will be that these mechanisms in combination explain the pattems of variation and stability observed
in daardoor, daarom and dus, starting from the categorization hypothesis.
3.1.
Cognitive factors producing variation in language use
How can we explain that lhe meaning of connectives sometimes is, and sometimes is not completely congruent with overt signals in the linguistic contexts
they are used in? The incongruence observed in Section 2 seems to be of a
specific type. The NPU discussed are ambiguous with respect to causality type;
they contain characteristics congruent with the causality type lhe connective
is hypothesized to belong to, and at lhe same time lhey contain characteristics
incongruent wilh it, but congruent wilh a difforent causality type. We propose
to analyze NPU as cases of "subjective construal" of the causality category. It
means lhat the same causal relation in reality may be categorized differently by
different speakers, according to their particular understanding of the sitnation
or lheir rhetorical purposes. This proposal is in line wilh lhe suggestion put forward in several studies of linguistic categorization that an expression's meaning
is not just an objective characterization of lhe sitnation described. Equally important for linguistic semantics is how lhe speaker chooses to "construe" the
sitnation and portray it for expressive purposes (Langacker [ 1990] 2002: 315,
1987; Verhagen 2007; and references cited !here). Subjective construals of categorization in terms of causality were found with causal verbs, marking canial
relations at lhe clause level (Verhagen 2000; see also Verhagen 1997; Verhagen
and Kemmer 1997). Thus, we can analyze lhe NPU of dus in volitional causal
contexts, discussed in section 2 and repeated here, as follows.
( 11)
(Bystanders rush to help out at the Volendam pub fire.) "lkwoon vtakhij,
dus ik ben brandwondencreme gaan halen."
'"1 live nearby dus I ran to get burn oinlment."'
(12)
(Letter to the editor discussing lhe consequences of lowering u'le age
limit for child adoption in lhe Nelherlands.) Het is allemaal heel goed
te begrijpen dllt de realistische adoptieouders, na de wachttijd van vele
jaren, hun kindje zo snel mogelijk willen hebben. Dus kiezen ze voor
een kindje uit China of een ander "sne/" land
'It is only natoral that realistic adoptive parents, after having waited for
many years, w.mt to have their child as soon as possible. Dus they opt
for a child from China or from another "quick" country.'
Categories of subjectivity in Dutch causal connectives: a usage-based analysis
127
Volitional causal relations ·are inherently objective. They refer to processes
which occur in observable reality. This is indicated in the contexts of (II) and
(12) by the assertive speech acts, of vvhich the truth is not questioned. At the
same time, the related segments of (11) and (12) contain linguistic elements
construing subjectivity (speaker reference in fll], perspectivizing devices in
[ 12]; see section 2), leaving the context ofuse as a whole ambiguous for causality
type. The overall interpretation of the causal relation seerns to be disambiguated
by the speaker's choice for a specific connective. Marking with dus in fragments
(11) and (12) indicates that lhe reader is expected to construe the causal process
as a case of subjective causality, despite its inherently objective character.
The sn~eetive construal analysis of causal connectives proposed here is in
line wilh lhe assumption lhat lhe mental representation of a linguistic utterance
(in a natural context) is only partly based on lhe overt signals it contains. This
assumption is accepted in various branches of linguistic study. The representntion a language user builds from a given utterance is, apart from lhe linguistic
signals, a product of lhe previous discourse, hackgrom1d knowledge and inferencing (Sanders and Spooren 2001: 3; for more elaborate discussions see
Fanconnier 1984; Sperber and Wilson 1995; Langacker 1987, 2000; Verhagen
1997, 2000; Radden et al. 2007).11ris inherent context dependency and underspecification of language intplies that individual usage contexts of a linguistic
element may vary to lhe extent lhat lheir characteristics conform to lhe more
abstract semantic representation associated wilh lhis element in language user's
long term memory (Verhagen 1997).
3.2.
Cognitive factors retaining stability in language representation
The mechanisms discussed in Section 3 .I explain why variation may occur in
lhe way abstract semantic knowledge (in our case: concerning causal connectives) is mirrored in actual language use. But in view oflhis variation, how can
we maintain the idea that daardoor, daarom and dus have a eoustant meaning?
Or put in more general terms: how is it possible lhat language users are able to
interpret conceptual models invoked by linguistic elements in a consistent way?
A common asswnption in functionally oriented branches of linguistic theory
is that an individnal's linguistic system is fundamentally grounded in "usage
events". This assumption has been elaberated notably wilhin what has become
known as the "usage-based approach to language" (Langacker 1987, 2000; Bybee 1985, 2006; Goldberg 2005; for an overview see Barlow andKemmer2000
and contributions to that volume). According to this approach, lhe language
system does not consist of rules generating grammatical instances of language
128
Ninke Stukke1; Ted Sanders andArie Verhagen
use, but rather of generalizations over individual nsage events, which in tum
categorize or license other usage events. In other words, according tu the usagebased approach to laugua1,-e, "granuuar is the cognitive organization of one's
experience with language" (Bybee 2006: 711 ).
Thus it is assumed that grammar ultimately emerges from individual USage events. This process is often compared to the more general psychological
process that the occurrence of any (psychological) event leaves some kind of
trace that facilitates their reoccurrence. Frequently eocountered patterns become
"entreoched" in memory; acquire the statos of cognitive routine that is retrieved
and applied without requiring conscious attention. Another process, which oc•
curs parallel to entrenchment, is that of abstraction: the emergeoce of a structure
through reinforcement of the commonality inherent in multiple experiences that
difter .in some other way (Langacker 2000: 4). In this process, peculiarities of
individual instances are filtered out from the representation the language nser
has of the entrenched unit This process is reflected in language acquisition patterns of childreo (Barlow and Kemmer 2000: xii, see references cited). In the
process of abstraction, the case of"schematization" is of particular importance
for our porposes. Langacker (2000: 4) defines a "schema" as the commonality
that emerges from distinct structores when one abstracts a'ovay from their points
of difference by portraying them with lesser precision and specificity.
TI1e phenomenon of schematization lies at the basis of language users' ability
to recognize a variety of specific iustances ofuse as an iustantiation of a specific
entrenched strllcture. This is the cognitive ability of categorization: recognize
similarities and differences between phenomena. Or from the perspective of our
problem: recognize a specific instance of use as a member of an entrenched
schema. A number of factors seem to play a role in thLq process, among whicl)
contextual priming (cf. discussion of the context dependency of language in
Section 3.1) and the amount of overlap between the "target" and the potential
categorizing structure (Langacker 2000: IS~ 17; for a more elaborate discussion
see e.g. MacWhinoey 2000).
Categories that emerge from linguistic experience exhibit prototype effects.
"Prototypicality" refers to the well-known cognitive phenomenon that conceptual categories are not homogenous; some members of a category are better examples than others. Oranges, apples and bauanas are better examples of FRUIT
than nuts and olives, which are nevertheless recognized as members of the same
conceptual category (Rosch 1973) 8 • TI1e "best examples" are the category's pro8
The participants in Rosch' experiment were US college students. It is highly likely
that prototypicality effects wry with regional or cultural factors (e.g, Lipka 1987;
cf. discussion in Ungerer & Schmid J 996: 49-52).
Categories ofsubjectivity in Dutch causal connectives: a usage-based analysis
129
totypical members; othermembtlrs vary to the degree to wl:ticll they have features
in common vvith the prototype (cf. Wittgenstein 1953; Rosch 1973; Rosch and
Mervis 1975). The usage-based conception oflanguage predicts that the more
frequent, or: the more entrenched, contexts of use are the more prototypical
members of a category, while the less frequently encountered usage contexts
are more "peripheral" members belonging to the same category, related to the
prototype by way of conceptual affu:tity. 9
We started this section with the question: How can we explain that language
users maintain a more or less stable representation of the meaning of causal
connectives, in view of the variation in usage-contexts encountered? Referriug
to the usage-based theory oflanguage, \\'e may answer this quel>iion as follows.
On the basis of the Iiuguistic and non-linguistic contexts of use, language users
will be able to recognize the PU and NPU of causal connectives as members of
one and the same category. We hypothesize that NPU of causal connectives are
motivated by the phenomenon of subjective construal based on the conceptual
model of sul:>jectivity. Findings reported in previous studies that seem to favor
our hypothesis, are: the fact that the c.onnectives' PU are relatively frequent, and
observations on the conuectives' NPU reported in previous studies (and more
particularly: our interpretation of them, see Section 2).
4. Corpus analysis
The usage-based perspective discussed in the previous section yields specific
predictiorn concerning the patterns manifested within the variation of comlectives' contexts of use. In this section, we report a corpus analysis in which we
investigate to what extent patterns of use found with daardom; daarorn and dus
can he accounted for by the categorizatiou hypothesis, under the assumptiou
that the "appearance" of the causality categories in language use is mediated by
9
Empirical evidence in favor of this effect of "token ftequency" is found for example in the phenomena of phonetic reduetion of high frequency words and phrases
(cf. Bybee 2006, 2007), the "conserving effeet'' ~the finding that high-frequency
sequences become more entrenched in their morphosyntaclic structure (e.g. Bybee
1985), and the "autonomy effecf' ~ tl1e fact that morphologically complex forms of
high frequency can lose their internal structure as they become autonomous from
etymologically related forms (Bybee 1985 ···· see discussion of these phenomena in
Bybee 2006; see for discussion of other types of frequency effects Hasher 1984;
Geeraerts, Grcndelaers & Bakema 1994; Verhagen 2000; Bybee 1985, 2007; Goldberg 2005; Schmid 2000).
130
Ninke Stukker, Ted Sanders and Arie fer hagen
prototypicality effects. In Sections 4.1 and 4.2 we present our hypotheses and
methods of analysis. Section 4.3 presents the resnlts of our corpus analysis.
4.1.
Hypothesis 1: Frequency of use as an indication for
prototypicality of PU
Building on the "frequency of use reflects the cognitive entrenchment" assumption (section 3.2), our first- quantitative- hypothesis is: the connectives daardoor, daarom and dus are significantly more frequently used in usage types
conforming to the categorization hypothesis (their PU).
Hypothesis 1:
Daardoor is prototypically used in content non-volitional relations;
daarom is prototypically used in content volitional relations; dus is prototypically used in epistemic relations.
Notice that this hypothesis reflects the "semasiological" perspective on word
meaning. Given that a lexical items couple word forms to semantic contents,
the semasiological question is: "Given linguistic item X, what meaning does it
express?" This perspective is complemented by the "onomasiologial" perspective on word meaning. The onomasiological question is: "Given concept Y, what
linguistic item(s) can it be expressed with?" (Geeraerts 1997: 17). Although
our present research question is primarily of a semasiological nature (PU ai\d
NPU are inherently semasiological concepts), both of the levels of analysis are
actually relevant for understanding how the meaning and use of causal connectives relate to conceptual structure. In the final analysis, we need to know
what the connectives "mean" (denote in terms of conceptual reference - the
semasiological perspective) and we need to know how (with what element or
elements) an articnlate conceptual category can be expressed linguistically (the
onomasiological perspective). We will return to this issue in the discussion.
4.1.1.
Method and operationalization
In line with the usage-based assumptions presented in section 3, a usage context's degree ofprototypicality is operationalized in terms of usage-frequency:
the more frequent a specific context is, the more prototypical it is taken to be.
The categories of causal relations mentioned in the hypotheses will be determined making use of the "basic operation paraphrase test" of Sanders ( 1997; cf.
Knott and Dale 1994). The paraphrase test makes it possible to determine the
Categories of subjectivity in Dutch causal connectives: a usage-based analysis
131
relational meaning with elimination the connective's contribution to the overall
interpretation of the relation. 10
The categories are operationalized as "paraphrases" (see Figure 1), explicating the relational meaning they add to the connected segments. In doing so,
the type of causal relation that relates the segments can be established in an
objectified and reliable way. The paraphrases are adapted from Pander Maat and
Sanders, 1995 (cf. Evers-Vermenl and Stukker 2003). The paraphrases represent the connectives' hypothesized prototypical usage schemas on two aspects
which have proven to be crucial for determining causality category (cf. Section
2): SOC type (implicit speaker SOC, explicit [speaker or actor] SOC, no SOC)
and event type in the consequent-segment (conclusion, volitional action, nonvolitional situation). "S 1" refers to the segment containing the causal antecedent.
