Post on 21-Jan-2023
This article appeared in a journal published by Elsevier. The attachedcopy is furnished to the author for internal non-commercial researchand education use, including for instruction at the authors institution
and sharing with colleagues.
Other uses, including reproduction and distribution, or selling orlicensing copies, or posting to personal, institutional or third party
websites are prohibited.
In most cases authors are permitted to post their version of thearticle (e.g. in Word or Tex form) to their personal website orinstitutional repository. Authors requiring further information
regarding Elsevier’s archiving and manuscript policies areencouraged to visit:
http://www.elsevier.com/copyright
Author's personal copy
Causality in verbs and in discourse connectives:
Converging evidence of cross-level parallels in
Dutch linguistic categorization
Ninke Stukker a,*, Ted Sanders a, Arie Verhagen b
a Utrecht Institute of Linguistics OTS, Utrecht University, Trans 10, 3512 JK Utrecht, The Netherlandsb Leiden University Centre for Linguistics, Leiden University, P.O. Box 9515,
2300 RA Leiden, The Netherlands
Received 27 November 2006; received in revised form 14 June 2007; accepted 27 October 2007
Abstract
Several authors have proposed to describe the meaning and use of causality markers with reference to
conceptual models of causality. If a parallel between semantic categories and conceptual categories exists,
we would expect that similar conceptual models of causality are manifest across different types of
linguistic constructions expressing causality. This cross-level similarity hypothesis is investigated in the
present paper. So far, causality markers of different grammatical types have typically been studied in
isolation. We argue that for a full understanding of the interaction between conceptual structure and
linguistic structure, an integrative perspective on different types of causality markers is needed. We focus
on causal verbs (manifest on the clausal level of the linguistic structure) and causal connectives (discourse
level) in Dutch. Pursuing the research strategy of converging evidence, we first present theoretical
considerations, and subsequently discuss data from language use suggesting that cross-level parallels exist
at an analytical level. Then, we report an experiment that aimed to test language users’ intuitions on the
cross-level similarity hypothesis. The results are interpreted as empirical evidence for the parallels in
meaning of causal verbs and discourse connectives. Remaining challenges for experimental studies of
language users’ intuitions are discussed.
# 2007 Elsevier B.V. All rights reserved.
Keywords: Causality; Causal auxiliary verbs; Discourse connectives; Conceptual models; Categorization; Converging
evidence
www.elsevier.com/locate/pragma
Available online at www.sciencedirect.com
Journal of Pragmatics 40 (2008) 1296–1322
* Corresponding author. Tel.: +31 30 253 6228; fax: +31 30 253 6000.
E-mail addresses: Ninke.Stukker@let.uu.nl (N. Stukker), Ted.Sanders@let.uu.nl (T. Sanders),
A.Verhagen@let.leidenuniv.nl (A. Verhagen).
0378-2166/$ – see front matter # 2007 Elsevier B.V. All rights reserved.
doi:10.1016/j.pragma.2007.10.005
Author's personal copy
1. Introduction1
Causality is a fundamental concept in human thinking and reasoning. It is not surprising that
most, if not all, languages in the world have a range of lexical expressions specifically designed
for communicating causal relations. This paper focuses on two grammatically different types of
causality markers in Dutch. We investigate the semantic contrasts expressed by different causal
auxiliary verbs, marking causal relations expressed within one clause, and those expressed by
different causal connectives, marking causal relations between clauses. Some examples2:
Causal verbs:
(1) [De extreme koude]cause deed zelfs [de rivieren bevriezen]effect.
[The extreme cold]cause made/caused even [the rivers (to) freeze]effect.
(2) [Ze]cause liet [haar zoon onder luid protest zijn bord leegeten]effect.
[She]cause made/had [her son empty his plate]effect, despite his complaints.
Causal connectives:
(3) [Het was extreem koud]cause. Daardoor [waren zelfs de rivieren bevroren]effect.
[It was extremely cold]cause. Because of that [even the rivers were frozen]effect.
(4) [Het was extreem koud]cause. Daarom [zochten we een cafe op]effect.
[It was extremely cold]cause. That’s why [we entered a cafe]effect.
(5) [Het is onbewolkt]cause. Dus [het wordt koud vandaag]effect.
[The sky is clear]cause. So [It will be cold today]effect.
These examples illustrate that Dutch, like most other languages, offers alternative options to
mark causal relations, and that the presence of different marking options recurs in constructions
of different grammatical type, manifest at different levels of the linguistic structure. The causal
‘auxiliary verbs’ doen (1)—roughly equivalent to English ‘make’, and laten (2)—roughly
equivalent to either English ‘let’ or ‘have’, are used in constructions referred to as ‘analytic
causatives’. They can be characterized as ‘two-verb constructions’ that express a predicate of
causation (finite form of doen or laten) and a predicate of effect, expressed as an infinitive
(cf. Kemmer and Verhagen, 1994; Wolff and Song, 2003).
Examples (3)–(5) are causal coherence relations, relating discourse segments, minimally
clauses, into a coherent whole (cf. Hobbs, 1979; Mann and Thompson, 1988; Sanders et al.,
1992). They may or may not be linguistically marked with lexical or grammaticized expressions.
An example of the latter type are the ‘connectives’ under investigation in the present study. Like
many other languages, Dutch offers a variety of connectives. In constructions expressing
‘forward causality’ (where in presentation order the cause precedes the effect) the most
frequently used ones are daardoor (not having a grammaticized counterpart in English; best
approximated by ‘because of that’), daarom (‘that’s why’) and dus (‘so’).
N. Stukker et al. / Journal of Pragmatics 40 (2008) 1296–1322 1297
1 This paper is based on Stukker (2005), especially chapter 6.2 We focus on the causal relations mainly from a conceptual perspective. Therefore, the English glosses of our Dutch
text material will not contain literal translations.
Author's personal copy
Why do language users need so many lexical contrasts for marking causal constructions? An
obvious answer would be: because the markers have different meanings, and because these
differences are salient enough to maintain specialized expressions for communicating them. An
explanation suggested by several cognitive semanticists is that causality markers function as
‘categorization devices’, assigning the causal relation expressed to a specific conceptual type of
causal relation. So far, causality markers manifest at different levels of linguistic structure have
typically been studied in isolation. As a consequence, the notions used to describe the meaning and
use of causality markers vary enormously among grammatical types. For example, causal verbs in a
variety of languages have been characterized making use of the notions ‘coercion’, ‘permission’,
‘volition’, ‘control’, ‘autonomy’, ‘directness’ (cf. Shibatani, 1976; Givon, 1976; Wierzbicka, 1988;
Talmy, 1988a; Verhagen and Kemmer, 1997; Degand, 2001; Wolff and Song, 2003). Causal
connectives, on the other hand, have been described with reference to notions that can hardly be
related to those used for describing causal verbs, for example ‘subjectivity’ (e.g. Pander Maat and
Sanders, 2000; Pit, 2003), or related concepts such as ‘speaker involvement’ (Pander Maat and
Degand, 2001), and ‘mental spaces’ (Dancygier and Sweetser, 2000; Verhagen, 2005).
Inview of the basic assumption underlying the studies mentioned, this is a surprising situation. If
a parallel between semantic categories and conceptual categories holds, we would expect that
similar conceptual models of causality are manifest across different types of linguistic constructions
expressing causality, notwithstanding obvious grammatical differences. In this paper, we argue that
an integrative approach of causality markers of different types is not only possible, but also highly
desirable, if we are to fully understand the mechanisms underlying the linguistic expression of
causality. We aim to make such a contribution to the study of causality, adding two new aspects to
previous discussions. Firstly, we apply cognitive semantic theories and concepts originally
developed for explaining clause-level phenomena, to connectives—linguistic items operating on
the level of discourse structure. Secondly, we test our cross-level similarity hypothesis empirically
against data from language use. We’ll proceed as follows: in section 3 we sketch the parallels
between the linguistic levels and illustrate them with corpus examples of the causal verbs doen and
laten, and the causal connectives daardoor, daarom, and dus. In section 4, we address the question
whether such analytical results have any cognitive relevance. In order to provide some more direct
evidence for the central idea of cross-level parallels, we set out to test the cross-level similarity
hypothesis in an experiment, investigating the intuitions of language users themselves. We believe
that this combination of methods is imperative if we ultimately aim at an integrative theory of
causality markers that is descriptively adequate and cognitively plausible. Before we start reporting
the empirical studies, we present a brief overview of arguments in favor of the ‘cognitive’
assumption underlying the cross-level similarity hypothesis (section 2). We also discuss how the
methodology of converging evidence used in this study contributes to a better understanding of the
relation between linguistic causality markers and cognitive models of causality.
2. Conceptual models in linguistic causality markers
The concept of causality lies at the core of our understanding of how the world functions. The
Oxford English Dictionary defines causality as: ‘‘the operation or relation of cause and effect’’.
Cause is defined as ‘‘that which produces an effect; that which gives rise to any action, phenomenon
or condition’’. Exactly how this ‘relation of cause and effect’ is to be defined has been a matter of
debate from Aristotle’s times until the present day, and seems to depend crucially on the chosen
level of analysis. In this paper we will focus on the way causality is defined in human cognition.
Findings from cognitive psychology and anthropology suggest that every-day human conceptual
N. Stukker et al. / Journal of Pragmatics 40 (2008) 1296–13221298
Author's personal copy
thought concerning causality is organized in models (cf. Michotte, 1963; Piaget and Garcia, 1974;
Piaget and Inhelder, 1969; D’Andrade, 1987). An important characteristic of these conceptual
causality models is that they differ fundamentally from scientific models explaining how the
world functions. The conceptual models structuring every-day thinking represent simplified
theories, reflecting conceptual systems that are rooted in naive physics and psychology
(cf. Lakoff, 1987; Talmy, 1988a). Psychological and anthropological research on causality
suggests that the human mind distinguishes different types of causality. Michotte (1963), for
example, suggests that in the human understanding of causality, categorization into different
types plays a role. These findings are corroborated by anthropological studies like D’Andrade
(1987), who presents evidence suggesting that we invoke an elaborate ‘Folk model of the
mind’ in our every-day thinking about the causing of ‘mental states’ (emotions, feelings,
intentions, etc.). In this conceptual model, too, categorization into different types of mental
states is an important notion.