The paraphrase itself refers to segment "S2", containing the causal effect.
Relation category
Paraphrase
Content non-volitional De situatie in S I leidt tot de volgende situatie: S2
The situation in Sl leads to the following situation: S2
Content volitional
De situatie inS 1 is een motivatie voor de volgende handeling:
sz'"
Epistemic
The situation in SJ is a motivation for the following action:
S2
De situatie in S 1 is een argument voor de volgende conclusie:
S2
The situation in SJ is an argument for the following conclusian: S2
Figure 1. Paraphrases of categories of causal relations marked by daarom, daardoor
anddus.
The paraphrase test consists of three steps, which we will illustrate analyzing
(13), a PU context of daarom, already discussed as fragment (7) and repeated
here:
{13)
(In Denmark and in the Netherlands, carcasses of cows older than 30
months are tested for the cattle plague BSE). Andere Ianden zijn n0g
niet klaar om elk voor de slacht aangeboden rund te onderzoeken. Zij
vernietigen daarom op grote schaal dieren.
10 Our analysis focuses on lexical elements (and sometimes contextual elements). Constructional aspects are not taken into account. For hypotheses concerning constructional differences entailing differences in conceptual interpretation, see e.g. EversVermeul (2005).
I 0 With S2 understood to be actually performed.
132
Ninke Stukker. Ted Sanders and Arie V,rhagen
'Other countries are not yet ready for testing any individual bovine
destined for consumption. Daarom they destroy animals on a large
scale.'
Step 1: Determine the text segments that are related by the connective.
[Andere Ianden zijn nag niet klaar om elk voor de slacht aangeboden rund te onderzoeken.}Sl [Zij vernietigen daarom op grate schaal
dieren.}S2
[Other countries are not yet ready for testing any individual bovine destined for consumption.]Sl Daarom [they destroy animals on a large
scale]S2
Step 2: Remove the connective marking the causal relation.
/Andere Ianden zijn nog niet klaar om elk voor de slacht aangebode!!
rund te onderzoeken.}Sl [Zij vernietlgen op grote schaal dieren.}S2
[Other countries are not yet ready for testing any individual bovine
destined for consnmption.]S 1 [They destroy animals on a large scale] S2
Step 3: Insert the paraphrases and determine which one fits the context uoder
consideration best. 12
{Andere Ianden zijn nag niet klaar om elk voor de slacht aangeboden
rund te onderzoeken.]Sl
[Other countries are not yet ready for testing any individnal bovine
destined for consumption.]Sl
# 13 The situation in Slleads to the following situation:
The situation in S 1 is a motivation for the following action:
#The situation in S 1 is an argument for the following conclusion:
[Zij vernietigen op grate schaal dieren.]S2
[They destroy animals on a large scale]S2
Since the connective's contribution to the interpretation of the relation as a
whole is eliminated in this procedure (•iep 2), "PU-hood" or "NPU-hood" of a
given context can he established as follows. If the paraphrase category coincides
with the connective's hypothesized prototypical context of use, the fragment at
hand is classified as PU. This is the case with (13), and also with (15) below,
which is a typical PU ofdus (discussed in section2 as [8]). In fragment (14}, on
12 In orderto construct a maximally natural text, tense and aspect ofthe paraphrase may
be adapted to the fragment. If necessary for a adequate interpretation, propositions
may he reconstructed (write ont referential expressions, integrate relevant implicit
information that can be deduced from the context).
13 The # symbol indicates !hat the suggested relation leads to an incoherent interpretation,
Categories ofsubjectivity in Dutch causal connectives: a usage-based ana~ysis
133
the other hand (discussed earlier as [9]), the paraphrase fitting bst (of volitional
causality) does not coincide with the hypothesized PU ofthe marking connective
dus. Hence, ( 14) is classified as NPU.
(14)
{"lk woon vlakbij, ]Sl
"['I live nearby;] Sl
# T11e situation in S t leads to the following situation:
Tire situation in S 1 is a motivation for the following action:
# The situation in S I is an argmnent for the following conclusion
[lk ben brandwondencreme gaan halen." }S2
['I ran to get burn ointment.']S2"
(15)
{Maar de huidige hypothese wijt de leukemieen aan een virus.JSI
['But the current hypothesis attributes the leukemias to a virus.']S 1
#The situation in S !leads to the following situation:
#The situation in St is a motivation for the following action
The situation in S 1 is an argument for the following conclusion
{Het is denkbaar dat de so/daten die nu leukemie hebben gekregen,
slachtoffer zijn van iets anders dan verarmd uranium.]S2
['It is conceivable that the soldiers who suffer from leukemia now, are
victims of something else than impoverished uranium.']S2
4.1.2.
Sample and procedure
The frequency of use hypothesis is tested against a corpus of newspaper texts,
taken from an electronic version of the Dutch national newspaper Trouw, from
the year 2001. 14 In order to control for possible genre-effects ( c£ Sanders 1997),
the sample was built from genres belonging to different text types. For a more
detailed description of the sample, seeAppendix I. Fromeachofthe sub corpora, 50 occurrences per connective were selected. 15 Only forms functioning as
markers of causal coherence relations were included, 16 and only relations holding between segments of minimally one clause were included in the sample.
Occurrences were included in the sample in chronological order. The newspa-
14 Available from "Krantenbank" Factlane, (Lexis Nexis Nederland bv), a service that
provides electronic access to the archives of a number of Dutch daily newspapers.
15 In order to minimize risks ofbia'i according to individual authors • styles and topic,
maximally two occurrences per article were included in the corpus.
16 For an overview of other usage-types of daarom and dus, see Evers-Vermeul &
Stukker, 2003; Evers-Vermeul, 2005.
134
Ninke Stukker, Ted Sanders andArie TJerhagen
pers were analyzed "issue by issue" (day by day). 17 The sample was analyzed
quantitatively making use of the statistical method of contrast analysis (see
Appendix 2).
Categories ofsubjectivity in Dutch causal connectives: a usage-based analysis
usage types of a given connective. Therefore, our analysis consists of identifying
general tendencies in the usage of the connectives under investigation.
4.2.2.
4.2.
Hypothesis 2: Conceptual affinity ofNPU to the prototype as an
indication of category membership
Our second hypothesis concerns the connectives' NPU, which are expected to
occur less frequently (see hypothesis 1). The degree of (non-)typicality of usage
contexts is operationalized quantitatively in terms of relative frequency of use
(see hypothesis 1). Non-typical contexts can thus be defined as "contexts of use,
defined in terms of causality category, occurring significantly less frequently
than others". On the basis of the assumptions discussed in section 3, we expect
that NPU are adequately analyzed as non-typical members of the same conceptual category the connective's PU belong to. Our general expectations are that
a) connectives' NPU can be analyzed as subjective coustruals of categorization,
and b) the overall interpretation of NPU shows conceptual affinity to the resp.
connectives' PU. The following hypotheses, specified per connective, are tested:
Hypothesis 2:
Given a connective a which is prototypically used in category A, and
which can non-prototypically be used in category B (C, D), the nonprototypical usage of a in B (C, D) is characterized by elements 'of
category A more often than when category B (C, D) is marked with
connective b (c, d).
4.2.1.
Method and operationalization
Neither the speaker's subjective interpretation of a situation nor his rhetorical
purposes are directly accessible for analysis. Therefore, subjective construals
are reconstructed on the basis of the connectives' (linguistic or non-linguistic)
contexts of use. It can only be hypothesized that non-typical usage types show
conceptual affinity to the more typical usage types of a given connective; it cannot be known in advance what kind(s) of connection will occur; given the role of
individual creativity it seems to be impossible in principle to predict all possible
17 Therefore, the distribution over text genres mirrors composition of the complete
newspaper, rather than that proportional selection of all of the genres was strived at.
135
Sample and procedure
The primary causality category of an NPU was determined on the basis of the
linguistic context, making use of the paraphrase test described in Section 4.1;
this step is performed when testing hypothesis 1. The analysis testing hypothesis
2, then, consists oflooking for demonstrable signs from the linguistic or extralinguistic contexts of use - either within the segments connected in a causal
relation, or in the broader context - which motivate the categorization of the
specific causal relation as the causality type associated with the connective
chosen.
4.3.
Results
Our data largely reflect patterns found in previous corpus studies of daardoor,
daarom and dus (Pander Maat and Sanders 1995, 2000; Pander Maat and Degand
2001).
4.3.1.
Hypothesis 1: Usage frequency ofconnectives'PU
Our first hypothesis is corroborated by the connectives in our sample (Z = 4 .30;
p < .001; for a more detailed discussion of the statistical analysis see Appendix I). Figure 2 presents an overview of the relation between connectives
and causality categories in the corpus from a semasiological perspective (cf.
section 4.1 ). When read vertically, it answers the question: "Given connective
a (b, c), what category of causality does it typically express?". Our paraphrase
method revealed that the connectives' PU contexts may )>e linguistically realized with different degrees of explicitness with respect to the categories they
represent. In this section we discuss the most important of the PU patterns we
encountered in our corpus.
Daardoor
In accordance with our PU hypothesis, we found that daardoor prototypically
is used in contexts of content non-volitional causality (96/100 cases). Half of
the cases in our sample consist of physical processes, such as fragment (16)
(discussed in section 2 as [8]):
136
Categories ofsubjectivity in Dutch causal connectives: a usage-based analysis
Ninke Stukker, Ted Sanders and Arie Yerhagen
CausaJity categories in conn&etives
Figure 2. The relation between connectives and causality types in our sample.
(16)
17
De Boeing 747, het duurste vliegtuig dat rondvliegt, daalt steeds sneller
in waarde. De afgelopenjaren is hetvermogen van vliegmaatschappijen
daardMr met vele miljarden dollars verminderd.
'The Boeing 747, the most expensive plane in the air, is continuing tO
diminish in value rapidly. Daardoor airlines' capital has decreased by
millions of dollars over the last few years.'
The other balf of the non-volitional daardoor contexts concerned cansal
processes with animate beings as a locus of effect, an example is (17).
(17)
Magje een onderzoek van vorige week geloven, dan speelt in Amsterdam een derde van de kinderen tussen zeven en negen nooit buiten.·
[. .. J Daardoor leren ze niet goed om te gaan met andere kinderen en
dat is weer slecht voor later, als ze voorldurend nieuwe mensen leren
kennen.
'If you are to believe a study from last \veek, a third of the children
in Amsterdam between the ages of seven and nine never play outside.
Daardoor they do not learn to interact with other children: that bas a
negative effect later on when they continue to meet new people.'
17 Two of dus' contexts of use and three of daarom's contexts of use in the sample
were categorized as speech act causality. Because of the extremely low frequency,
these cases were not further taken into account. For this reason, the sums total of
daarom and dus in figure 3 do not add up to I 00. The speech act cases were discarded
137
Despite anima~'}' of the locus of effect, (17) is not a case of volitional causality; the paraphrase "The situation in SJ is a motivation for the following action:
S2" does not adequately reflect the purport of the relation. Since (17) reports
on an investigation conducted by other persons than the writer of the fragment,
categorization as an epistemic causal relation is not very likely either. This is
reflected in the fact that the paraphrase for epistemic causality "The situation
in Sf is an argument for the following conclusion: SZ" does not fit this context
either. The paraphrase for non-volitional causality "The situation in Sl leads
to the following situation: S2 " is the only one fitting this context adequately.
This ean be explained by the fact that "they" cannot be analyzed as a subject of
consciousness. "Learn to interact with other children" is a mental process that
can only take place in an animate being, but occurrence of the process is not
dependent on the experiencer's intentions of doing so.
Daardoor may also occur in non-volitional relations containing an action
predicate in S2, as in fragment (18).
( 18)
De schaatser Frans de Ronde omschreefde Jaap Edenbaan als een grote
kattebak. "Overallag zand. Dao.rt/Qor schaatsten velen met bramen op
hun ijzers. [. .. ]".