The idea that conceptual models and the act of categorization play an important role in language
as well, has been present for decades (cf. Rumelhart, 1975; Schank and Abelson, 1977; Fillmore,
1976). It has recently been elaborated specifically in the field of cognitive semantics, a foundational
characteristic of which is the assumption that a direct relation exists between semantic knowledge
and ‘world knowledge’. Cognitive semanticists view language as ‘‘a structured collection of
meaningful categories that help us deal with new experiences and store information about old ones’’
(Geeraerts, 1997:8; see also Langacker, 2000; Haiman, 1980; Lakoff, 1987; Taylor, 1995). An
assumption that is particularly important for the purposes of the present paper is that with respect to
content and structure, linguistic categories are analogous to conceptual categories (that exist
independently from language). Following this line of reasoning, several authors have proposed that
the semantic contrast between causality markers is adequately described with reference to the
conceptual models of causality discussed above; more specifically, to the different categories of
causality within these models (cf. Talmy, 1976, 1988a; Lakoff, 1987; Verhagen and Kemmer,
1997). These authors propose that causality markers function as categorization devices: when
selecting a specific marker among the options available in a language, the language user assigns the
causal relation expressed to a specific type (‘category’) of causal relation. Starting from this
assumption, Talmy (1988a) argues that different types of causality markers should be analyzed with
reference to one invariable conceptual model, namely, the conceptual model of Force Dynamics,
which describes ‘how entities interact with respect to force’. He argues that this model is rooted in
human conceptual understanding of causality and claims that Force Dynamics makes it possible to
generalize over semantic properties of grammatically different types of causal expressions in
English, ranging from causal verbs and prepositions to modal expressions and causal connectives
(Talmy, 1988a:50; see also Degand, 19963).
Talmy’s Force Dynamics theory has been influential until today. We want to investigate
whether other sources than the analyst’s intuitions – introspection is the only source of evidence
in Talmy’s proposal – point in the same direction. Recent insights have shown that the analyst’s
intuitions alone do not always provide a reliable source of evidence for questions concerning
patterns of language use and their relation to conceptual structure (cf. Gibbs, 2006; Sanders and
N. Stukker et al. / Journal of Pragmatics 40 (2008) 1296–1322 1299
3 We limit our discussion to the parallel proposed between conceptual categories and semantic categories, modeling the
contrasts between various markers available for highly similar causal constructions (verbs or connectives). An integrative
approach of causal expressions focusing on causal constructions (prepositions, verbs, and connectives) is proposed by
Degand (1996, 2001). She shows that Talmy’s ‘common causative situation’ underlies all of the constructions investigated,
while they differ in terms of metafunctions and strata as proposed in systemic functional grammar (cf. Halliday, 1985; Dik,
1978).
Author's personal copy
Spooren, 2007). Subsequent studies, based on analyses of corpora of natural language use, show
that marking patterns in specific causality markers are often co-determined by more general
cognitive models (for elaboration of this argument, see Kemmer and Verhagen, 1994; Verhagen
and Kemmer, 1997). Furthermore, like many other cognitive semantic theories, Talmy’s proposal
predominantly focuses on expressions that function within clauses. Causal conjunctions,
typically used to connect clauses at the level of discourse structure, are taken into account mainly
in their non-typical function as prepositional markers (because of).
The studies reported in the present paper aim to extend the integrative perspective on causality
markers as proposed by Talmy beyond the clause-level, and aim to realize it in empirical research.
We make use of the methodology of ‘converging evidence’ (Gries et al., 2005; Sanders and
Spooren, 2007). As a first step, the meaning and use of the causal verbs doen and laten, and of the
causal connectives daardoor, daarom and dus in corpora of natural language use was investigated.
In section 3, we summarize results of the corpus analyses and show that parallels in meaning and use
of causal verbs and causal connectives exist at an analytical level. This first step yields a hypothesis
that has descriptive adequacy. As a second step, the cognitive plausibility of our analytical results
from language use was assessed. The actual usage of linguistic items is expected to contain strong
indications of the knowledge language users have concerning their meaning and use (cf. Langacker,
1987, 2000; Bybee, 1985; Barlow and Kemmer, 2000, and contributions to that volume). However,
more direct evidence regarding the cognitive plausibility of the cross-level similarity hypothesis –
do analytical findings have a parallel in the mind of language users? – can be obtained by testing
hypotheses against language users’ intuitions. Therefore the cross-level similarity hypothesis was
additionally tested in an experimental study, which is reported in section 4.
3. Categorization of causality in Dutch language use
3.1. Causality marking at the clause level: animacy
Verhagen and Kemmer (1997, 1992), Kemmer and Verhagen (1994) investigate the relation
between conceptual models and the lexical semantics of the Dutch causal verbs doen (‘make’)
and laten (‘have’, ‘let’). When analyzing empirical data from actual language use, they note a
strong asymmetry in the ‘participant types’ doen and laten are combined with: doen is typically
used when the causal process relates inanimate elements, while laten is typically used with
animate causal participants. Verhagen and Kemmer relate this patterning to the conceptual model
of Naive Dualism, which captures the fundamental distinction we tend to make in our every-day
thinking, namely between events ultimately originating from some mind, and events that
originate from our inanimate, physical environment (Verhagen and Kemmer, 1997:70–72;
Verhagen, 1995; see also other contributions to Stein and Wright, 1995).
Verhagen and Kemmer observe that doen is predominantly used if the cause-part of the
relation refers to an inanimate entity. According to the model of Naive Dualism, physical entities
are taken to act directly on other entities or persons. This motivates an interpretation as ‘direct
causation’4: doen is used when the activity of the cause-part is conceptualized as causing the
N. Stukker et al. / Journal of Pragmatics 40 (2008) 1296–13221300
4 The notion of ‘directness’ has been used for characterizing causal expressions by other authors as well. It should be
noted, however, that the content of this label varies among studies. In many studies, the notion is used to contrast causal
constructions with varying degrees of syntactic and semantic ‘integratedness’ (cf. Kemmer and Verhagen, 1994). For
example, lexical causatives (‘direct’) are distinguished from other types of causatives (analytical, morphological) which
count as more ‘indirect’ (cf. Givon, 1975; Shibatani, 1976; Wierzbicka, 1988; Shibatani and Pardeshi, 2001; Wolff and
Song, 2003).
Author's personal copy
effected predicate ‘immediately and directly’ (Verhagen and Kemmer, 1997:71–73).
Representative examples5 are (6) and (7):
(6) Belangrijk is dat [‘groene’ sneeuwklokjes] op een ’s zomers koele plaats geplant worden;
warmte doet het blad te vroeg afsterven zodat de bol niet groeit, of geeft gevaar voor
schimmels. It is important that [‘green’ snowdrops] be planted in a summery place that
is cool in summer; heat causes the leaf to die prematurely so that the bulb will not grow,
or makes it prone to fungus.
(7) (From a review of the Dutch magazine Filosofie Magazine). Het wegvallen van allerlei
bindende ideologieen, gepaard aan een ongekend grote keuzevrijheid, doet individuen
zoeken naar hun eigen weg door het bestaan. The fading away of all sorts of uniting
ideologies, coupled with unprecedentedly large freedom of choice, makes individuals
search for their own way through life.
In (6) the ‘heat’ causes ‘the leaf to die’ directly; if the heat is present, nothing can prevent the leaf
from dying. A similar understanding of (7) is likely: when interpreting this situation we assume
that the mere occurrence of the ‘fading away of all sorts of uniting ideologies’ and the
‘unprecedentedly large freedom of choice’ bring about the ‘searching for their own way’
unavoidably; there is nothing the ‘individuals’ can do to avert the effect.
Laten, on the other hand, is almost exclusively used in constructions with animate causes
(Verhagen and Kemmer, 1997:65). The meaning and use of laten are best described as marking
‘indirect causation’—see examples (8) and (9):
(8) Jari Litmanen komt de komende tweeenhalf jaar uit voor Liverpool. [. . .] De Fin krijgt
medewerking van Barcelona, dat hem ondanks een contract tot 2002 transfervrij laat
vertrekken. Jari Litmanen will play for Liverpool for the next two and a half years[. . .]..
The Finn received cooperation from Barcelona, which despite a contract valid until
2002, lets him leave without transfer.
(9) [From an interview with dancer and repetitor Tatiana Leskova] Omdat de techniek van de
huidige dansers beter is geworden, mag en moet je een oud ballet daaraan wel aanpassen.
[. . .] Dus laat ik de enkele pirouettes van destijds nu dubbel draaien. Because the skills
of today’s dancers have improved, you may and you must adapt an old ballet to the
circumstances. [. . .] Therefore, nowadays I have the single pirouettes from those days
turned twice.
According to the model of Naive Dualism, animate beings can only act on other animate beings
via the intervening physical world. In other words: it is not possible to reach into another person’s
mind and directly cause them to act (Verhagen and Kemmer, 1997:72–73). This is exactly the
interpretation that fits laten-marked causative constructions. In (8), the soccer club ‘Barcelona’ is
initiating the causal process, but cannot control its complete course; the effect of ‘him’ ‘leaving’
will take place only if an intermediary factor, associated in this example with Litmanen’s
intentions to join a different club, come into play. Hence, this intermediary force is seen as the
N. Stukker et al. / Journal of Pragmatics 40 (2008) 1296–1322 1301
5 The fragments discussed in section 3 are examples from natural language use, taken from a corpus of newspaper texts
(the Dutch daily newspaper Trouw, electronically available from Factlane (LexisNexis Nederland)—see Stukker, 2005).
Author's personal copy
force most directly involved in bringing about the effect. A similar interpretation holds for (9):
the causer ‘I’ induces the causal process, but the effect of ‘pirouettes turned twice’ only occurs
because of an activity of the dancers who perform the pirouettes.6 Verhagen and Kemmer define
Indirect Causation as ‘‘a situation that is conceptualized in such a way that it is recognized that
some other force besides the initiator is the most immediate source of energy in the effected
event’’ (Verhagen and Kemmer, 1997:67). The prototypical usage types7 of doen and laten can be
characterized as in Fig. 1.