'The skater Frans de Ronde defined the Jaap Eden rink as a big kitty
litter bin. ''There was sand everywhere. Daardoor many skaters skated
with scratches on their blades."'
Interestingly, in spite of the presence of the inherently volitional action in
the consequence segment, the paraphrase of volitional causal relations "The
situation in Sl was a motivation for the following action: S2" does not fit very
well. The adequate interpretation of this relation is not that the fact that "sand
was all over the skating rink" led to the volitional action of "skating of many",
but that the situation depicted in S 1 led to the unvolitional situation of "many
skaters skating with scratches on their blades".
Note that the deviant instantiations ofdaardoor's PU non-volitional causality
we discuss here, and other connectives' PU we discuss in this section, are not to be
classified as NPU. The reconstructions indicate that classification in accordance
with the connective's PU causality type yields the most plausible interpretation.
Daarom
the present psper for reasons of space. See for a discussion of speech act causality
As noted in previous studies (cf: Pander Maat and Sanders 2000; Pander Maat
and Degand 2001), daarom is the most "generalist" of dte three connectives
analyzed in this study. It is the only connective that is conventionally used in
Sanders et al., this volwne).
all of the categories disfinguished Still, in accordance 1Nith our PU hypothesis,
from statistical analysis. We did not discuss the category of speech act causality in
138
Ninke Stukker, Ted Sanders and Arie Verhagen
daarom has a statistically significant preference for volitional causal relations
(50/100). The majority of these contexts contain an overtly expressed action iu
S2. An example of this pattern is (19), discussed in section 2 as (9).
(19)
(In Denmark and in the Netherlands, carcasses of cows older than 30
months are tested for the cattle plague BSE). Andere Ianden zijn nog
niet klaar om elk voor de slacht aangeboden rnnd te onderzoeken. Zij
vemietigen daarom op grate schaal dieren.
'Other countries are not yet ready for testing any individual bovine
destined for consumption. Daarom they destroy animals on a large
scale.'
However, volitionality of the causal relation need not be explicitly specified
with an action predicate. The paraphrase test identified volitional causal relations
without overtly expressed actor SOC's. As an example, consider (20), in which
the inherent volitionality ofthe causal relation is "hidden" in an agentless passive
construction.
(20)
Toen Napoleon ook Holland in bezit kreeg, was de vaart ineens niet
meer nodig. De Noordervaart is daarom nooit verder gegraven dan
tot Beringe.
'Once Napoleon controlled Holland as well, the waterway was no longer
necessary. Daarom the Great Northern Canal was never dug any further
than Beringe.'
Some contexts in our sample even contained state predicates which appeared
to be relevant in the interpretation of the relation as "result of an action". These·
cases were analyzed as volitional causal relations. An example is (21 ).
(21)
Microsoft-oprichter Bill Gates waagde zichzelf als eerste aan het spel
tijdens een computerbeurs in gokparadijs Las Jlegas. "Wlj wilden iets
bouwen dat een doorbraak voor computerfreaks is '', omschrijft de ·
voormalige topman Microsofts ambities. Het zwart gekleurde apparaat
heeft daarom vier aansluitingen voor besturingsknuppels meegekregen, waar de gangbare "gameboys " er maar twee hebben.
'Microsoft's founder Bill Gates was the first to try the game during a
computer conference in gambler's paradise Las Vegas. "We wanted to
build something that would be a breakthrough for computer freaks"
explains Microsoft's ex- top executive. Daarom the black colored machine has been provided with four conoections for game pads, while
the current "gameboys" have ouly two.'
Categories ofsubjectivity in Dutch causal connectives: a usage-based analysis
139
The perfect heeft meegekregen in (21) (literally: 'has received') suggests an
act of transfer, including an volitional actor (hence the translation 'has been
provided with'). Thus both in (20) and in (21), the paraphrase for volitional
causal relations, "The situation in SJ was a motivation for the following action:
S2 " is the only one that fits these contexts adequately.
Dus
Dus is prototypically used in the category of epistemic causality (761100). This
is in accordance with our PU hypothesis. A typical example of this category is
(22), discussed in section 2 as (10).
(22)
(Dutch soldiers who served in Bosnia relate the higb incidence of
leukemia among them to frequent exposure to impoverished uranium).
Maar de huidige hypothese wijt de leukemieiin aan een virus ( .. ).
Het is dus denkbaar dat de so/daten die nu leukemie hebben gekregen,
slachtoffer zi}n van iets anders dan verarmd uranium.
'But the current hypothesis attributes the leukemias to a virus. (... )
Dus it is conceivable that the soldiers who suffer from leukemia now,
are victims of something else than impoverished urauium.'
Epistemic causal relations are, in a large majority of cases, easily recognized by
way of subjective elements (signs of "self-expression" of a SOC: evaluations,
modal elements, etc., cf. Langacker 1990; J. Sanders and Spooren 1997; Pit
2003). But sometimes, the intended epistemic interpretation is linguistically
"underspecified". Yet, the paraphrase of epistemic causality "The situation in Sl
is an argument for the following conclusion: S2" is the only one that fits these
contexts adequately. An example is (23).
(23)
(A "pavese" is a weapon shield) Twee eeuwen lang, tussen 1300 and
1500, zijn paveses overal in Europa in gebrnik geweest, dus ook in
Nederlandse legers.
'During two centuries, between 1300 and 1500, paveses were used
everywhere in Europe, dus in Dutch armies as well.'
Underspecified epistemic causal relations concern almost without exception instances of "noncausal epistemic relations" in which either the real-world
causality has a different direction than the epistemic one (so-called "abductive"
causality 19 ), or real world causality is not relevant at all, as in the case of (23),
19 An example from our corpus is "Dan moet je eerst uitvinden van welk station je
vertrekt. Rouaan ligt ten westen van Parijs. Dus het Noordstation, waar je in de
140
Ninke Stukker, Ted Sanders and Arie ferhagen
an instance of assumption based reasoning ( cf. Pander Maat and Degand 2001:
221-224). Noncausal epistemic relations can be identified making use of the
paraphrase test.
A non-standard lingnistic construal of an epistemic causal relation is (24).
Judging from form characteristics, the relation between S 1 and S2 can be
interpreted both as a motivation for the (genuine) question "why shouldn't
Braakhekke sing on stage?" (in which case it should have been categorized
as speech act causality, cf. Sweetser 1990; see also Sanders et al. this volume),
and as giving an argument for the conclusion that is formulated as a rhetorical
question "Braakhekke should sing on stage as well", which leads to categorization as an epistemic causal relation. The latter interpretation seems to be
the more adequate one in this context, where the cook justifies his involvement
with theatre. This is corroborated by the fact that the paraphrase for epistemic
causality "The situation in SJ is an argument for the following conclusion: S2"
adequately reflects the purport of the relation in (24).
(24)
(Television chef Braakhekke also appears on stage these days) Is
Braakhekke wellicht een aandachtsjunk, wil de interviewer weten.
"Neeeee ", kaatst de lange uithaal tot over het Leidseplein. Braakhekke
"likt aileen graag aan het theater en zingt altijd onder het koken, dus:
waarom niet op het toneel?"
'The interviewer wonders whether Braakhekke is perhaps an attention junkie?, "Noooo", resounds his answer across the Leidseplein.
Braakhekke "only likes to lick the theatre and sings while he is cooking, dus: Why not on stage?"'
Conclusion from the PU analysis
On the basis of the "frequency reflects entrenchment assumption", then, we
interpret the results of our PU analysis as an indication that the usage contexts
conforming to the categorization hypothesis are more entrenched in the language
user's semantic knowledge than the usage contexts which do not conform to the
categorization hypothesis.
buurt Belgische frieten kunt kopen?" Then first you have to find out from which
station you'll be leaving. Rauen lies west ofParis. "Dus "the North station, in which
surroundings you can buy Belgien fries?
Categories ofsubjectivity in Dutch causal connectives: a usage-based analysis
4.3.2.
141
Hypothesis 2: Conceptual affinity of connectives' NPU to
their PU
Our second hypothesis concerns the non-typical contexts of use (NPU) of daardoor, daarom and dus. We hypothesized (section 4.2) that the connectives'NPU
reflect characteristics of their PU, to such an extent that the NPU can be analyzed
as peripheral members from the same semantic category the connectives' PU
belong to. In this section, we describe NPU patterns we found in our sample.
Per pattern we discuss one representative example. We describe in what way
the NPU is conceptually related to the connective's PU, and also how the NPU
pattern diverges from the same causality category marked with one of the other
cmmectives.
Daardoor
Daardoor is typically used to mark non-volitional causal relations (see section 4.3.1 ). In a small minority of cases in our sample (4/100), it is used to mark
volitional causal relations. An example is (25):
(25)
De Chinezen krijgen dit jaar vee[ meer vrije dagen. De regering hoopt
dat de bevolking daardoor meer spaargeld gaat uitgeven om de groei
van de economie op peil te houden.
'The Chinese get more vacation days this year. The government hopes
that daardoor the populace will spend more of its savings to keep the
economy growing.'
The causal relation in (25) is constructed between the first sentence the
complement clause of the second sentence. The fragment is taken from an article
which discusses a characteristic of the economic climate in China, namely that
the Chinese people tend to save all their money instead of spending it and thus
stimulate economic activity. The causal relation in (25) is best explicated with
the paraphrase for voliti<,mal causality, "The situation in SJ is a motivation for
the following action: S2 ". The paraphrase for non-volitional causality, "The
situation in SJ leads to the following situation: S2" doesn't seem to reflect the
purport of (25) adequately.
However, in accordance with our second hypothesis, daardoor-marked volitional causal contexts systematically exhibit conceptual affinity to daardoor's
PU on the aspect of the relation's "locus of effect". They differ from daarom
and dus marked volitional contexts with respect to degree of intentionality of
this locus of effect.
142
Ninke Stukker, Ted Sanders and Arie Verhagen
In the case of (25), the locus of effect "the populace", is ambiguous with
respect to SOC-hood. On the one hand, the predicate "spend" in S2 is inherently
intentional. On these grounds, we would have to interpret (25) as containing an
actor SOC. On the other hand, "spend" is presented as instantiating systematic
patterns. The measure of allotting the people more holidays is prompted by an
(assumed) social law: "holidays are spent shopping" or "the more time off, the
more shopping is done". A linguistic indication in favor of this non-volitional
interpretation is that the causal relation is strongly presented from the perspective
of Chinese government. Although a "normal" intentional causal interpretation
of the situation reported is certainly conceivable on a conceptual level (Chioese
people consider more holidays a reason for going shopping more frequently), the
present wording favors an interpretation from the perspective of the government.
The first indication of this is the embedding of the cause-segment in a matrix
sentence that conveys a mental state of the government. The next indication is
located in the second part of the effect-segment, a goal presented internally in
the sentence, namely, the action of spending. It is improbable that "to keep the
economy growing" specifies the goal that "the populace" has for "spending", as
the construction of the sentence suggests. Most probably, this element should
be understood as a motivation for introducing the measure from the perspective
of the government. This is possibly why the volitional causal paraphrase "The
situation in Sl is a motivation for the follawing action: S2" fits this context
well, while at the same time the intentions of the aetna! intentional agents are
not very relevant in interpreting the situation.
Similar elements indicating "restricted intentionality" can be identified in
the other volitional contexts from daardoor, but they are absent in any of !he
daarom- or dus-marked volitional relations (see table I; cf. the examples of
volitional daarom discussed in section 4.3.1 and the volitional dus example in
[31] discussed below). Although the effect-denoting segments of the daardoormarked cases contain an action predicate and an animate locus of effect, these are
not to be construed as a genuinely intentionally acting "SOC''s (cf. Section 2).
It is highly "unlikely that the loci of effect in these cases consciously interpreted
the situation presented in the cause-segment as a valid reason for performing
the action depicted in the effect-segment; their "intentionality'' seems to be of a
restricted kind. These observations indicate that daardoor-marked cases show a
conceptual relatedness to daardoor's PU, non-volitional causal relations, where
human intentionality does not play a relevant role. 20
20 Note that fragment (25), just like the three other volitional contexts of daardoor,
differ from daardoor-marked fragments with "non-volitional action predicates",
one ofthe patterns instantiating daardoor's PU (cf. [18], discussed in section 4.3.1 ).