This patterning – doen with inanimate causal participants and laten with animate ones –
characterizes the vast majority of usage-contexts of Dutch causal verbs (Verhagen and Kemmer,
1997; Degand, 2001; Stukker, 2005). Apart from these prototypical usage contexts, doen and
laten occur in contexts that are ‘non-standard’ in terms of participant configurations. An example
is (10), a causal process containing animate participants, laten’s prototypical usage context,
marked with doen.
(10) (From a book review) Askew raakt meer en meer de weg kwijt [. . .] Kit weet contact
met hem te krijgen en hem te doen inzien dat hij [. . .]de sterke schouder kan zijn
waarop zijn disfunctionele gezin kan leunen. Askew is losing his way more and more
[. . .] Kit manages to stay in contact with him and to make him see that he [. . .] can
be a strong shoulder for his dysfunctional family to lean on.
These exceptions to the rule, however, should not be interpreted as counter-examples to the idea
that causal verbs are directly related to the conceptual model of direct causation. What these
conventional contexts have in common, is that the effect of the markers are best explained with
reference to the prototypical usage contexts themselves. Corpus data suggest that ‘doen + ani-
mate participants’ (see (10)) only occurs in contexts which demonstrably allow for an inter-
pretation as direct causation, and these are contexts where doen brings about a rhetorical effect
that fits the overall interpretation of the fragment within its context well (Verhagen and Kemmer,
1997:73–77; see also Stukker, 2005:63–67). The effect ‘see’ in (10), for example, may or may not
be intended by the causer ‘Kit’, but the occurrence of the process is ultimately beyond her
complete control. The extent to which the causee, the experiencer of the causal effect ‘him’,
controls the coming about of the effect ‘see that he can be a strong shoulder for his dysfunctional
N. Stukker et al. / Journal of Pragmatics 40 (2008) 1296–13221302
Fig. 1. Prototypical usage-types of doen and laten.
6 Note that in this respect Dutch laten differs from equivalents in other languages, such as English let, and French
laisser. The latter can only be used to mark ‘permissive’ or ‘enablement’ relations, in which the effect occurs because of
an inherent tendency of the causee, e.g. in example (8) forces associated with ‘him’ (Talmy, 1988a,b, 2000; Verhagen and
Kemmer, 1997:68). Dutch laten, on the other hand, can also be used to mark outright causal processes, defined as
situations in which the causer ‘forces’ the causee to carry out an activity they are not inherently inclined to perform,
cf. (9). As a consequence, laten’s meaning is relatively schematic, and best captured by the notion of ‘indirect causation’.
For a more extensive discussion, see Verhagen and Kemmer (1997:69–70).7 In line with the usage-based approach to language (cf. Langacker, 1987; Barlow and Kemmer, 2000, and contributions
to that volume; Bybee, 2006), the term ‘prototypicality’ is used here with reference to the level of usage of the
constructions. The degree of prototypicality is determined on the basis of usage frequency (for a more elaborate
discussion, see Stukker et al. (in prep.)).
Author's personal copy
family to lean on’ is somewhat ambiguous, depending on contextual factors. These factors
explain why doen fits the context of (10) so well: its effect can be characterized as backgrounding
Askew’s (factually present) contribution to bringing about the effect. This analysis fits the most
obvious interpretation of this specific causal process, namely, that in (10) the causal effect comes
about more or less against Askew’s intentions; it is not likely that he contributed to it actively. In
other words, an interpretation of (10) as an instance of direct causation, construed as a process
with a cause-participant that can act directly on the effect-participant, rather than indirect
causation, is licensed by specific and demonstrable context factors.
This type of ‘exploitative usage’ is relatively rare, but not uncommon (on doen see also
Verhagen, 2000; examples of exploitative usage of connectives will be discussed in section 3.2).
It has been explained as a normal fact of language use and it fits in with assumptions held in many
branches of linguistics, namely the assumption that linguistic utterances are relatively
‘underspecified’: the mental representation built from an utterance is a result not only of explicit
linguistic signals, but it is also influenced by the language user’s knowledge of the specific
context of use. Individual linguistic elements in the utterance give minimal, but sufficient clues
for finding the domains and principles appropriate for building the mental representation
(Fauconnier, 1994:xviii; see Verhagen, 1997 for a more elaborate discussion of the necessity of
context-sensitivity of language; see discussion of implications of this idea for theories on
coherence in Sanders and Spooren, 2001). The rhetorical function of doen in the context of (10)
can be explained with reference to a basic tenet in cognitive semantics, viz. that an expression’s
meaning is not just ‘‘an objective characterization of the situation described’’. Equally important
is how the speaker chooses to ‘construe’ the situation and portray it for expressive purposes;
language use is inherently ‘perspectivized’ Langacker (2002:315, 1987); cf. Talmy’s (1988b,
2000) ‘imaging systems’). Evidently, causality markers do not function as mere ‘plug-ins’, able
to bring about any interpretation in any context; a minimum amount of congruence between a
linguistic element’s meaning and the context it is used in, is required.
It is along these lines that the marking of (10) with doen instead of laten can be explained. The
ambiguous context allows for different construals in terms of directness of the causal process.
Where laten would have favored a construal as indirect causation, with a relatively autonomous
role for the causee Askew—marking with doen highlights the non-intentional aspects present in
this specific context. In itself, the exploitative usage of doen underlines its function as a
categorization device assigning the causal process to the conceptual category of direct causation,
construing it as a causal relation between inanimate elements. The fact that contexts of use may
be ambiguous for ‘causality type’ plays a crucial role in the experiment reported in section 4.
3.2. Causality marking at the discourse level: subjectivity
The meaning and use of the Dutch causal connectives daardoor (‘because of that’), daarom
(‘that’s why’) and dus (‘so’) have been characterized with reference to the cognitive concept of
‘subjectivity’. This concept is rooted in the human cognitive ability to relate information to a
speaking or thinking subject of consciousness (Sanders and Spooren, 1997; Verhagen, 2005) and
may be defined as language users’ ability to express themselves and to adopt other people’s
perspectives (Pit, 2005:26). In linguistic theory, subjectivity has been defined as ‘speaker
foregrounding’, or: ‘the speaker’s self expression’ (Langacker, 1987, 2002; Lyons, 1995).
Presence or absence of speaker foregrounding in the causal relation is commonly assumed to
be the major dimension determining the distribution of daardoor, daarom and dus (Pander Maat
and Sanders, 2000, 2001; Pander Maat and Degand, 2001; Pit, 2003; Stukker, 2005). The degree
N. Stukker et al. / Journal of Pragmatics 40 (2008) 1296–1322 1303
Author's personal copy
of speaker foregrounding in causal coherence relations can be analyzed with reference to the
subject of consciousness (SOC; Lyons, 1995; see for a similar concept Langacker, 1990), defined
in this context of discourse causality as ‘‘a person whose intentional acting is seen as the ultimate
source of the causal relation’’ (Pander Maat and Sanders, 2000:64).8 Subjectivity then is defined
as the conceptual distance between the speaker and the SOC responsible for the causal relation
(Pander Maat and Sanders, 2000:77). If the SOC coincides with the present speaker, the causal
relation is maximally subjective. Subjective relations are typically marked with dus, and they are
typically relations where an (implicit) speaker SOC performs an act of reasoning (Pander Maat
and Sanders, 2000; Stukker, 2005). This type of causal relations is commonly referred to as
‘epistemic causal relations’ (cf. Sweetser, 1990). An example is (11), where the causal relation is
constructed between the situation that both the speaker and the addressee have a position of
power, and the speaker SOC’s conclusion that they ‘can talk on the same level’.
(11) (The editor in chief of the radio show, With the Eye on Tomorrow tells about a
letter he once received from a listener.) ‘‘De goede man schreef: ‘Mijnheer, u bent
de baas van Het Oog, en ik ben de baas van mijn vrouw. We kunnen dus op niveau
praten.’’’ Waarop de vraag volgt of Van Hoorn nu eindelijk iets kan doen aan die
vermaledijde begintune. ‘‘The good man wrote: ‘Sir, you are the boss of The Eye,
and I am the boss of my wife. Therefore we can talk on the same level.’’’ After this,
the question follows whether Van Hoorn can finally do something about that cursed
opening tune.
Daarom and daardoor, on the other hand, are prototypically used in contexts where the causal
relation is constructed between two situations in observable reality (‘content causality’,
Sweetser, 1990). In these cases, the causal relations are, as a whole, reported by the speaker,
who is not, contrary to epistemic causal relations, by default the relation’s SOC. In these contexts,
the distance between speaker and SOC is relatively large; accordingly, they are commonly
categorized in language use as ‘objective causal relations’. Within the category of objective causal
relations, Dutch connectives make a distinction between ‘intentional causal relations’ with an agent
SOC (in (12), zij ‘they’), typically marked with daarom, and ‘non-intentional causal relations’
without an SOC-role. The latter type are relations of ‘pure cause’, constructed between physical or
uncontrollable mental processes, where human intervention does not play a role, typically marked
with daardoor (Pander Maat and Sanders, 1995; Stukker, 2005); an example is (13).
(12) Ook Koert Bakker en Jessica Gysel van ‘relatiebemiddelingsbureau’ Fanclub menen
dat de traditionele reclame klinisch dood is. Daarom organiseren zij voor Adidas
hiphopparty’s en straatvoetbalfeesten. Koert Bakker and Jessica Gysel from Fanclub
PR Service also claim the traditional advertisement is as good as dead. That’s why
they are organizing hiphop parties and street soccer activities for Adidas.
(13) De [schaatser Frans de Ronde] omschreef de Jaap Edenbaan als een grote kattebak.
‘‘Overal lag zand. Daardoor 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. Because of this, many skaters skated with burrs on their
blades[. . .]’’.
N. Stukker et al. / Journal of Pragmatics 40 (2008) 1296–13221304
8 For equivalent concepts and definitions, see Pander Maat and Degand (2001) and Pit (2003).
Author's personal copy
In (13) one situation ‘in the real world’ leads to another: the presence of sand on the skating rink
causes the burrs on the skates’ blades, without intervention of a human intentional agent. By
contrast, in causal relations with an agent SOC the cause-part of the relation presents a motivation
the agent may have had for performing the action reported in the effect-part of the causal relation.