In the latter case, ambiguity of the intentional aspect in the relation is solved at
Categories of subjectivity in Dutch causal connectives: a usage-based analysis
143
Table 1. Intentionality in content volitional causal relations
Restricted intentionality
Full intentionality
Total
Daardoor
4
0
4
Daarom
0
50
50
Dus
0
22
22
Total
4
72
76
Daarom
Daarom occurs in NPU contexts more frequently than daardoor and dus do. It
is also the only connective that has NPU in two different causality types, namely
non-volitional causality (16/100) and epistemic causality (311100). We start our
discussion with the latter type, an example is (26).
(26)
(Selecting tall players for volleyball increases chances of international
success) "Nederland heet een langvolk te zijn, maar via de clubs vinden
we de lange talenten niet. Ze !open we[ op straat rand; vaak geftustreerd al vroeg met sport gestopt, omdat hun motoriek tijdens de eerste
puberjaren achterloopt bij die van ldeinere leefiijdsgenootjes. Daarom
moeten wij zelfnaar de scholen gaan om ze te vinden en om ze te overtuigen dat ze juist door volleybal meer eigenwaarde kunnen krijgen."
"'The Dutch are supposed to be a tall people, but we can't find the tall
talents through the clubs. They're walking around on the streets; often
already long frustrated and having given up the sport, because their
motor skills are not that of their smaller peers when they're teenagers.
Daarom we have to go to schools ourselves to find them and convince
them they'd be greatly appreciated in volleyball."'
Fragment (26) contains an intentional action predicate in S2, but this predicate is embedded under a modal verb. Therefore, the paraphrase for volitional
causality, which is daarom's PU, "The situation in Sl is a motivation for the
following action: S2" does not fit this context adequately. The paraphrase for
epistemic causality ''The situation in Sl is an argument for the following conclusion: S2 " by contrast, does. Epistemic causality is distinguished from other
types by its inherent high degree of speaker involvement. The source ofthe causal
the segmental level (see discussion in 4.3.1 ), while ambiguity of intentionality in
volitional relations with daardoor, such as (25), occurs at the relational level.
144
Ninke Stukker, Ted Sanders and Arie Verhagen
relation is by default the speaker2 1
, relating an argument to a conclusion in ongoing discourse. This characteristic is lacking in daarom's PU volitional causality.
Our hypothesis therefore predicts that epistemic causal relations marked with
daarom are relatively less subjective than the dus-marked ones. We expect that
the contexts of use of epistemic daarom relatively more often contain characteristics associated with daarom's PU: causality. In our data, we found two patterns
corroborating this expectation.
The first indication concerns modality type. The category ofepistemic causality arguably contains two different types of conclusions. The first involves "real"
epistemic modality: conclusions regarding the probability of the truth of proposition P. An example of this type is S2 of (27), discussed as a PU context of dus
in section 4.3.1 (22), and repeated here as (27).
Categories of subjectivity in Dutch causal connectives: a usage-based analysis
145
but dus appears to prefer "real epistemic" modality, while daarom appears to
be distributed over the two types rather evenly. Deontic modality with dus is
exceptional. Table 2 presents an overview.
Table 2. Modality type in dus and daarom.
Epistemic modality
Deontic modality
Daarom
14
12
Total
2621
Dus
70
6
76
Total
84
18
102
We found an asymmetry in distribution of dus and daarom over modality
types (Z = 3.97; p < .001) which is in line with our expectation concerning their differences. Both connectives can express both of the modality types,
Both in epistemic and in deontic modality, the speaker expresses an attitnde
towards the proposition, which classifies both types as subjective. However, deantic and epistemic modality differ in terms of degree of subjectivity. Deontic
modality (or: agent-oriented modality [Bybee et al. 1994], root modality [J.
Sanders 1994]) reports the existence of internal and external conditions on an
agent with respect to the completion of the action expressed in the main predicate. As a report, the agent-oriented modality is part of the propositional content
of the clause (Bybee et al. 1994: 177). Deontic modality refers to a situation in
observable reality, and is not uniquely linked to the speaker as a source ofjudgment of the sitnation depicted. Epistemic modality, on the contrary, is typically
concerned with the "truth" of a proposition in which the speaker is a primary
source of responsibility of the epistemic judgment (cf. Traugott 1989; Sweetser
1990; Bybee et al. 1994). The relatively large proportion of deontic contextswhich count as relatively objective contexts, despite its inherent subjectivity marked with daarom can be interpreted as an indication that daarom 's epistemic
causal NPU exhibit conceptual affinity to daarom's PU of objective, volitional
causality.
A second pattern distinguishing epistemic daarom from epistemic dus concerns the "information statns" of the segment functioning as the conclusion in
the epistemic relation. Whereas in dus-marked epistemic relations the conclusion is always performed in the actual discourse situation, in epistemic contexts
marked with daarom, the conclusion may concern "given information". An example of an epistemic relation with a conclusion performed in aetna! discourse
is (27) above. An example of an epistemic relation with a "given" conclusion
is (28):
21 This may also be an embedded speaker, to whom the actual speaker "lends" perspective (cf. J. Sanders 1994; see also discussion of this type of fragments in Sanders et
a!. this volume).
21 Five of the epistemic daarom- fragments in the corpus could not be classified as
epistemic or deontic modality. This is why the sum total in this analysis amounts to
only 26.
(27)
(Dutch soldiers who served in Bosnia relate the high incidence of
leukemia among them to frequent exposure to impoverished uranium).
Maar de huidige hypothese wijt de leukemieen aan een virus ( .. ). Het
is dus denkbaar dat de so/daten die nu leukemie hebben gekregen,
slachtoffer zijn van iets anders dan verarmd uranium.
'But the current hypothesis attributes the leukemias to a virus. ( ... )
Dus it is conceivable that the soldiers who suffer from leukemia now,
are victims of something else than impoverished uranium.'
The other type involves so-called deontic modality (conclusions about the
desirability of some course of action expressed in P (Pander Maat and Sanders
2000: 74; cf. Sweetser 1990). An example of a causal coherence relation expressing deontic modality in S2 is (26) above. Following Sweetser (1990:49)
"deontic" and "epistemic" modality are defined as follows:
Modality type
Definition
Deontic modality
real-world obligation, permission or ability
Epistemic modality
necessity, probability or possibility
Figure 3. Distinction between deontic and epistemic modality (Sweetser 1990).
146
(28)
Ninke Stukker, Ted Sanders and Arie V<!rhagen
"Maar bij emstige brandwonden is het hele lichaam ziek", zegt Hermans. "De lever, de nieren, alle organen doen mee. Daarom is de zorg
voor deze patiiinten zo ingewikkeld."
"'Serious burns make the whole body ill", says Hermaos. "The liver,
the kidneys, all orgaos are affected. Daarom the care for these patients
is so complicated."'
Categories ofsubjectivity in Dutch causal connectives: a usage-based analysis
for this conclusion are. 23 Table 3 summarizes the distribution of"performativity
of the conclusion" over epistemic causal relations marked with dus or dilarom 24
Table 3. Distribution of"perfonnativity of the conclusion" over epistemic causal relations marked with dus or daarom.
Conclusion performed
Fragment (28) is taken from a "backgrouod" article which discusses the
question whether the capacity of care for burn injuries in the Netherlaods is
to be extended or not. The interviewee, Professor Hermans, argues against this
proposition. The cause-part of the relation in (28) must be interpreted as an
argument for the conclusion in the effect-part, stating that care for burn victims
is complicated; the paraphrase for epistemic causality "The situation in SJ is
an argument for the following conclusion: S2 " conveys this interpretation. best.
However, this is not the only aspect cornrnuoicated in this fragment. Judging
from the referential expression zo (roughly translated as 'this') preceding the
evaluative expression ingewikkeld ('complicated'), this evaluation is not new. As
a matter of fact, Hermans is quoted twice in the article. Fragment (28) represents
the second quote. Fragment (29) precedes the fragment cited in (28) aod contains
the first quote:
(29)
"Het gaat niet om het personeel. VOor deze zorg zijn zeer gespecialiseerde verpleegkundigen, internisten, anesthesisten en chirurgen
nodig. Als daar meer van komen, hebben ze ieder te weinig werk om
routine op te bouwen. En dat kan riskant zijn."
"'The arnouot of personnel is not relevaot. For this type of patient" care,
highly specialized nurses, internists, aoaesthesists, and surgeons are
needed. If their number increases, individually, they won't be able to
practice their skills to a sufficient extent. That may be daogerous."'
The proposition that "the care for these patients is (... ) complicated" is
conceptUally present immediately after this first quote, as ao inference from
Hermans' statement that "for this kind of patient care, highly specialized ... are
needed". Yet, the inferred proposition is stated explictly in the second quote, aod
the proposition innnediately preceding it firoctions as a justification for stating
this proposition. Thus, the conclusion segment of the causal relation in (28) is
ambiguous for information status. The information conveyed in the conclusion
part itselfis not entirely "new", but its status as a conclusion, and the justification
147
Daarom
24
Conc~usion
given
Total
7
31
Dus
76
0
76
Total
100
7
107
23 This ambiguity of information status of the relation as a whole renders "nonperformative conclusion relations" somewhat ambiguous for causality type as well,
namely between epistemic and non-volitional causality. Since the paraphrase test
explictly addresses the "informational surplus" of a causal relation (cf. Sanders,
Spooren & Noordman 1993) we chose to categorize this type as epistemic. Whether
categorization of "non-performative conclusion relations" as epistemic causal relations is actually the most adequate analysis remains a matter of debate (see also
Sanders & Spooren 2007, who claim that the inference of"new" knowledge is a distinguishing feature ofepistemic relations). Categorization of these cases as content
non-volitional would not alter the overall picture of the analysis of distribution of
daardoor, daarom and dus presented in this paper (Z=I.28; p<.001; cp. the values
of our present analysis [Z~4.30; p<.001] reported in section 4.3.2).
24 Epistemic causality with a "given conclusion'' marked with daarom must be distinguished from the marking of"given information" status with dus in the function of
discourse marker of information management (cf. Schiffrin 1987). An example is
Op het moment dat de eerste burger zich openlijk aforoeg wie de openin?Jshandeling
wilde verrichten, ging een groat aantal vingers de Iucht in. Het werd dus Hilde Zaer.
'At the moment the mayor explicitly asked who wanted to do the opening ceremony,
many people raised their hands. Hilde Zoer became '~en" the person to do it.' In
this fragment dus does not function as a connective marking a causal coherence
relation, but as a marker of information status, signaling how the information within
its scope is to be interpreted with respect to the rest of the information presented
in the text. See for a more elaborate discussion Evers-Verrneul & Stukker (2003);
Evers-Verrneul (2005:chapter 7). (Example taken from Evers-Vermeul & Stukker
2003: 130)
148
C-ategories ofsubjectivity in Dutch causal connectives: a usage-based analysis
Ninke Stukker, Ted Sanders and Arie Verhagen
Let us now tum to the non-volitional contexts of daarom; (30) is an example.
(30)
(Chess-player Kasparov shows disrespect with regard to his compatriot
and fellow chess-player Kramnik.) Kramnik zou iemand zijn. die nooit
of te nimmer aanvalt en daarom zo zelden verliest.
'Kramnik was supposed to be a person who never ever attacks and
daarom thus seldomly loses a game.'
The causal relation in (30) is coustructed hetween the two complement clauses.
The situation described in S2 "thus seldomly loses a game" carmot be interpreted
aB an volitional action. 1herefore, the paraphrase for volitional causal relations
"The situation in SI is a motivation for the following action: S2" does not
explicate the causal relation in (30) adequately. The paraphrase fornon-volitional
causal relations "The situation in Slleads to the following situation: S2 ",docs.
The defining characteristic of non-volitional relations is that the source of the
causal process is rwt located in an volitionally acting human being. In other
words, it does not contain an SOC. The locus of effect is either au inanimate entity
or an animate being construed as an "experiencer" (non-agent; e.g. Frawley
1992; Croft 1998). On the grounds of our second hypothesis, we expect that
non-volitional relations marked with daarom contain characteristics ofdnarom's
PU volitional causality, containing an actor SOC.
We found one pattern corroborating our expectations, namely that nonvolitional relations marked with daarom have a numerical preference for animate loci of effect "This preference is lacking in non-volitional relations marked
with daardoor (cf. the discussion of PU daardoor in section 4.3.1) (Z = 2.22;
p .02).25 Table 4 presents an overview.
Table 4. Distribution of animacy oflocus of effect over content non-volitional relations
marked with d<Jarom or daardoor.
Inanimate locus of effect
Daarom
Total
-·------···
Animate locus
Total
43
53
96
2
14
16
45
67
112
Daa:rdoor- and daarom-marked relations do not seem to differ in terms of
the nature of the role of their anitrulte loci of etTect; in all of the cases it is
25 The overall preference for animate loci of effect may be caused by a genre-effect:
newspaper articles tend te report actions of human beings ruther than describe physical processes.
149
construed not as a SOC, but as an "experiencer" of the causal effect who does not
contribute actively to bringing the causal relation about l see also our discussion
of fragments [17] and [18] in section 4.3.1).
Dus
NPU contexts of dus concern volitional causal relations (22/100). An example
is (31) -discussed in section 2, and repeated here.
(31)
(Bystanders rush to help out at the Volendam pub fire.) "Ikwoon vlakbij,
dus ik ben brandwondencreme gaan halen."
'"I live nearby dus I ran to get burn ointment."'
The paraphrase for epistemic causal relations The situation in Sl is an argumentfor the following conclusion: "S2"- which reflects dus' PU- does not fit
this context The paraphrase for volitional causal relations ·'The situntion in Sl
was a motivation for the following action: S2 ",does. Volitional causal relations
marked with dus systematically exhibit a conceptual affinity to its PU of epistemic causality. In the majority of dus-marked cases, the SOC seems to be not
only relevant as an 'volitional agenf' - which is the defining characteristic of
volitional causal relations -but also as a subjective agent "justi:tying" tbe particular course of action chosen. This aspect may cause a higher level of speaker
involvement in the causal relation, aud in this respect dus' NPU of volitional
causality reflects a defining characteristic of its PU of epistemic causality. In
this respect, volitional causal relations marked with daardoor or daarom seem
to form the mirror image from the patterns found with dus.
A pattern which we consider indicative for this tendency concerns perspective
configuration. In volitional causal relations, the cause-part of the relation may or
may not be presented from the perspective of the SOC, from v.-hose perspective
the effect-part of the relation is presented by default. Pander Maat and Sanders
(2000) found that dus-marked volitional causal relations t<:nd to be present<:d
from "continuous actor perspective" more often than daarom-marked ones. The
causal relation in (31) is presented from a continuous actor perspective, which
is signaled in this fragment by the quotation marks, and by continuous reference
to the speaker with "l" (see discussion of this example in section 2). As an
effect of this explicitly marked continuous perspective, the causal relation is
construed as containing a high level of speaker involvement; in particular, we
tend to interpret the cause-part of the relation not as a mere description of a
motivation for the action performed but rather as a justification of that action.
By way of e<:>ntrast, consider fragment (32) (discussed in section 4.3.1 as
au example of daarom's PU context of volitional causality) which contains an
150
Ninke Stukker, Ted Sanders and Arie Verhagen
Categories ofsubjectivity in Dutch causal connectives: a usage-based analysis
example of a volitional causal relation with non-continuous perspective configuration.
(32)
Microsoft-oprichter Bill Gates waagde zichzelf als eerste aan het spel
tijdens een computerbeurs in gokparadijs Las vegas. "Wij wilden iets
bouwen dat een doorbraak voor computeifreaks is'', omschrijft de
voormalige topman Microsofts ambities. Het zwart gekleurde apparaat
heeft daarom vier aansluitingen voor besturingsknuppels meegekregen, waar de gangbare "gameboys" er maar twee hebben.
'Microsoft's founder Bill Gates was the first to try the game during a
computer conference in gambler's paradise Las Vegas. "We wanted to
build something that would be a breakthrough for computer freaks"
explains Microsoft's ex-top executive. Daarom the black colored machine has been provided with four connections for game pads, while
the current "gameboys" have only two.'
S 1 of the causal relation in (32) is presented from the perspective of "Microsoft's ex-top executive" (quotes, explicit first person SOC), while S2 is presented from neutral perspective; the SOC who is conceptually present is linguistically backgrounded (cf. discussion of this fragment in section 4.3.1). Our
sample corroborates Pander Maat and Sanders' (2000) finding that dus' NPU of
volitional causality more often contain continuous perspective configurations
than volitional causal relations with daarom do (Z = 2.1; p = .036; see Appendix 2)26 . Our findings are sunnnarized in Table 5.
Conclusion from the NPU analysis
Our findings suggest that the NPU of daardoor, daarom and dus systematically exhibit conceptual affinity to their respective Pl.J. Figure 4 sunnnarizes the
tendencies of conceptual affinity found in our sample.
Daardoor
Daarom
Dus
Content
non-volitional
Non-volitional processes, animate or
inanimate LOE
Non-volitional
processes, animate LOE
0
Content
volitional
Restricted intentionality Full intentionality,
Neutral perspective
objective
Epistemic
0
Distribution of perspective configurations over content volitional relations
marked with daardoor, daarom or dus.
Daardoor, daarom
Dus
Total
Continuous perspective
Non-continuous perspective
Total
25
16
41
29
6
35
54
22
76
These findings suggest that dus' NPU contexts of volitional causality reflect
conceptual affinity to dus' PU of epistemic causality.
26 Since we did not have specific expectations concerning differences between daardoor and daarom, they were merged into one category in the statistical analysis.
Full intentionality, subjective
Continuous SOC perspective
Neutral perspective
Figure 4.
Deontic and epistemic
modality
Conclusion ''perfonned"
or "given"
Epistemic modality
Conclusion "performed"
Observed patterns of conceptual affinity of typical and non-typical usage
types of daardoor, daarom and dus. Bold face printed cells correspond to the
connectives' PU.
5. Discussion aud conclusion
5.1.
Table 5.
151
Summary and interpretation ofthe results
The starting point of our paper was the assumption that causal connectives are
directly related to conceptual categories of causality, and that causal connectives
function as categorization devices. When selecting one of the options available
in a language, the speaker assigns the causal relations expressed to a specific
conceptual type of causality. Findings from previous studies seemed to contradict this categorization hypothesis. Corpus analyses had revealed that a 1-to1-relation between categories and patterns of connective use does not exist. In
this paper, we reinterpreted these finding for three frl'quently used Dutch causal
connectives: daardoor, daarom and dus, within a usage-based framework. The
''usage-based approach to language" assumes that variation is an inherent characteristic oflanguage use, and it seeks to explain occurring patterns of variation
with reference to more general cognitive mechanisms.
Our re-analysis suggests that the patterns of use exhibited by these connectives can adequately be described and explained with reference to the well-known
cognitive phenomenon of"prototypicality structure". On the whole, our data re-
!52
Ninke Stukker, Ted Sanders and Arie ferhagen
veal two sources of evidence for this hypothesis. First, in a large majority of
cases, the connectives under investigation occur more in contexts of use which
directly reflect the relevant conceptual categories. Within a usage-based framework, this can be interpreted as an indication that these types of connective
use are relatively more enteenched in the language user's knowledge concerning
causal connectives. The more entrenched usage types can be thought of as a semantic category's prototypical center, which we referred to as the connectives'
prototypical usages (PU).
Second, the usage contexts which do not directly reflect the relevant categories, systematically exhibit conceptual affinity to their prototypical usage
schemas. We referred to these as the connectives' non-prototypical usages
(NPU). The conceptual affinity of NPU with PU could systematically be described witl1 reference to the conceptual model of subjectivity, which was shown
to govern categorization of the connectives' PU. We interpreted these findings
as an indication that the variety ofusage types per connective belong to one and
the same semantic category.
The usage-based approach to language assumes that the levels of language
structure, conceptual structure and language use interact. What does this interaction look like for the Dutch forward causal connectives we investigated in this
paper? Our findings suggest that it can be schematized as in figure 5.
Categories of subjectivity in Dutch causal connectives: a usage-based analysis
!53
edge and 1anguage use concerning the connectives daardoor, daarom anddus.
ii Linguistic category; 411 Conceptual category; C Domain of usage.
while dam-om and daardoor are typically used in contexts of objective, content
causality: Within the domain of obj e<:tive causality, daarom and daardoor distinguish between volitional and non-volitional causality. This distinction is closely
related to the conceptoal model of naive dualism - which reflects the conceptually important distinction between volitional animate beings and inanimate
entities (Verhagen and Kemmer 1997; Stukker, Sanders and Verhagen 2008).
The assumption that a direct relation exists between these conceptual categories of causality and the linguistic categories of daardam; daarom and dus,
is symbolized in Figure 5 by two sets of circles: Tbe upper level represents the
level oflinguistic knowledge; the lower level represents the level of conceptual
knowledge. The relative position of the circles symbolizes their mutoal connection: The linguistic categmies invariably reflect the conceptual categories. The
"bold" lines symbolize our ;mggestion that the linguistic categories of causal
connectives and the conceptual categories of causality they refer to, are invariable, clearly delineated, categories. Figure 5 also symbolizes the finding that the
usage of the cmmectives is not restricted to their prototypical centers - all of the
circles have extensions into other conceptual domains. This is indicated by the
oval forms. Their form and relation to the meaning circles symbolize our suggestion that the extended usages belong to the very same conceptual category
as the respective connectives' prototypes - creating an ambiguous construal in
terms of causality categories given that other linguistic or contextual elements
indicate another category of causality- but at the same time, that the possibility
of extension is restricted by the respective prototypes. Thus, we suggest that the
variability of connective use observed in corpora is to be accounted for at the
level of language use itself, and not directly at the level of linguistic and conceptual categories, even though the categories are based on experience in nsage.
Since this variability is. rooted in contextual factors, specifically .in the speaker's
subjective interpretation of a given situation and his rhetorical purposes, we cannot, as a matter of principle, predict the forms the cmmectives' NPU may takeP
This inherent variability at the level of connective use is symbolized in Figure
5 by the dotted lines. 1be asymmetric forms of the ovals representing the level
of language use, however, represent our finding that the relative frequency of
extensions from the prototypical centers vary per connective: NPU of daarom
and dus into their respective conceptual domains occur more frequently than
NPU of daardoor.
The connectives' PU reflect well-known and widely attested conceptual categories. A first conceptual distinction which is reflected in the usage of the
connectives under investigation is the distinction between objective and snbjective causality: dus is typically used in contexts of subjective, epistemic causality,
27 We can, however, make predictions. about the interpretation of a connective's NPU,
which may in turn be tested. This was the point of sections 4.2 and 4.3.
Daardoor
Daarom
Dus
Objective
Objective
Subjective
Non-volitional
Volitional
Epistemic
Causality
Causality
Causality
Figure 5. The relation between categories of linguistic knowledge, conceptual knowl-
155
154
Ninke Stulcker, Ted Sanders and Arie llerhagen
Categories ofsubjectivity in Dutch causal connectives: a usage-based analysis
5.2.
Discussion of the results from a "categorization" perspective
Maat and Sanders 2000, 2001; Pander Maat and Degand 2001). We agree that
this conceptual connection exists, but on the other hand, our results suggest that
the conceptual differences between volitional causality and epistemic causality
may be more fundamental than has been suggested in the subjectivity scale account of causal connectives. A systematic difference concerns the SOC's roles
and SOC distance to the speaker. In volitional causal relations, the SOC is part
of the causally related propositions by definition, as the agentive subject of the
action predicate expressed in the relation's consequent segment. Thus, in volitional relations, a certain distance between speaker and SOC is the default. In
epistemic relations on the other hand, the SOC is not part of the causally related
propositions. In typical cases, speaker and (extra-propositional) SOC coincide.