For example, in (12) the perception that ‘traditional advertisement is as good as dead’ leads to the
intentional act performed in the real world of ‘organizing hiphop parties and street soccer activities
for Adidas’. The prototypical usage types of dus, daarom and daardoor can be characterized as in
Fig. 2.
Apart from these prototypical usage types (see discussion in section 3.1), dus, daarom and
daardoor may occur in contexts that are linguistically construed differently. Examples are (14)
and (15):
(14) (The first sentences from a review of the book ‘Bird history of Amsterdam’)
Nederland verstedelijkt. Daarom is het niet vreemd dat ook de stadse natuur steeds
vaker aandacht krijgt van natuurvorsers. The Netherlands is becoming urbanized.
That’s why it is not at all odd that urban nature is getting more attention from
naturalists.
(15) (Bystanders rush to help out at the Volendam pub fire.) ‘‘Ik woon vlakbij, dus ik ben
brandwondencreme gaan halen.’’ ‘‘I live nearby so I ran to get burn ointment.’’
Fragment (14) contains an epistemic causal relation. The situation that ‘the Netherlands is
becoming urbanized’ serves as an argument for the conclusion that ‘it is not at all odd that urban
nature is getting more attention from naturalists’. This causal relation evidently contains an
implicit speaker SOC: the speaker constructs the causal relation by presenting the situation in the
first sentence as an argument for the conclusion presented in the next sentence. This is a typical
usage-context for dus; however, (14) is marked with daarom. Corpus data suggest that daarom is
used instead of dus in epistemic causal contexts where the speaker appears to background his or
her SOC-role as a concluder in the causal relation (Stukker, 2005:126–131; Stukker et al., in
prep.). They differ for example in degree of ‘performativity’: while the conclusions in dus-
marked epistemic relations always concern ‘new’ knowledge, daarom may also be used to mark
epistemic relations with ‘conclusions’ that either concern knowledge that is contextually
available and already accepted,9 or conclusions that are not regarded as ‘disputable knowledge’.
Fragment (14) is of the latter type. The reviewer quotes from the book being reviewed; therefore
not the reviewer herself is responsible for the conclusion presented in the second sentence, but the
N. Stukker et al. / Journal of Pragmatics 40 (2008) 1296–1322 1305
Fig. 2. Prototypical usage-types of daardoor, daarom and dus.
9 An example where the ‘givenness’ of the conclusion is linguistically marked (with zo ‘so’), is: ‘‘Maar bij ernstige
brandwonden is het hele lichaam ziek’’, zegt Hermans ‘‘De lever, de nieren, alle organen doen mee. Daarom/#dus is de
zorg voor deze patienten zo ingewikkeld.’’ ‘‘Serious burns make the whole body ill’’ says Hermans. ‘‘The liver, the
kidneys, all organs are affected. That’s why/#so the care for these patients is so complicated’’. In these cases, marking
with dus (‘so’) is ungrammatical (Stukker, 2005:129–131).
Author's personal copy
author of the book is. In line with this interpretation, daarom signals that the speaker must not be
construed as the causal relation’s SOC, the person responsible for this particular causal relation.
The opposite strategy, namely rhetorical speaker foregrounding, occurs in causal relations
containing an explicit actor SOC, a context where daarom is the default marker, marked with dus.
Corpus data suggest that dus is used instead of daarom in these contexts if the context contains an
argumentative flavor (Stukker, 2005:119–126; see also Pander Maat and Sanders, 2000:71–74
presenting a similar analysis), and, in this respect, resembles an epistemic causal relation. In (15)
the relation as a whole is presented from the perspective of an embedded speaker (cf. Sanders and
Spooren, 1997), the interviewee, which is signaled by the speaker of the text with quotation
marks. This embedded speaker does not only convey the motivation for her action ‘to run to get
burn ointment’, but also seems to argue that this was exactly the right thing to do under the
circumstances given, thus invoking the conventional action schema (cf. ‘topos’, Anscombre and
Ducrot, 1983): ‘if in case of emergency some tool is needed, the person that has easiest access to
it must go and get it’. The examples of non-prototypical usage of connectives discussed here are
not exhaustive, but they illustrate an important point: similar to what was reported for causal
verbs (see section 3.1), usage-types of causal connectives that diverge from the prototypical
usage configurations can be interpreted as speaker construals serving rhetorical purposes. With
connectives as well, the effects of these construals is adequately analyzed with reference to the
prototypical usage-context of the connective itself (Stukker, 2005:117–131; Stukker et al., in
prep.). In other words, just like causal verbs, connectives can be used exploitatively whenever a
specific context of use is sufficiently ambiguous with respect to causality type.
3.3. Cross-level parallels in Dutch linguistic categorization
The discussion in the previous sections suggest that, despite obvious constructional
differences, important parallels exist between the way language users categorize causal relations
at the clause-level with causal verbs, and the way they categorize causal relations at the
discourse-level with causal connectives—at least from an analytical point of view. First, in both
types of expressions it is the specific nature of the interaction between the cause factor and the
effect factor that determines categorization. A further parallel exists in the way these force
interactions are conceptualized. Both in verbs and in connectives, it is the distinction between
animate and inanimate entities that determines categorization of causal events. Or more
precisely, in causal relations marked with a connective, it is the presence or absence of an SOC,
and in causal relations marked with a causal auxiliary verb it is the type of causal participants
(causer or causee) that determines categorization (see Stukker, 2005:chapter 7, for a more
elaborate discussion). In both types of causality markers, then, we see reflected the conceptual
model of Naive Dualism. At this point, the following hypothesis on cross-level parallels between
causal verbs and causal connectives in Dutch can be formulated:
Cross-level hypothesis
I: The clause level causality marker doen parallels the discourse level marker daardoor in
marking ‘inanimate causality’. II: The clause level causality marker laten parallels the
discourse level markers daarom and dus in marking ‘animate causality’.
Yet, the parallel between verbs and connectives is not perfect. An important difference is that
connectives allow for expressing more causality types than verbs do: dus, prototypically marking
subjective causality, does not seem to have an equivalent causal verb at the clause-level. For this
N. Stukker et al. / Journal of Pragmatics 40 (2008) 1296–13221306
Author's personal copy
moment we will leave this question aside and focus on daardoor, daarom, doen and laten only,
but we will come back to this question in section 5.
4. Experiment: do language users experience parallels in causal verbs and discourseconnectives?
The corpus analyses reported in the previous section yielded a hypothesis that has descriptive
adequacy. But do the analytical conclusions on cross-level similarities have any cognitive
relevance?
In line with the methodology of converging evidence (see the discussion in section 2), the results
of the corpus analyses will be supplemented by experimental evidence. The remainder of this paper
reports an experiment testing language users’ intuitions on conceptual parallels between causal
verbs and causal connectives. In other words, the theoretically motivated hypothesis is tested in
speakers’ own terms. A further advantage of experimental testing is the possibility of manipulating
text material so as to put the comparability of causal verbs and causal connectives on edge.
4.1. Experimental task and hypotheses
The aim of the experiment is to investigate whether categorizations that language users make
with doen and laten parallel the categorizations they make with daardoor and daarom in situations
of ‘real’ language use. The general claim under investigation is: Language users experience
similarities between the causal verb doen and the causal connective daardoor; and between the verb
laten and the connective daarom. Participants were asked to paraphrase intra-clausal causal
relations marked with either doen or laten with an inter-clausal paraphrase of the relation, marked
with daardoor or daarom.10 For a number of reasons we chose to direct performance on the task
somewhat by prefabricating usage-contexts and response options, rather than using, e.g. an ‘open’
elicitation task. A prestructured task is believed to reduce the complexity of a task which in itself
demands considerable skills in abstraction and metalinguistic reasoning. Another advantage was
taken to be that homogeneity of the output of the experiment was favored. Finally, when using
standardized answering options, it was expected that interpretative role of the analysts was reduced
to a minimum. An example of an item used in the experiment is (16)11:
(16) Zaterdagavond stond het openluchtfestival van Blokzijl op het punt te beginnen.
[Enkele spetters regen]cause deden [de organisatoren het ergste vrezen]effect.
On Saturday night, the open air festival of Blokzijl was about to start
[A few drops of rain]cause made [the organizers fear for the worst]effect.
Paraphrase
[Er vielen enkele spetters regen]cause.
[Some drops of rain fell]cause.
a. Daardoor [vreesden de organisatoren het ergste]effect.
Because of that [the organizers feared for the worst]effect.
N. Stukker et al. / Journal of Pragmatics 40 (2008) 1296–1322 1307
10 This experimental design was inspired by Sanders and Verhagen (1996). However, the exact purposes of their test and
the way items were constructed differ from the version presented here.11 Underlining is added here only for ease of presentation; in the layout of the real experiment it was of course absent.
A complete list of items used in the experiment is presented in Stukker (2005:appendix 6-3) (accessible online via http://
igitur-archive.library.uu.nl/dissertations/2006-0428-200107/app.pdf).
Author's personal copy
b. Daarom [vreesden de organisatoren het ergste]effect.
That’s why [the organizers feared for the worst]effect.
In each item, a causative sentence was marked with a causal verb, either doen or laten. This
sentence was followed by two inter-clausal paraphrases of the same causal relation. The para-
phrases differed in only one respect: they were marked with either the connective daarom or
daardoor. Subjects were asked to choose the paraphrase that, according to their intuitions,
resembled the original intra-clausal causal relation best. The hypotheses tested are:
Hypotheses
I: Language users prefer a paraphrase marked with daardoor when the original causative
constructions are marked with doen. II: Language users prefer a paraphrase marked with
daarom when the original causative constructions are marked with laten.
4.2. Construction of items
The material used in this experiment had to meet a number of demands. In order to enhance
representativity and generalizability, usage-contexts were selected from corpora of natural
language use.12 Ideally the sample would have been built proportionally from prototypical and
non-prototypical usage-contexts (cf. section 3). However, this was prevented by the experimental
design chosen. There is specifically one demand that places considerable restrictions on the
material to be used, namely that in the inter-clausal paraphrases, alternative markers (daarom or
daardoor) fit equally well. Obviously, the task ‘choose the paraphrase that resembles the causal
sentence best’ is credible only if the paraphrases themselves don’t differ in quality, and are
perfectly ambiguous in the given context. The selected usage contexts must prevent that subjects
make their choice for one of the alternatives on the basis of differences in appropriateness of the
markers in those contexts. Therefore, only contexts of use were included in the experiment
that were neither markedly ‘intentional causal’ nor ‘non-intentional causal’; these are the
non-prototypical usage contexts (cf. section 3.2).
Two types of contexts were employed to ensure this type of ambiguity. The first type contains
an effect denoting predicate that refers to an intentional act, combined with a modifier that
possibly (but ambiguously) amends the intentional aspect. An example is (17):
(17) Het hoorspel ‘War of the worlds’ maakte in het jaar 1938 heel wat emoties los. De
nieuwsberichten over marsmannetjes waren nogal realistisch, en daarom/daardoor
renden de mensen in paniek de straat op. In the year 1938 the radio play ‘War of
the worlds’ caused a lot of fuss. The news messages about Martians were rather
realistic, and that’s why/because of that people ran into the street in panic.
In (17), the effect-denoting predicate renden ‘ran’, that in itself is inherently intentional, is
modified by in paniek ‘in panic’ The mental state of ‘panicking’ is by definition non-intentional.
Because of the juxtaposition of these two elements, the effect-denoting sentence has an
ambiguous reading. It can be interpreted focusing either on the intentional aspect or on the
non-intentional aspect (cf. the discussion in section 3). In these contexts both daardoor and
daarom can be used in a natural way.
N. Stukker et al. / Journal of Pragmatics 40 (2008) 1296–13221308
12 Mainly newspaper texts, see Stukker (2005:177) for an overview.
Author's personal copy
The second strategy consisted of selecting effect predicates denoting a ‘mental state’ that is
ambiguous to the degree in which the ‘experiencer’ can control its occurrence, such as belief
states and feelings (cf. D’Andrade’s (1987) Folk model of the mind). An example is (16) above,
containing an animate locus of effect which is ambiguous for SOC-hood. The causal effect
‘fearing’ may or may not be brought about intentionally. From the viewpoint of adequacy,
daardoor and daarom fit this context equally well. Evidently, subtle meaning change occurs
when alternating the markers. Daardoor backgrounds the degree of ‘control’ the locus of effect
may have had in bringing about the effected event, while daarom foregrounds this aspect.
Of course we expected that it is precisely this subtle meaning difference that determines subjects’
choice; we expected that they would prefer the daardoor paraphrase if the preceding sentence
was marked with doen, and daarom if the preceding sentence was marked with laten.
A further concern was to construct two-clause paraphrases in such a way that elements
essential in determining the choice for a causality marker in either analytic causatives or causal
coherence relations be left unchanged. Because of the grammatical differences between analytic
causative constructions and causal coherence relations, special care must be taken to ensure that
structure and content of the two-clause ‘paraphrases’ correspond to the simple clause ‘originals’
at crucial aspects. We proceeded as follows: It was assumed that in both construction types, the
causality category (intentional or non-intentional) is determined as a result of the nature of the
interaction between the causal participants (see the discussion in section 3). This implies that
when constructing an inter-clausal paraphrase out of an analytical causative construction, it is
important that the ‘point of application’ remain constant from the perspective of the causal effect,
as the nature of the causee in analytic causatives and the SOC (or non-SOC) in causal coherence
relations determines categorization (see section 3) to an important extent. Generally this is not a
problem, since the only conversion needed is to replace the infinitival form in the analytic
causative construction by the corresponding finite form in the effect clause of the inter-clausal
causal relation (cf. (16) in which infinitival vrezen ‘fear’ from the causal sentence is replaced by
the finite form vreesden ‘feared’).
It is equally important, for the same reason, that the ‘point of application’ in the causal relation
doesn’t change essentially from the perspective of the cause-part. In extending the (nominal)
causer to a complete clause, adding extra elements to the causal chain should be avoided. We
made use of constructions in which mentioning of the causer was accompanied by different kinds
of specifying information in modifier (adjectival and adverbial) phrases. In the analytic causative
construction in (16), for example, the causal relation is constructed between the causer ‘drops of
rain’ and the causee ‘the organizers’ who ‘feared for the worst’. The cause-part consists of only a
noun phrase: the noun spetters ‘drops’, complemented with the modifiers enkele ‘some’ and
regen ‘(of) rain’. In the inter-clausal paraphrase, this cause-part is extended into a complete
clause by adding the verb vielen ‘fell’ and Dutch presentative er. An important difference lies of
course in the switching of a noun phrase (part of proposition) to a situation (complete
proposition), but the change in conceptualization that is brought about is limited to ‘zooming out’
on the causal process: the content remains constant, it is only the relative attention different
elements receive that has changed.
4.3. Statistical analysis
Statistical analysis was carried out with an itemized, one sample T-test. This test compares
means per item with ‘chance’, which would be the expected score value if subjects’ choice of
paraphrase were not influenced by the preceding causal verb (=H0). Each time an individual
N. Stukker et al. / Journal of Pragmatics 40 (2008) 1296–1322 1309
Author's personal copy
response was in accordance with H1, a score ‘1’ was allotted. If an individual response was
contradicting H1, score ‘0’ was allotted. The effect per item was computed on the sum of
individual scores. The direction of the effect is indicated by the value of these mean scores: if it is
>.5, participants that responded in accordance with our hypotheses outnumber the ones who
didn’t; if it is<.5, it is the other way round. The level of acceptance was set to 5% ( p = .05). Since
the T-test output is two-tailed and the hypothesis tested in this study is one-sided, a value of
p < .05 does not automatically entail that the results of the item involved should be regarded as
‘in accordance with hypothesis’; this can only be concluded if p < .05 and the mean score >.5.
4.4. Pilot experiment
The experimental design was first tested in a pilot experiment.13 Five doen-marked items
and five laten-marked items were tested. The items were constructed following the procedure
sketched above. ‘Ambiguity’ of the paraphrases with respect to appropriateness of both daarom
and daardoor as markers of the causal relation was evaluated intersubjectively by independent
judges. In order to distract subjects’ attention from the experiment’s goal, one third of the
experiment consisted of fillers. Participants were 88 first-year students of Dutch Language and
Culture and 9 students of other Language studies at Utrecht University attending a course on
Text analysis, and 42 second-year students of Speech Therapy at the school of Higher
Vocational Education in Rotterdam (HRO) attending a course on statistics. Hardly any effects
were found in this version of the experiment. These findings were unexpected and contradicted
previous findings (Sanders and Verhagen, 1996). Careful scrutiny of both the test instrument
and the experimental procedure led to the tentative conclusion that the unexpected results
may have been caused by a ‘mismatch’ between the participants in the experiment and the level
of abstractness of the task. The rather complex experimental task probably demands
experimental participants with considerable abstract-analytical capacities, which the students
participating in our experiment lacked. Several participants reported finding the task ‘difficult’.
Furthermore, post hoc analysis suggests that we did not succeed in constructing perfectly
ambiguous contexts for each of the items (see Stukker, 2005:186–192 for further details).
These facts in combination may form an explanation for indications we have, that a
considerable number of participants did not carry out the task the way it was intended. They
rather appeared to compare the paraphrases with each other and pick out the ‘nicest one’, the
one that was the best expression in the text, instead of comparing each paraphrase with the
preceding causal sentence and picking out the paraphrase that ‘resembled the original sentence
best’, as they were instructed to do. Hence, we decided to modify the experiment in two
respects: participants with better developed analytic skills were selected, and the quality of the
test instrument with respect to ambiguity of the paraphrases was ameliorated.
4.5. Experiment
The experimental design of the pilot version was basically maintained. In order to reduce error
variance as much as possible, different types of items were used. Thirty items were tested; Fig. 3
presents an overview.
N. Stukker et al. / Journal of Pragmatics 40 (2008) 1296–13221310
13 It was prepared and carried out as a part of the master’s thesis of Van Maaren (2002). A complete list of items used in
the pilot experiment is presented in Stukker (2005:appendix 6-1) (accessible online via http://igitur-archive.library.uu.nl/
dissertations/2006-0428-200107/app.pdf).
Author's personal copy
All of the items were newly constructed. A first adjustment to material construction was the
character of the items. Twenty items were constructed following the procedure described in
section 4.2. These will be referred to as ‘natural text items’. In an attempt to test the hypothesis
more directly, a new type of items was added: the ‘chameleon items’ (cf. Sanders, 1997). These
items differed from the ‘natural text items’ in that their usage context was ambiguous not only
with respect to the use of daarom and daardoor in the paraphrases, but also with respect to the
causal verbs: The analytic causative constructions in these items could be marked with both doen
and laten. An example is (18):
(18) [Tijdens een persconferentie lichtte de Minister van OCW de voorgestelde bezuinigingen
toe.]
De kritisch doorvragende journalisten deden/lieten hem vrezen voor de weergave van
zijn plannen in de media.
[During a press conference the Minister for Education presented the proposed
cut-backs.]
The persistently critical journalists made/had/let 14 him worry about the representation
of his plans in the media.
N. Stukker et al. / Journal of Pragmatics 40 (2008) 1296–1322 1311
Fig. 3. Item types.
14 The ambiguity in the Dutch version may not be present in English, acceptability of have or let in this context is
debatable.
Author's personal copy
Paraphrase
De journalisten bleven kritisch doorvragen;
The journalists continued to pose critical questions;
a. daardoor vreesde hij voor de weergave van zijn plannen in de media.
because of that, he worried about the representation of his plans in the media.
b. daarom vreesde hij voor de weergave van zijn plannen in de media.
that’s why he worried about the representation of his plans in the media.
The purpose of this ‘double ambiguity’15 was to test the hypothesis in the same usage context in two
directions, as follows. Each ‘chameleon item’ was inserted in the test instrument twice, once
marked with doen and once marked with laten. If H1 holds, a chameleon item marked with doen
leads to an overall preference of the daardoor-paraphrase. If, on the contrary, the same chameleon
item is presented with laten, expectations are that in paraphrasing, preferences will shift to daarom.