Typical cases of epistemic causal relations concern relations constructed by a
speaker SOC who is not explicitly referred to (that is, to its role as a SOC) in the
related propositions. An example is The light in the neighbours living room are
out. So they are not at home. In less typical cases (at least, in newspaper text)
the actual speaker reports epistemic processes constructed by other persons. An
example would be Harry saw that the lights in his neighbours living room were
out. He concluded that they were not at home (examples taken from Pander
Maat and Sanders 2000: 67). But even in these cases, the explicitly mentioned
concluder SOC is not part of the causally related propositions. The matrix sentence He concluded that functions as a mental space builder, indicating that
the epistemic relation's "illocutionary force" is to be located in an embedded
speaker (indicated by the matrix sentence) and not in the present speaker (see
for a discussion of mental space configurations like these Sanders et al.in the
present volume. See also Verhagen 2005: chapter 3).
This distinction coincides with a fundamental distinction proposed in many
theories on subjectivity, e.g. Langacker's (1987, 1990) distinction between subject ("ground") and object of conceptualization, and his defiuition of subjectivity
in terms of presence or absence of the "ground" in the linguistic construal of a
particular situation; cf. also Traugott's definition of subjectification as "meanings grounded in the speaker's subjective attitude to or a belief about what is
said" (Traugott and Konig 1991: 189; for a more elaborate discussion of various
definitions of subjectivity, see the contribution of Sanders eta!. to this volume).
The results of the usage-based corpus analysis we presented in Section 4
suggest that it is exactly this distinction between propositional and nonpropositional SOC that is mirrored in the contrasts between connectives28 . If
Some previous studies of causal connective use have suggested that their meaning must be characterized as positions on a scale of subjectivity rather than as
conceptual categories, but we have argued that the phenomena involved can actually be integrated in a natural way in a usage based approach to categorization
(cf. discussion in Section 3). This is not to deny the relevance of the observations
and analyses adduced in favor of the scalar approach - on the contrary. Below,
we discuss how a categorization perspective leads to a different interpretation of
these similar findings, offering alternative explanations of usage facts that have
been observed before, and sometimes suggesting new explanations for facts that
have been observed previously, but were left unexplained.
Onr findings suggest that~ while contexts of use of a causal connective may
differ to the extent that they reflect characteristics of the causality category it
is prototypically associated with - the connectives' function remains constant
over the complete range of contexts it is used in. We suggested that connectives'
PU and NPU belong to one and the same semantic category, and this category
is adequately characterized with reference to conceptual categories of causality.
We also suggested that connectives have one constant function: they categorize
the causal relation they are used in as an instance of a specific conceptual type.
Our suggestion to treat daardoor as representing a distinct conceptual and semantic category will be relatively uncontroversial. Previous studies have argued
for a "status aparte" for this connective as well, on the grounds of conceptual
and usage argnroents. Pander Maat and Sanders (2000) propose that daardoor
conceptually stands apart from daarom and dus because of the fact that the
condition for its use is strict: it can only mark SOC-less causal relations. Judging from usage data, in terms of frequency of occurrence of NPU, daardoor
certaioly represents the most restrictive category of the three connectives under
investigation. However, despite its low frequency, NPU of daardoor into the
domain of volitional causality do occur in a rather natural way construing the
relation as "SOC-less" (see the discussion ofNPU patterns of daardoor in section 4.3.2). This NPU pattern of occurrence does not differ from those occurring
with daarom and dus.
Onr suggestion that daarom and dus must be analyzed as similarly distinct
categories is perhaps more controversial. A pivotal argument in favor of the
"scalarity" interpretation was that the conceptual categories of volitional and
epistemic causality show conceptual aflinity: in both types, the source of the
causal relation is an volitionally acting animate being, the "Subject of Consciousness" (cf. Section 2). The fact that dus and daarom can be used in both
categories was explained with reference to this conceptual connection (Pander
28 This distinction between explicit (propositional) SOC and implicit (nonpropositional) SOC characterizes the difference between daarom and dus (PU and
NPU) in our sample better than "absence or presence of reference to the speaker"
!56
Ninke Stukker, Ted Sanders and Arie Verhagen
Categories of subjectivity in Dutch causal connectives: a usage-based analysis
dus is used in an episternic relation - its PU - the relation is "performative"
by definition. Analogously, this performative aspect is systematically present in
dus' NPU volitional causal contexts of use. A systematic pattern in these contexts of use is the suggestion that the action described is justified, and not only
descriptively motivated, which is the standard interpretation when a volitional
relation is marked with daarom. Daarom's core meaning, on the other hand,
seems to be "objective description". Tills characteristic is not only systematically present in daarom's PU volitional causality, but it is also preserved in its
epistemic contexts of use, backgrounding the epistemic context's inherent high
level of speaker involvement.
These constancies in the effects ofdus and daarom over various contexts have
been observed in other studies as well. They were accounted for by the assumption that a scalar concept of subjectivity allows for a certain amount of variation
in degrees of subjectivity "tolerated" by the respective causal connectives (cf.
Pit's [2003] "tolerance zones"). In the categorization approach proposed here,
the constant effect of connectives on the construal of the causal relation can be
explained with reference to the cognitive concept of prototypicality structure,
and by the cognitive firnction of categorization itself. The cognitive semantic
concept of "subjective construal" we introduced in Section 3.1 now adds to the
mere observation of constancies, a firnctional explanation of the occurrence itself ofNPU. Our analysis suggests that the ambiguity observed in connectives'
NPU contexts serves rhetorical purposes. In the overall interpretation of causal
relations in concrete contexts of use, the inherent character of a connective may
be "blended" with other contextual factors. Volitional causal relations, fo~ example, are inherently objective. At the same time, the fact that a SOC necessarily
does. This finding contradicts the results reported in Pander Maat & Sanders (2000),
who found that usage of dus was concentrated in categories that are close to the
speaker according to the subjectivity cline of SOC: "unspecified < nominal <
pronominal 3'd person < explicit speaker" (2000: 68-70), while daarom was more
often used with categories at greater distance from the speaker. Stukker (2005:
chapter 4) discusses implications of either pattern for a definition of "subjectivity"
in causal coherence relations.
Table Nl Distribution of dus and daarom over SOC formulation types within the
category "explicit SOC" relations.
Total
Unspecified Nominal 3id Pronominal 1st person
znd or 3rd
person
Daarom
Dus
Total
II
5
16
8
6
14
person
12
5
17
19
6
25
50
22
72
157
plays a role opens the way to introduce elements of subjectivity in such contexts
(cf. Sanders et al. this volume). Conversely, inherent epistemic causal relations
may become "objectified" to the extent that their conclusions have their source
in outside reality rather than in the speaker's personal consciousness. 29 NPU
concerning "category crossing" usages within the domains of objective volitional and non-volitional causality can be analyzed in similar terms: volitional
daardoor backgrounds the SOC's volitionality in bringing about the inherently
intentional causal relation; non-volitional daarom foregrounds aspects of SOChood in the inherently non-intentional causal relation.
An inherent advantage of a categorization approach to connectives is that it
allows for integration with related fields of study. There is ample evidence that
the categorization firnction per se is crucial in causality markers manifested at
the clause-level of linguistic structure as well (e.g. Tahny 1988; Kemmer and
Verhagen 1994; Verhagen and Kemmer 1997; Wolff and Song 2003; Stukker,
Sanders and Verhagen 2008). Furthermore, the categorization perspective fits in
with more general theories of discourse structure- not only because categorical
concepts are more easily integrated with the idea that connectives are markers of
coherence relations, but also because the categories of objective and subjective
causality have a direct parallel in taxonomies of coherence relations (Sanders
2005); moreover, it also aligns with the categorical difference between the connectives in terms oftheir interaction with grammatical negation (Verhagen 2005:
197-205).
Despite the fact that we consider these results important, they should be
corroborated by an independent investigation from an onomasiological perspective (cf. Geeraerts 1997 for an elaborate discussion of the relation between
the semasiological and the onomasiological perspectives). When we introduced
Hypothesis 1, we noticed that it inherently addressed the relation between form
and meaning in a semasiological way. Given the forms, what do they express:
categories or extended areas on a scale? As figure 5 depicts, our present conclusion includes the idea of a tight connection between linguistic and conceptual
structure. Tills hypothesis not only predicts the kind of relations between PU and
NPU as discussed so far; it specifically also predicts that a particular type of concept, say volitional causality, will typically, and most frequently, be expressed
by the associated form. This is the onomasiological question.
To see that this is, indeed, an independent issue, consider the distribution of
uses we started out with (figure 2), given in table 6.
29 For an analysis ofNPU of daarom and dus in terms of blending of mental spaces,
see the contribution of Sanders et al., this volume
!58
Ninlre Stukker, Ted Sanders and Arie rerhagen
Categories ofsubjectivity in Dutch causal connectives: a usage-based analysis
Table 6. Distribution of daardoor. daarom and dus over domains of use in the sample, 30
159
These ratios can now be used to construct table 7 from table 6:
·~
Daardoor
Daarom
Dus
Total
96
16
0
Content Volitional
4
50
Epistemic
0
31
97
22
76
112
76
107
295
Content Non-volitional
Total
100
98
Looking at the colunms, it becomes evident that the prototypical use of
daardoor is content non-volitional, that of daarom is content volitional, and that
of dus is epistemic. Looking at the rows, it seems that the prototypical way of
expressing a content non-volitional relation is daardoor, for a content volitiollal
relation it is daarom, and for an epistemic relation it is dus. However, the validity
of such a conclusion depends on the ratios of use of the three connectives, which
is ignored in the table because we sampled I 00 iustances of each connective (in
order to have sufficient cases for analysis). For example, if dus would be used
five times more often than daarom, the number of cnntent volitional relations
marked by dus would be 110 rather than 22, and since this is considerably more
than the 50 cases marked with daarom, the protoypical marker for such relations
would be dus rather than daarom. Such a situation could then be constructed as
an argument for the conclusiou that dus is a kind of"unmarked case", ioheren!ly
allowing a range of different uses, whereas daarom would be a "marked case"
with an ioherently restricted range oftJSCs.
In principle, of course, this issue requires an independent investigation, using
a different and independent sampling method. Thus, we cannot give a definitive
answer to the onomasiological question here, but we are able to provide good
indications why we expect our prediction to be corroborated, viz. by reconstructing the data presented in table 6 in terms of the actual ratios of use of the three
connectives in the corpus. The corpus (cf. note 16) was searched for the three
connectives in a particular way independent of the kind of relation expressed. 31
On this basis, the ratios of use in this corpus were established as follows:
daardoor : daarom : dus
I : 1.81 : 3 .II
31 As the search engine does not allow searches that will result in over 3000 hits, 4 combinations ofthe connectives with common words ftom different parts ofspeech were
used: with weer ('again'), soms ('sometimes'), dag ('day'), and altijd ('always').
The respective numbers and the resulting ratio•s returned for these are presented in
tbe following table:
Table 7. Distribution of daardoor, daarom and dus over domains of use in the sample,
actua1 ratios reconstructed,
Daardoor
--···
Daarom
Dus
Total
Content Non-volitional
96
29
0
125
Content Volitional
4
91
163
Epistemic
0
56
68
236
292
100
176
304
580
Total
---···
When we now look at the numbers in the rows, we can see that the relationships predicted by our hypothesis actually hold: the prototypical marker for a
content non-volitional relation is daardoor (96 out of 125), for a content volitional relation it is daarom (91 out of 163), and for an epistemic relation it
is dus (236 out of 292) (Z = 4.57; P < .001). Still, it will be clear that an
independent investigation of this issue is called for, and we plan to report this
in future research.
5.3.
More methodological considerations and future research
We consider our results as presenting strong indications in favor of the categorization hypothesis, but the results of the present study generate new research
questions as well. For example, taxonomies of coherence relations have treated
the distinction between volitional and non-volitional relations as being of secondary importance (cf. Sanders 1992; Sanders et aL 1992, 1993). Our findings
suggest that the importaoce of ••animacy" deserves reconsideration in a cognitively plausible theory on discourse structure. 1be categories of volitional
---··
Table Nl Reconsruction of ratios of use.
soms
Dus
2429 3.07 948 2.73 1192 3.14 1712 3.42
Daarom
1374 1.73 558 1.6 712 1.87 101& 2.03
Daardoor 791 1
347 1
380 1
501
1
weer
Total
6281
3.11
3662 1.81
2019 1
···~--···
As the table shows, the variation between the different combinations is limited, and
we therefore conclude that the average ratios in the last column provide a reliable
indication of the acrual ratios of use of the connectives in the entire corpus, also in
view of the large number of observations.