Findings like these would constitute even more direct and fine grained evidence for H1.
One important drawback of ‘chameleons’, however, is that they don’t occur in real language
use very frequently. Consequently, not all of the chameleon items included in the material are
natural examples. The best ‘chameleon candidates’ were causative constructions in which a
rather typical laten-context had been marked with doen. This the type of ‘exploitative use’ of
doen discussed in section 3.1. Because of this specific character of the chameleons (animate
causer, animate causee), we were quite sure that laten fitted the context as good as doen would,
therefore the ambiguity of these causative constructions was not tested.
All of the inter-clausal paraphrases (i.e.: all test items, leaving out the preceding intra-clausal
relation) were pretested for ambiguity. They were presented in print as ‘‘causal text fragments’’ to
thirteen test subjects representative of the test population. Participants were asked to mark the
connective that, in their opinion, fitted the context best. It was assumed that the more ambiguous a
given context was, the more dispersed the judgments of the group of subjects as a whole would
be. We found indeed that with some items, preferences were distributed more or less evenly,
while others appeared to give rise to a clear collective preference for one of the two options. The
latter items were either revised or replaced by other contexts. Before including them in the test
instrument, the resulting new paraphrases were pre-tested following the same procedure over
again.
Finally, in order to facilitate the performance of the experimental task, special attention was
given to ‘interpretability’ of the test items. This was done both with respect to the causal relations
reported, and with respect to the text fragments as a whole. Only items expressing familiar causal
schemes were included. To further facilitate interpretation, each item was introduced by a single
sentence adding some context for interpreting the causal relation to follow, see for example (18).
Information that was part of the original material for the ‘natural text items’ but was not essential
to interpreting the causal process (mainly modifiers), was left out. In order to distract subjects’
attention from the experimental goal, about half of the material consisted of fillers.
N. Stukker et al. / Journal of Pragmatics 40 (2008) 1296–13221312
15 One reviewer noticed that the doen-version of item in (18) may suggest an interpretation of ‘‘the questions of the
journalists’’ as the causer. Although this would strictly speaking be incorrect (journalists being the head of the nominal
phrase), this interpretation is interesting because it implies an adaptation to the meaning of doen as a causality marker,
illustrating the ‘construal’ function of causality markers, described in section 3.1. In this specific context, doen evokes an
interpretation of the ‘journalists’ causing the ‘fearing’ immediately and directly, as induced by a physical law (cf. the
discussion of (6), (7) and (10) in section 3.1). In an interpretation along these lines, physical aspects of inherently animate
participants in the process become foregrounded. It is exactly this type of variation in interpretation according to the
marker chosen, that is exploited in this experiment.
Author's personal copy
4.6. Participants and procedure
With the results of the pilot experiment in mind, advanced university students were selected as
test subjects in the final version of the experiment. The experiment was presented as an integrated
part of an advanced BA3-level course on cognitive processes of reading and writing. As an
obligatory preparation for taking part in this course, all students had previously attended a course on
text analysis. Therefore, all of the subjects were familiar with the idea that sentences in a text are
connected by ‘coherence relations’, and with the idea that different types of causal relations may
exist. Furthermore, all of the subjects were advanced level students either in Dutch Language and
Culture or in Liberal Arts specializing in Language and Communication. Unlike the participants in
the pilot-experiment, these participants can be expected to be more experienced analysts of
language, and more proficient in performing an abstract analytical task such as the one at hand.
Approximately 40 students enrolled for the course. The task was split up into three parts because
of the heavy workload. The experimental sessions were held at fixed moments during three of the
four seminars scheduled. Attendance per session fluctuated between 25 and 30 persons total. Each
session lasted about 15–20 min. Preceding the first session, a spoken instruction was provided by
the experimentator. A summary of this instruction was attached as a title page to each copy of the
experiment; participants were asked to reread it each session before they started working on the
task. In the written as well as in the oral instruction, the need to compare the inter-clausal
paraphrases to the intra-clausal version of the causal relation was emphasized.
4.7. Results
The results were analyzed per cluster of items (cf. Fig. 3), they will be discussed accordingly.
4.7.1. Natural text items
Table 1 summarizes the findings for the doen-marked items. Though the exact strength varies
over items, overall the doen-items show a strong tendency in the expected direction. Eight out of
nine items have a mean score value >.5. This indicates that in these cases a majority of subjects
judged in accordance with the hypothesis tested, which predicted that the daardoor-paraphrase
resembled the original doen-marked sentence best. Five out of these eight items show a (strong)
effect.
How should these findings be interpreted? In order to determine their significance, the
probability that the present findings must be attributed to chance can be estimated. In other
N. Stukker et al. / Journal of Pragmatics 40 (2008) 1296–1322 1313
Table 1
Results hypothesis ‘doen parallels daardoor’: natural text items
Item code Degree of correspondence t (d.f.) p
1 .23 �3.195 (25) .002
2 .59 .961 (26) .173
3 .84 5.044 (30) <.001*
4 .77 3.592 (30) <.001*
5 .54 .386 (25) .352
6 .81 3.904 (25) <.001*
7 .77 3.195 (25) .002*
8 .70 2.275 (26) .016*
9 .81 .900 (25) .189
Significant results are indicated with *.
Author's personal copy
words: we have to compute the probability that H0 is true while finding ‘good’ and ‘bad’ results in
the present proportion. To that end, a binomial test was carried out. It turned out that the chance
that H0 holds in the present situation approaches zero ( p = .00003). This result indicates that the
findings for doen-marked natural text items can be interpreted as support for the H1 of this
experiment: language users experience similarities between the causal verb doen and the causal
connective daardoor.
The natural text items marked with laten reveal a highly similar pattern, see Table 2. Of the
seven items tested, five show a tendency in the expected direction. Four of these items show a
(strong) effect. The significance of the proportion of items supporting the hypothesis was again
estimated with a binomial test. Again, the probability that the present findings for laten must be
attributed to chance appeared to very small ( p = .0002). This finding can be interpreted as an
indication that the H1 of this study holds for the laten-items as well: language users experience
similarities between the causal verb laten and the causal connective daarom.
4.7.2. Chameleon items
The ‘chameleon items’ reveal a totally different picture. The results of the doen-marked
versions of the items are summarized in Table 3. The results of the laten-marked versions are
summarized in Table 4. None of the expectations formulated in section 4.1 are borne out by these
data. Seven of the doen-marked chameleon items show a trend in the expected direction. But of
only two of them, the effect is statistically significant. Three items score in the opposite direction.
As a consequence, the ‘statistical significance’ of these items can be ignored. According to an
N. Stukker et al. / Journal of Pragmatics 40 (2008) 1296–13221314
Table 2
Results hypothesis ‘laten parallels daarom’: natural text items
Item code Degree of correspondence t (d.f.) p
1 .58 .895 (30) .189
2 .11 �6.310 (26) .000
3 .96 12.500 (26) .000*
4 .50 .000 (25) .5
5 .71 2.530 (30) .009*
6 .85 5.505 (26) .000*
7 .74 3.028 (30) .003*
Significant results are indicated with *.
Table 3
Results hypothesis ‘doen parallels daardoor’: chameleon items
Item code Degree of correspondence t (d.f.) p
D1 .61 1.270 (30) .107
D2 .52 .189 (26) .426
D3 .55 .533 (30) .299
D4 .52 .189 (26) .426
D5 .77 3.195 (25) .002*
D6 .30 �2.275 (26) .016
D7 .65 1.617 (25) .059
D8 .88 6.019 (25) <.001*
D9 .45 �5.33 (30) .299
D10 .15 �5.050 (26) <.001
Significant results are indicated with *.
Author's personal copy
estimation based on a binomial test, the probability that the present results must be attributed to
chance is much bigger than with the ‘natural text’-items: p = .07. On the basis of these results, it
cannot be concluded that language users experience similarities between doen and daardoor as
causality markers.
A similar picture emerges from the laten-versions. Here, only three out of ten items show a
trend in the expected direction. Of these items, two items’ scores are statistically significant. Not
surprisingly, a binomial test of these findings produces the same result as with the doen-items,
namely that the probability of finding these proportions when H0 is true, is .07. Therefore, for the
laten-versions H0 cannot be rejected either. Hence, the data for the chameleon items do not
corroborate the idea of a cross-level similarity intuitively shared by language users.
5. Discussion and conclusion
This paper started from the idea that choosing one causality marker over another to express a
causal relation can be seen as an act of linguistic categorization. A basic assumption in cognitive
semantic theory, the framework that lies at the basis of the issues investigated here, is that
linguistic categories reflect conceptual categories in human cognition. Following this line of
reasoning, we have argued that such a categorization approach results in the hypothesis of cross-
level similarities in language: If a parallel between semantic categories and conceptual categories
exists, similar conceptual models of causality are expected to be manifest across different types
of linguistic constructions expressing causality. We have focused on parallels between
constructions that belong to different levels of the linguistic structure: causal verbs (clausal level)
and causal connectives (discourse level) in Dutch. We tested the cross-level similarity hypothesis
that Dutch language users experience parallels between the meaning and use of causal verbs and
causal connectives. In section 3, we have first shown parallels in the meaning and use of causal
verbs and causal connectives, as it was reflected in a corpus of newspaper texts. Subsequently, in
section 4, we reported an experiment that aimed to test the cognitive plausibility of the cross-level
similarity hypothesis. This section starts with an overview of the experimental findings. Next, we
will discuss the methodological aspect of the study reported. Finally, we will discuss how this
study contributes to our ultimate goal, which is to construct a descriptively adequate and
cognitively plausible integrative theory on the meaning and use of causality markers at different
levels of the linguistic structure.
N. Stukker et al. / Journal of Pragmatics 40 (2008) 1296–1322 1315
Table 4
Results hypothesis ‘laten parallels daarom’: chameleon items
Item code Degree of correspondence t (d.f.) p
L1 .52 .189 (26) .426
L2 .39 �1.270 (30) .107
L3 .35 �1.617 (25) .059
L4 .50 .000 (25) .500
L5 .44 �.570 (26) .287
L6 .69 2.083 (25) .024*
L7 .39 �1.270 (30) .107
L8 .23 �3.592 (30) <.001
L9 .37 �1.369 (26) .092
L10 .90 7.470 (30) <.001*
Significant results are indicated with *.
Author's personal copy
5.1. Overview of experimental results
Our cross-level similarity hypothesis was tested with different types of experimental items.
These types turned out to ‘behave’ differently. The ‘natural text items’, based on ‘real life’
occurrences of doen and laten, suggest that, even though not all individual items supported the
hypothesis, the overall conclusion is that the doen-marked items as well as the laten-marked
items clearly corroborate the cross-level similarity hypothesis. The other type of items tested
were the ‘chameleon items’. These were ambiguous not only with respect to marking with
daarom or daardoor, but also with respect to marking with laten or doen. Surprisingly, the results
for these ‘chameleons’ revealed a totally different picture. Only a small amount of the chameleon
items showed statistically significant effects in the expected direction. Statistical evaluation
however indicates that these findings are most probably not caused by an effect of H1.
How can we account for the fact that the chameleons differ from the natural text items so
strongly? The chameleon items were, out of necessity, often entirely constructed texts. Natural
examples of this type do exist, but they are rare. Even making them up was a difficult job. This
suggests that a ‘lower limit’ holds to the degree of ‘underspecification of context’ a causality marker
can bear in order to still properly function as a categorization device (cf. discussion in section 3.1).
The invented chameleon items that we used all show contexts that have become rather ‘faded’ in
terms of the features that determine the categorization of causality in normal contexts. A possible
explanation for the different responses may therefore be that in these contexts, unlike in the contexts
of natural texts, subjects just didn’t have enough ground for identifying causality type in the causal
sentence. We assume that as a consequence, information for making an apt comparison between an
intra-clausal and inter-clausal version of the same causal relation was lacking. An explanation along
these lines is of course speculative, but it is in accordance with response patterns found in the pilot
experiment (see section 4.4). These patterns suggest that the degree of abstractness experienced by
the participants influences the quality of their performance on this task. Participants in the pilot
experiment either seemed not to know at all what to choose (result: no effect found), or they seemed
to base their choice solely on input from the paraphrase contexts, in other words: forgot about
‘comparing the two items’ and just picked out the ‘nicest’ inter-clausal version of the causal relation.
Assuming then, that the natural text items are better suited for testing parallels between causal
verbs and causal connectives, the assessment of the overall significance relies on these items. It is
evident that the results of the experiment form a pattern that is more stable than the one found in
the pilot version. There are still items that behave differently, but no systematic patterning is
found. Our conclusion is that overall the instrument measured the intended construct, but that not
all items measured it to the same extent. At the same time, it has become clear that there are limits
to the kind of issues that can be tested with experimental tasks. This issue is taken up in section 5.2.
Despite these complexities in the experimental task, the results can be interpreted meaningfully as a
first empirical indication that the meanings of causal verbs and causal connectives show parallels,
according to the intuitions of language users.
5.2. Converging evidence
The results of the experiment constitute a direct type of evidence, in addition to the results of
corpus analyses, supporting the claim that causal verbs and causal connectives in Dutch are related
to more general conceptual models of causality. A clear advantage of experimental testing was the
possibility of manipulating text material so as to put the comparability of causal verbs and causal
connectives on edge. Moreover, the results of this experiment add cognitive plausibility to the
N. Stukker et al. / Journal of Pragmatics 40 (2008) 1296–13221316
Author's personal copy
integrative perspective on causality markers advocated here. We have argued repeatedly in favor of
the methodology of converging evidence, including the experimental study of language use. In our
view, linguists working in the field of language and cognition should remain open to
complementary research strategies like corpus studies and experiments, because these provide
a deeper insight into language use. Fortunately, a growing number of researchers combine methods
of theory-building and testing (see Gibbs, 2006 and contributions to Gonzalez-Marquez et al.,
2007). Focusing on issues of categorization here, it can be concluded that linguistic intuitions are
more and more checked with those of language users. Examples include Sandra and Rice (1995) on
the polysemy of prepositions, Sanders and Spooren (1996) on epistemic modality and Sanders et al.
(1992, 1993) and Sanders (1997) on the categorization of coherence relations. The methods used
vary from experimentation with conscious to experimentation with unconscious behavior. For the
purposes of the present study, it was essential to make use of conscious behavior, in which subjects
are asked to give judgments or to react to a meta-linguistic task.
The challenge was, of course, in the question whether language users could use their intuitions
to compare two types of items. Earlier studies had used less complex tasks – card sorting, filling
in connectives – which did not involve ‘translating’ one type of causality in terms of another. It is
fair to conclude that the task indeed seemed to be too difficult for the relatively naive students that
participated in the pilot-experiment, being first year university students and second year students
in higher vocational education (see section 4.4). The fact that results for the second version of the
experiment were more unequivocal can be interpreted as an indication that the participants in this
experiment were better up to the task. This implies that our comparison task requires participants
to have some experience with ‘judging language’; they should be experienced in reflecting on
language data, but they should of course also be ‘naive’ regarding the specific linguistic theory
under investigation. Evidently, this type of participants is hard to find. Moreover, there is little
experience with highly complex experimental tasks of the type used in our experiment.
Notwithstanding these difficulties, our results present a clear indication of the fruitfulness of
the research strategy of converging evidence: they show how results of corpus analyses and
experimental methods complement each other. The corpus analyses and experimental results
added to the descriptive adequacy and the cognitive plausibility of our theoretical approach.
We intend to test the robustness of the patterns found in future research, by extending the range of
causal expressions to be tested (such as prepositions and ‘backward’ causal connectives), and by
testing the cross-level similarity hypothesis in converse direction, asking participants to
paraphrase interclausal causal relation marked with a connective with an intraclausal causal
relation marked with a causal verb.16
5.3. Conclusion: towards an integrative perspective on causality markers
This study yields arguments in favor of the cross-level similarity hypothesis we presented in
this paper. We found strong indications that Dutch language users rely on the same conceptual
models of causality when they express causality in discourse with connectives and in clauses with
auxiliary verbs. The contrast between ‘animate’ and ‘inanimate participants’ in the causal
N. Stukker et al. / Journal of Pragmatics 40 (2008) 1296–1322 1317
16 Results of a pilot experiment, however, suggest that participants find a task in which causal information has to be
‘reduced’ (from text into one clause) even more complex than the experimental task presently used, where causal
information had to be ‘extended’ (from one clause into a more elaborate text, see Van Maaren, 2002). Similarly, it is to be
expected that the paraphrasing task will be more complex when the presentation order of cause and effect segments of
‘original’ and ‘paraphrase’ diverges, which will be the case when including backward causal connectives. Therefore,
perhaps the most important challenge lies in constructing a better match between the experimental task and participants.
Author's personal copy
process seems to play a crucial role in categorizations made both at the clause-level with causal
verbs, and at the discourse level with causal connectives. However, the experiment focused on
only two of the marking options for causal coherence relations in Dutch: daardoor and daarom,
and did not take into account dus ‘so’—a connective which is used at least as frequently for
marking forward causal relations in Dutch as daardoor and daarom are. Dus is prototypically
related to the conceptual model of subjective causality (cf. discussion in section 3.2). In section 3.3
we suggested that this type of causality cannot be expressed within an analytical causative
construction. Some examples will illustrate this point:
(19) De straten zijn nat, dus het regent.
The streets are wet, so it is raining.
(20) Het regent, dus de straten worden nat.
It is raining, so the streets are getting wet.
(21) *De natte straten laten/doen het regenen.
*The wet streets are making it rain.
(22) De regen doet/?laat de straten nat worden.
The rain is making the streets get wet.
Neither of the inter-clausal epistemic relations, relating an argument in the first clause to the
conclusion presented in the second clause, can be reformulated as an analytic causative
construction. This either leads to ungrammaticality (21) or to an ‘objective causality’ inter-
pretation, in which case the causal relation holds in ‘the real world’ (22). This difference in
interpretation, resulting from simple ‘transposing’ the relation from one construction to the other,
must be accounted for as well. A possible line of reasoning is located in the assumption,
elaborated in the field of cognitive linguistics, that ‘constructional’ (syntactic) aspects contribute
to the overall semantics and interpretation of a linguistic utterance, in other words: that
grammatical differences may entail conceptual differences (cf. Langacker, 1987; Goldberg,
1995; Thompson, 1985; Ford et al., 2002; see for evidence from Dutch causal connectives Evers-
Vermeul, 2005). Applied to the apparent lacking of a marker of subjective causality in analytic
causative constructions, this is our explanation: Subjective causal relations always consist of two
separate propositions, which are related by a communicative act of the speaker. For example, in
(19) and (20) the second sentences contain conclusions based on the arguments presented in the
first sentences (cf. section 3.2). It seems logical that this type of relations can only be expressed in
grammatical constructions that consist of (at least) two separate propositions, which is the
case with inter-clausal causal constructions, and not with the intra-clausal analytic causative
constructions (see Degand, 2000, 2001 for a similar stance regarding causal prepositions in Dutch).
The findings and suggestions presented in this paper can be summarized as in Fig. 4.
The findings reported in this paper add a new type of evidence in favor of the long-standing
hypothesis that patterns in the linguistic expression of causal relations are governed by human
conceptual structure. We argued that an integrative approach to causality markers of grammatical
different types refines our understanding of the mechanisms underlying the linguistic expression
of causal relations. Apart from the results corroborating our cross-level similarity hypothesis, we
found indications that cross-level parallels are mediated by constructional aspects. In addition,
our findings suggest that discourse-level expressions are subject to the same principles and
N. Stukker et al. / Journal of Pragmatics 40 (2008) 1296–13221318
Author's personal copy
mechanisms as the clause-level expressions regarding the linguistic communication of causality.
A final point we want to underline is the fruitfulness of the research strategy of converging
empirical evidence. It is only by selecting and combining different empirical methods that
generalizing patterns of the type investigated in the present study could be brought to light.
Acknowledgements
We would like to thank Huub van den Bergh for his invaluable advice on statistic analysis, and
two anonymous reviewers for their comments on an earlier version of this paper. All remaining
errors are our own. The first two authors acknowledge the support of NWO-grant 277-70-003,
awarded to Ted Sanders, while preparing this paper.