160
Ninke Stukker, Ted Sanders and Arie Verhagen
and non-volitional causality, relevant for distinguishing Dutch daardoor and
daarom, show an interesting parallel to the categories of animate and inanimate
causality as identified with Dutch causal verbs (Verhagen and Kemmer 1997·
see for discussion of the exact parallel Stokker, Sanders and Verhagen 2008:
Stokker 2005).
Another intriguing finding is that grammaticalized markers of "subjective
causality" (a force dynamic interaction at the level of representation) are only
found at the discourse-level of language use. These issues will be addressed in
future research (see also Stokker, Sanders and Verhagen 2008).
Obviously, our stody has limitations as well. A first source oflimitations we
want to discuss here is the nature of the methodology we used, which especially
affected our analysis ofNPU. Our methodology was built on the usage-based
assumption that linguistic knowledge has the form of usage schemas, specifying
the type of context in which- in our case- causal connectives may adequately
be used. Our method of analysis consisted of investigating observable characteristics of the individual, contextual instances of the more abstract usage
schemas. The context's causality type was determined on the basis of paraphrases, which focused on SOC-type and event type of the consequent segment
(cf. Section 4.1 ). If the causality type of the context was congruent with the
connective's hypothesized inherent meaning, the context was classified as a PU,
if it was not congruent with the connective's inherent meaning, it was classified
asaNPU.
NPU were further analyzed on the basis of other segmental characteristics
which could be related to the concepts of animacy and subjectivity (cf. Section 4.2). This analysis revealed converging trends of conceptual affinity,. but it
did not include all of the connectives' individual usage contexts. This is partly
due to the fact that the present analysis had an explorative character, mainly
focusing on patterns and phenomena already reported in previous studies on
connectives, or in related fields of stody. But more importantly, since our analysis was restricted to detecting explicit, objectively observable, contextoal indications of subjectivity and animacy, NPU contexts of use not containing explicit
indications could not be taken into account. On the basis of the inherent underspecification oflanguage use, we must expect that natural contexts of use exist
in which the connective chosen is the only visible indication of the intended
causality category (cf. the discussion of NPU of causal verbs in Section 4.1;
cf. on connectives Pander Maat and Sanders 2000:76--7). These cases were not
included in our analysis.
A further limitation, related to the one previously mentioned, is that our
methodology did not allow for testing the categorization hypothesis' restrictive-
Categories of subjectivity in Dutch causal connectives: a usage-based analysis
161
ness32 . Prototypicality structure inherently defines constraints on the extendibility ofNPU; it predicts that if usage contexts of connectives do not contain any
conceptual affinity to the connective's prototype, this usage context is unacceptable33. We can illustrate this prediction with reference to the following
examples:
(25)
Het was extreem koud. # Daardaor zochten we een cafe op.
'It was extremely cold. Daardoor we entered a cafe.'
(26)
De straten zijn nat. # Daarom zal het wei geregend hebben.
'The streets are wet. Daarom it must have rained.'
The connective daardoor leads to an incoherent interpretation (signaled by the
symbol#) in the volitional causal context of (25). On the basis of the patterns of
use reported for daardoor in Section 4.3, and on the basis of general assumptions
in usage-based theory, we can explain the incoherence of this fragment as follows: the context of use of(25), in particular the combination of verb and subject
in the consequence clause, is inherently volitional. Daardoor always preserves
characteristics of its prototype "non-volitional causality", therefore it can only
be used in volitional contexts containing elements of restricted intentionality
(which is not the case in this fragment). A similar line of reasoning holds for
(26). In this example, a conclusion regarding the plausibility of the proposition
in S2 is justified with the argument in S I, while there is no causal relation between the propositions themselves (it is a case of so-called abductive reasoning).
In this type of epistemic relations, a high level of speaker involvement is an inherent aspect of the interpretation. Daarom always preserves characteristics of
its prototype "volitional causality", not standardly containing speaker involvement; therefore it can only be used in epistemic contexts with moderate speaker
involvement (which (26)) is not).
In future research, we will try to reduce the impact of the inherent limitations
of our usage-based method of analysis by constructing a more elaborate model
32 In part, the inability to define clear boundaries to extendibility of use may be a
matter of principle. The usage mechanisms described here imply that it is simply
impossible to predict all possible usage types on the basis of theory alone - the
inherent underspecification of language use, and the subjective basis of linguistic
construal entail that utterance meaning cannot be derived compositionally from its
constituting parts, nor be predicted completely in advance (cf. Verhagen 1997).
33 The suggestion that restrictions on connectives' usability may exist was noticed by
previous studies as well (cf. Pit's [2003] "tolerance zone" of connectives), but left
unexplained. Within the usage-based framework, this fact, too, can be explained
with reference to the connectives' prototypical contexts of use.
!62
Ninke Stukker, Ted Snnders and Arie Perhagen
of analysis, which allows us to measure linguistic and non-linguistic indications
of animacy and ~ectivity in a more systematic way. Such a model will enable
us to refme and extend our analyses of NPU pattorns (Stukker and Sanders
submitted).
A second limitation concerns the generalizability of our results. We investigated only one variety of discourse: written newspaper texts. Recent studies
have suggested that conventions of connective use are co-determined by context factors, of which modality of communication ( VITitten, spoken, electronic)
is an important one (cf. Spooren, Sanders, Huiskes and Degand to appear). A
concrete example of the way medislity influences the linguistic expression of
causal relations is the distribution of cansality types. In our corpus we hardly
encountered "speech act causality", assumed to be a causality category as important as content and epistemic causality are (Sweetser 1990; Spooren et al. to
appear; see for discussion Sanders et al. this volume; Sanders and Spooren this
volume). Evidently, ifwe are to give a complete account of the relation between
the linguistic expression of cansality and cognitive concepts, we should extend
our field of study to other discourse types as well (cf. Sanders 2005).
Furthermore, the analysis presented here does not decide how the usage
patterns found are reflected in the language user's mental representation of
the semantic categories of the connectives under investigation. Our findings
suggest that the connectives' PU \lave a special status in tbe language users'
lingnistic knowledge. But we don't know whether besides these prototypical
cores other usage types have attained the status of conventionalized, entrenched
usage schema (cf. Langacker 1987, 2000), e.g. as conventionalized "blending"
schemas (cf. Sanders et al. this volume). For example, how must we interpret
frequency differences among NPU of daarom and duson the one hand, and daardoor on the other? The specific distribution patterns we found suggest that the
relatively frequent occurrence ofNPU of daarom and du.s may have formed an
independently entrenched, conventionalized subschema in the language user's
mental representation of semantic knowledge (cf. Langacker 2000; Bybee 2006,
2007) which is less likely with respect to the relatively unfrequent NPU of
daardoor. But even if this is the case, our analysis suggests that the categories of
subjectivity as proposed here (i.e. volitional causality and epistemic causality)
provide for the essential level of generalization. Future analyses of other text
genres and medialities will address the question to what extent the frequeuey.
differences arnoog NPU patterns form a stable pattem across varieties of discourse (e.g. because of a relatively high fimctional value of ambiguity between
objective and subjective causality in human communication; or because of in~
herent conceptual affinity- cf. Pander Maat and Sanders 2000) or whether they
are restricted to the newspaper discourse investigated.
Categories ofsubjectivity in Dutch causal connectives: a usage-based analysis
163
A final - but crucial issue for further research we want to address here
is: what is the role of the connective in the overall interpretation of the causal
relation, as opposed to other lingnistic indications of subjectivity in the context?
Our construal analysis suggests that a connective has the effect of foregrounding
aspects of subjectivity io the connective's context of use that are congruent with
the connective's meaning. Tills effect comes out cle.arest in our analysis ofNPU,
defined in our analysis as contexts ofuse which are ambiguous for causality type.
The construal analysis predicts that the coiDleLiive used to mark the arnbigunus
relation foregrounds the aspects that are congruent with the connectives function
of ascribing the causal relation at hand to the conceptual type of causality the
connective is associated with. Elements in the contexts that are incongruent
with this connective's function may be backgrounded, but they \viii still play a
rol in the relation's overall interpretation. The resulting interpretation may be
characterized as a "blend" of causality types, in the sense proposed by Sanders
et al. (this volume).1bis question will be investigated further in future research
that aims at analyzing the contexts of connective use in greater detail (Stokker
and Sanders submitted).
5.4.
Conclusion
Testing hypotheses eoncemiog language structure against data from namrallanguage use is a method commonly accepted in functional and cognitive linguistics. However, the extent to which different approaches acknowledge the role of
language use mechanisms in resulting theories and descriptions varies strongly.
If corpus analysis is used as a method for testing hypotheses concerning the relation between language and cognition; an obvious prereqnisite is that cognitive
mechanisms manifest at the level oflanguage use are taken into account as well.
The analysis presented in the present paper shows how a usage based approach
to language use data adds to our understanding of the relation between language
structure, cognitive structure, and language use. We believe that the usage-based
analysis of connectives proposed here has advantages from an empirical as well
as from a theoretical perspective. It allows for generalization over usage contexts
per connective, just like the subjectivity scale proposed by previous approaches
did, but it also explains concrete, observable usage putterns, such as clustering
of usage types and flexibility of use, as a consequence of more general mechanisms governing language nse and of general cognitive mechanisms such as
the categorization function and prototypicality structure. Thus, the usage-based
perspective proposed in this paper adds to the descriptive power as well as to
the cognitive plausibility of the analysis of causal cmmectives.
164
Ninke Stukker, Ted Sanders and Arie Yerhagen
Categories ofsubjectivity in Dutch causal connectives: a usage-based analysis
Acknowledgements
Appendix2
We would like to thank Barbara Dancygier and Dirk Geeraerts for their insightful
comments on an earlier version of tltis paper, which made us reconsider some
aspects of our proposal, and Huub van den Bergh for invaluable advice on
statistic analysis. As usual, the responsibility for all claims made here is entirely
our own. The first two authors acknowledge the support of NWO vici-grant
277-70-003, awarded to Ted Sanders, while preparing this paper.
Contrast analysis
165
In the contrast analysis, each of the hypotheses was restated in terms of expeeted
relative frequencies. Actual distribution of dnardom; daarom and dus was compared to the expectations. The bypotheses are supported if the Z-score resulting
of this sualysis has as a value Z > 1.96 (p < .05).
TableA-3. Expected relative distribution of daardoor, dam-om and dus over relation
categories.
Appendix]
Sample of texts
Daardaor
Daarom
Dus
Content Non-volitional
+2
-l
-l
Content Volitional
~I
+2
-I
-I
I
+2
--~
Table A-I. Composition of the text sample.
...
~~~~----····~~-
Daardoor
Daarom
Dus
50
50
50
50
50
Informative text genres
Persuasive text
50
Table A-4. Distribution of daardoor, daarom and dus over domains of use in the sample34
Table A-2. Specification of text genres in the sample (indicated by terms used by Lexis
Nexis' nserinterface).
lnfonnative text
achtergrond (ac
"background")
biogrnfie (bio ~"biography")
Daardoor
Daarom
Dus
Total
!6
50
31
97
0
112
22
76
76
98
295
Content Non-volitional
Content Volitional
96
Persuasive text
brief (br ~"letter")
Epistemic
0
100
column (co- ''column")
opinie (opi- "opinion piece")
opening (op "opening")
recensie (rec- '"review")
portret (]lOr~ ''portrait")
hoofdartikel (ha ~"leading article'')
reportage (rep ... "[running] commentary")
necrologic (nee ~"obitoary")
wedstrijdverslag (sp "sporting event article") . _ _ __
4
Total
-----------
107
z = 4.30; p < .001
Table A-5. Expected relative distribution of daarom anddus over modality type in epistemic causal relatlons.
Epistemic modality
z=
Deontic modality
.;:___
Daarom
~1
+l
Dus
+I
~1
3.97; p < .001
166
Ninke Stukker, Ted Sanders and Arie Verhagen
Categories ofsubjectivity in Dutch causal connectives: a usage-based analysis
NoSOC
soc
Daw·doot
+1
-I
Bybee, J., R. R"'Idns & W. Pagliuca
The evolution of grammar; Tense, aspect, and modality in the lan1994
guages of the world. Chicago/Louden: The University of Chicago
Press.
Croft, W.
1998
The structure of events and the structure oflanguage.ln: M. Tomasello
(ed. ), The new psychology of language. Cognitive and fonctional approaches to language structure. Mahwah/New Jersey: Erlhaum.
D'Andrade
A folk model ofthe mind. In: Holland & Quinn (eds.), Cultural models
1987
in language and thought, 112-148. Cambridge: Cambridge University
Press
Degand, L. & Pander Maat, H.
A contrastive study of Dutch and French causal connectives on the
2003
Speaker Involvement Scale. In: A. Verhagen & J. van de Weijer(eds.),
Usage based appmaches to Dutch, 175-199. Utrecht: LOT
DeSmet, H. & J.-C. Verstraete
Coming to terms with subjectivity. Cognitive Linguistics 17: 365-392.
2006
Evers-Vermeul,l
The development of Dutch connec-tives. Change and acquisition as
2005
windows on form-function relations. Ph.D. dissertstion, Lar, Utrecht
University.
Evers-Vermeul, J. & N. Stukker
S~ectificatie in de ontwikkeling van causale connectieven? De dia2003
chronic van daarom, dus, want en omdat. Gramma/rrr 9(213): 111-
Daarom
-I
+I
Fauconnier. G.
Iable A-6. Expected relative distribution of "perfonnativity of the conclusion" over
epistemic causal relations marked with dus or daarom.
Conclusion performed
_____________ -1
__m
----~D-awv
Conclusion given
+1
-_1________
D~=s________________+~l___________________
____
z=
5.41; p < .001
Table A-7. Expected relative distribution ofperspective configurations over content volitional relations marked with daardoor, daarom or dus.
------~-----
...
Continuous persp:~•:::cti:::.v_:e_ _N:....:..on:..-_c_ontinuous perspective
Daardoor. daarom
Dus
-!
+I
~~-------------~
+I
-1
Tab/eA-8. Expected relative distribution of SOC type over content non-volitional relations mmked "ith daarom or daardoor.
·----------------
z=
167
·----
-------
2.22;p = .02
Reforences
Barlow, M. & S. Kemmer (eds.)
Usage based models oflanguage. Stanford: CSLI Publications.
2000
Bybee, J.
Morphology: a study ofthe relation between meaning andform. Am1985
sterdam/Philadelphia: John Benjamins.
Byhee,J.
From usage to grammar: The mind's response to repetition. Language
2006
82: 711-733.
Bybee,J.
Frequency of use and the organization of language. Oxford et al.:
2007
Oxford University Press.
139.
1994
Frawley, W.
1992
aspects qfmeaning construction in natura/language.
Cambridge: MIT Press.
~'dental spaces:
Linguistic semantics. Hillsdsle/New Jersey; Lawrence ErlbaumAssociates.
Geeraerts; D.
Diachronic prototype semantics. A contribution to historicallexicology. Oxford: Clarendon Press.
Geeraerts, D., S. Grondelaers' & P. Bakema
The structure of lexical variation. Meaning naming, and context.
1994
Berlin et aL: Mouton de Gruyter.
Goldberg, A.
2005
Constructions at work. The nature ofgeneralization in. language.' Oxford: Oxford University Press.
1997
168
Ninke Stulcker, Ted Sanders andArie Verhagen
Hasher, L.
1984
Categories ofsubjectivity in Dutch causal connectives: a usage-based analysis
169
Pander Maat, H. & T. Sanders
Automatic processing of fundamental information: The case of frequency of occurrence. American Psychologist 39: 1372-1388.
Hobbs, J. R.
1979
Coherence and coreference. Cognitive science 3: 67-90.
Kemmer, S. &A. Verhagen
1994
The grammar of causatives and the conceptual structure of events.
Cognitive Linguistics 5-(2): 115-156.
Knott, A. & R. Dale
1994
Using linguistic phenomena to motivate a set of coherence relations.
Discourse Processes 18 (1): 35--<i2.
Langacker, R. W
Foundations of cognitive grammar, Volume 1. Theoretical prerequf1987
sites. Stanford: Stanford University Press.
Langacker, R. W
[1990] Concept, image, and symbol. The cognitive basis ofgrammar.
2002
Berlin, New York: Mouton de Gruyter.
Langacker, R. W
A dynamic usage-based model. In: M. Barlow & S. Kemmer (eds.),
2000
Usage based models oflanguage. Stanford: CSLI Publications.
Lipka, L.
Prototype semantics or feature semantics: an alternative? In: W
1987
LOrscher & R. Schulze (eds.), Studies in linguistics, literary criticism, and language teaching and learning. To honour Jterner Hiillen
on the occasion ofhis sixtieth birthday, 282-298. Tiibingen: Nar.
Lyons,J.
Linguistic semantics. An introduction. Cambridge: Cambridge Uni1995
versity Press.
MacWhinney, B.
2000
Connectionism and language learning. In: M. Barlow & S. Kemmer
(eds.), Usage based models oflanguage. Stanford: CSLI Publications.
Mann, W. C. & S. A. Thompson
1988
Rhetorical structure theory: towards a functional theory of text organization. Text 8: 243-281.
Pander Maat, H.
2001
Tekstanalyse. Wat teksten tot teksten maakt. Bus sum: Coutinho.
Pander Maat, H. & L. Degand
2001
Scaling causal relations and connectives in terms of speaker involvement. Cognitive linguistics 12(3): 211-245.
Pander Maat, H. & T. Sanders
1995
Nederlandse causate connectieven en het onderscheid tussen inhoudelijke en epistemische relaties. Leuvense Bijdragen 3: 349-374.
2000
Domains of use or subjectivity: The distribution of three Dutch causal
connectives explained. In: E. Couper-Kuhlen & B. Kortmann (eds.),
Cause, condition, concession, and contrast: Cognitive and discourse
perspectives, 57-82. Berlin/New York: Mouton de Gruyter.
Pander Maat, H. & T. Sanders
2001
Subjectivity in causal connectives: An empirical study oflanguage in
use. Cognitive linguistics 12(3): 247-273.
Pit,M.
2003
How to express yourself with a causal connective? Subjectivity and
causal connectives in Dutch, German and French. Ph.D. dissertation,
Utrecht University. Amsterdam: Rodopi.
Radden, G., K. M. Kopcke, Th. Berg & P. Siemund
2007
Aspects ofmeaning construction. Philadelphia: John Benjamins.
Rosch, E.
1973
Natural categories. Cognitive Psychology 4: 328-350.
Rosch, E. & C. Mervis
1975
Family resemblances: studies in the internal structure of categories.
Cognitive Psychology 7: 573-605.
Sanders, J.
1994
Perspective in narrative discourse. Dissertation Katholieke Universiteit Brabant.
Sanders, J. & W Spooren
1997
Perspective, subjectivity, and modality from a cognitive linguistic
point of view. In: W. Liebert, G. Redeker & L. Waugh (eds.), Discourse and perspective in cognitive linguistics, 85-112. Amsterdam/Philadelphia: John Benjamins.
Sanders, T.
1997
Semantic and pragmatic sources of coherence: on the categorization
of coherence relations in context. Discourse Processes 24: 119-147.
Sanders, T.
2005
Coherence, Causality and Cognitive Complexity in Discourse. In:
M. Aurnague, M. Bras, A. Le Draoulec & L. Vieu (eds.), Proceedings/Actes SEM-05, First International Symposium on the exploration
and modelling ofmeaning, 105-114. Toulouse: Universite de Toulouse
le Mirail..
Sanders, T. & W Spooren
2007
Discourse and text structure. In: D. Geeraerts & H Cuyckens (eds.),
Handbook ofcognitive linguistiCs. Oxford: Oxford University Press.
Sanders, T., J. Sanders & E. Sweetser
this volume Causality, cognition and communication: A mental space analysis of
subjectivity in causal connectives.
170
Ninke Stulcker, Ted Sanders and Arie Verhagen
Sanders, T. J. M., W. P.M. Spooren & L. G. M. Noordman
1992
Toward a taxonomy of coherence relations. Discourse Processes 15:
1-35.
Sanders, T. J. M., W. P. M. Spooren & L. G. M. Noordman
1993
Coherence relations in a cognitive theory of discourse representation.
Cognitive Linguistics 4: 93-133.
Schiffiin, D.
1987
Schmid, H.-J.
2000
Discourse markers. Cambridge: Cambridge University Press.
English abstract nouns as conceptual shells. From corpus to cognition.
Berlin: Mouton de Gruyter.
Sperber, D. & D. Wilson
1995
Relevance. Communication & cognition. Oxford et al.: Blackwell.
Spooren, W., Sanders, T., Huiskes, M. & Degand, L.
to appear
Subjectivity and Causality: A Corpus Study of Spoken Language. In:
J. Newman & S. Rice (eds.), Conceptual Structure in Discourse and
Language.
Stukker, N.
2005
Causality marking across levels of language structure. A cognitive
semantic analysis of causal verbs and causal connectives in Dutch.
Dissertation, LOT, Utrecht University.
Stukker, N. & T. Sanders
2009
Another('s) perspective on subjectivity in causal connectives: a usagebased analysis of volitional causal relations. Discourse, 4.
URL: http://discours.revnes.org/index7260.html.
Stukker, N., T. Sanders & A. Verhagen
2008
Causality in verbs and in discourse connectives. Converging evidence
of cross-level parallels in Dutch linguistic categorization. Journal of
Pragmatics 40, 1296-1322.
Sweetser, E. E.
From etymology to pragmatics. Metaphorical and cultural aspects of
1990
semantic structure. Cambridge: Cambridge University Press ..
Talmy,L.
Force dynamics in language and cognition. Cognitive Sciencel2: 491988
100.
Taylor, J.R.
Linguistic categorization. Prototypes in linguistic theory. Oxford:
1995
Clarendon Press.
Traugott, E. C.
On the rise of epistemic meanings in English: an example of subjec1989
tification in semantic change. Language 65:31-55.
Categories ofsubjectivity in Dutch causal connectives: a usage-based analysis
171
Traugott, E. C.
1995
Subjectification in grammaticalisation. In: D. Stein & S. Wright
(eds.), Subjectivity and subjectivisation: linguistic perspectives, 3154. Cambridge: Cambridge University Press.
Traugott, E. C. & E. Konig
1991
The semantics-pragmatics of grammaticalization revisited. In: E.C.
Traugott & B. Heine (eds. ),Approaches to grammaticalization, volume
J: 189-218. Amsterdam/Philadelphia: Benjamins.
Ungerer, F. & H.-1. Schmid
1996
An introduction to cognitive linguistics. Harlow et al.: Longman.
Verhagen, A.
1997
Verhagen, A.
2000
Verhagen, A.
2005
Context, meaning, and interpretation, in a practical approach to linguistics. In: L. Lentz & H.L. W. Pander Maat (eds. ), Discourse analysis
. and discourse evaluation, 7-39. Amsterdam: Rodopi.
Interpreting usage: construing the history of Dutch causal verbs. In:
M. Barlow & S. Kemmer (eds.), Usage based models of language,
261-286. Stanford: CSLI Publications.
Constructions of intersubjectivity. Discourse, syntax, and cognition.
Oxford: Oxford University Press.
Verhagen, A. & S. Kemmer
1997
Interaction and causation: causative constructions in modem standard
Dutch. Journal of Pragmatics 27:61-82.
Wittgenstein, L.
1958
Philosophical investigations. G.E.M. Anscobe (trans.). Oxford:
Blackwell.
Wolff, Ph. & G. Song
2003
Models of causation in the semantics of causal verbs. Cognitive Psychology 47: 276-332.