References
Anscombre, Jean-Claude, Ducrot, Oswald, 1983 [1995]. L’argumentation dans la langue. Mardaga, Bruxelles.
Barlow, Michael, Kemmer, Suzanne (Eds.), 2000. Usage Based Models of Language. CSLI Publications, Stanford.
Bybee, Joan, 1985. Morphology: A Study of the Relation Between Meaning and Form. John Benjamins, Amsterdam.
Bybee, Joan, 2006. From usage to grammar: the mind’s response to repetition. Language 82, 711–733.
Dancygier, Barbara, Sweetser, Eve, 2000. Constructions with if, since, and because: causality, epistemic stance, and
clause order. In: Couper-Kuhlen, E., Kortmann, B. (Eds.), Cause, Concession, Contrast: Cognitive and Discourse
Perspectives. Mouton de Gruyter, Berlin, pp. 111–142.
D’Andrade, Roy, 1987. A folk model of the mind. In: Holland, D., Quinn, N. (Eds.), Cultural Models in Language and
Thought. Cambridge University Press, Cambridge, pp. 112–148.
Degand, Liesbeth, 1996. Causation in Dutch and French: interpersonal aspects. In: Hasan, R., Cloran, C., Butt, D.G.
(Eds.), Functional Descriptions. John Benjamins, Amsterdam, pp. 207–237.
Degand, Liesbeth, 2000. Causal connectives or causal prepositions? Discursive constraints. Journal of Pragmatics 32,
687–707.
Degand, Liesbeth, 2001. Form and Function of Causation. A Theoretical and Empirical Investigation of Causal
Constructions in Dutch. Peeters, Leuven.
Dik, Simon C., 1978. Functional Grammar. Foris, Dordrecht.
Evers-Vermeul, Jacqueline, 2005. The development of Dutch connectives. Change and acquisition as windows on form-
function relations. Ph.D. dissertation. LOT, Utrecht University (Accessible online via: http://igitur-archive.librar-
y.uu.nl/dissertations/2005-0817-200059/index.htm).
Fauconnier, Gilles, 1994. Mental Spaces: Aspects of Meaning Construction in Natural Language. MIT Press, Cambridge.
Fillmore, Charles, 1976. Topics in lexical semantics. In: Cole, P. (Ed.), Current Issues in Linguistic Theory. Indiana
University Press, Bloomington, pp. 76–138.
N. Stukker et al. / Journal of Pragmatics 40 (2008) 1296–1322 1319
Fig. 4. Summary of conceptual parallels between verbs and connectives.
Author's personal copy
Ford, Cecilia E., Fox, Barbara A., Thompson, Sandra A., 2002. Constituency and the grammar of turn increments. In:
Ford, C.E., Fox, B.A., Thompson, S.A. (Eds.), The Language of Turn and Sequence. Oxford University Press,
Oxford, pp. 14–38.
Geeraerts, Dirk, 1997. Diachronic Prototype Semantics. A Contribution to Historical Lexicology. Clarendon Press, Oxford.
Gibbs Jr., Raymond, 2006. Introspection and cognitive linguistics: should we trust our own intuitions? In: Ruiz de
Mendoza Ibanez, F.J. (Ed.),Annual Review of Cognitive Linguistics, 4. John Benjamins, Amsterdam, pp. 135–151.
Givon, Talmy, 1975. Cause and control: on the semantics of interpersonal manipulation. In: Kimball, J.P. (Ed.), Syntax
and Semantics 4. Academic Press, New York, pp. 59–89.
Givon, Talmy, 1976. Some constraints on Bantu causativization. In: Shibatani, M. (Ed.), Syntax and Semantics. The
Grammar of Causative Constructions. Academic Press, New York, pp. 325–351.
Goldberg, Adele E., 1995. Constructions: A Construction Grammar Approach to Argument Structure. University of
Chicago Press, Chicago.
Gonzalez-Marquez, Monica, Mittelberg, Irene, Coulson, Seana, Spivey, Michael J. (Eds.), 2007. Methods in Cognitive
Linguistics. John Benjamins, Amsterdam.
Gries, Stefan, Hampe, Beate, Schonefeld, Doris, 2005. Converging evidence: bringing together experimental and corpus
data on the association of verbs and constructions. Cognitive Linguistics 16, 635–676.
Haiman, John, 1980. The iconicity of grammar: isomorphism and motivation. Language 56, 515–540.
Halliday, M.A.K., 1985. An Introduction to Functional Grammar. Arnold, London.
Hobbs, Jerry R., 1979. Coherence and coreference. Cognitive Science 3, 67–90.
Kemmer, Suzanne, Verhagen, Arie, 1994. The grammar of causatives and the conceptual structure of events. Cognitive
Linguistics 5, 115–156.
Lakoff, George, 1987. Women, Fire and Dangerous Things. What Categories Reveal about the Mind. University of
Chicago Press, Chicago.
Langacker, Ronald W., 1987. Foundations of Cognitive Grammar. Vol. I, Theoretical Prerequisites. Stanford University
Press, Stanford.
Langacker, Ronald W., 2000. A dynamic usage-based model. In: Barlow, M., Kemmer, S. (Eds.), Usage Based Models of
Language. CSLI Publications, Stanford, pp. 1–63.
Langacker, Ronald W., 2002 [1990]. Concept, Image, and Symbol. The Cognitive Basis of Grammar. Berlin, etc., Mouton
de Gruyter.
Lyons, John, 1995. Linguistic Semantics. An Introduction. Cambridge University Press, Cambridge.
Mann, William C., Thompson, Sandra A., 1988. Rhetorical structure theory: towards a functional theory of text
organization. Text 8, 243–281.
Michotte, A., 1963. The Perception of Causality. Basic Books, New York.
Pander Maat, Henk, Degand, Liesbeth, 2001. Scaling causal relations and connectives in terms of speaker involvement.
Cognitive Linguistics 12, 211–245.
Pander Maat, Henk, Sanders, Ted, 1995. Nederlandse causale connectieven en het onderscheid tussen inhoudelijke en
epistemische relaties (‘Dutch causal connectives and the distinction between content and epistemic relations’).
Leuvense Bijdragen 3, 349–374.
Pander Maat, Henk, Sanders, Ted, 2000. Domains of use or subjectivity: the distribution of three Dutch causal connectives
explained. In: Couper-Kuhlen, E., Kortmann, B. (Eds.), Cause, Condition, Concession, and Contrast: Cognitive and
Discourse Perspectives. Mouton de Gruyter, Berlin, pp. 57–82.
Pander Maat, Henk, Sanders, Ted, 2001. Subjectivity in causal connectives: an empirical study of language in use.
Cognitive Linguistics 12, 247–273.
Piaget, Jean, Garcia, Rolando, 1974. Understanding Causality. W.W. Norton & Co., New York.
Piaget, Jean, Inhelder, Barbel, 1969. The Psychology of the Child. Basic Books, New York.
Pit, Mirna, 2003. How to express yourself with a causal connective? Subjectivity and causal connectives in Dutch, German
and French. Dissertation Utrecht University. Rodopi, Amsterdam.
Pit, Mirna, 2005. Subjectiviteit en causale connectieven (‘‘Subjectivity and causal connectives’’). Tijdschrift voor
Taalbeheersing 27, 24–41.
Rumelhart, David, 1975. Notes on a schema for stories. In: Bobrow, D.G., Collins, A.M. (Eds.), Representation and
Understanding: Studies in Cognitive Science. Academic Press, New York, pp. 211–236.
Sanders, Jose, Spooren, Wilbert, 1996. Subjectivity and certainty in epistemic modality. A study of Dutch epistemic
modifiers. Cognitive Linguistics 7, 241–264.
Sanders, Jose, Spooren, Wilbert, 1997. Perspective, subjectivity, and modality from a cognitive linguistic point of view.
In: Liebert, W., Redeker, G., Waugh, L. (Eds.), Discourse and Perspective in Cognitive Linguistics. John Benjamins,
Amsterdam, pp. 85–112.
N. Stukker et al. / Journal of Pragmatics 40 (2008) 1296–13221320
Author's personal copy
Sanders, Ted, 1997. Semantic and pragmatic sources of coherence: on the categorization of coherence relations in context.
Discourse Processes 24, 119–147.
Sanders, Ted, Spooren, Wilbert, 2001. Text representation as an interface between language and its users. In: Sanders,
T., Schilperoord, J., Spooren, W. (Eds.), Text Representation: Linguistic and Psycholinguistic Aspects. John
Benjamins, Amsterdam, pp. 3–25.
Sanders, Ted, Spooren, Wilbert, 2007. Discourse and text structure. In: Geeraerts, D., Cuyckens, H. (Eds.), Handbook of
Cognitive Linguistics. Oxford University Press, Oxford.
Sanders, Ted J.M., Spooren, Wilbert P.M., Noordman, Leo G.M., 1992. Toward a taxonomy of coherence relations.
Discourse Processes 15, 1–35.
Sanders, Ted J.M., Spooren, Wilbert P.M., Noordman, Leo G.M., 1993. Coherence relations in a cognitive theory of
discourse representation. Cognitive Linguistics 4, 93–133.
Sanders, Ted, Verhagen, Arie, 1996.In: De cross-linguıstische realisering van causaliteit in werkwoorden en connectieven
(‘The cross-linguistic realization of causality in verbs and connectives’). Paper Presented at the TIN-dag, Utrecht
University, 20 January 1996.
Sandra, Dominiek, Rice, Sally, 1995. Network analysis of prepositional meaning: mirroring whose mind—the linguist’s
or the language user’s? Cognitive Linguistics 6, 89–130.
Schank, Roger, Abelson, Robert, 1977. Scripts, Plans, Goals, and Understanding. Lawrence Erlbaum Associates,
Hillsdale, NJ.
Shibatani, Masayoshi, 1976. The grammar of causative constructions: a conspectus. In: Shibatani, M. (Ed.), Syntax and
Semantics. The Grammar of Causative Constructions. Academic Press, New York, pp. 1–40.
Shibatani, Masayoshi, Pardeshi, Prashant, 2001. The causative continuum. In: Shibatani, M. (Ed.), The grammar of
causation and interpersonal manipulation. John Benjamins, Amsterdam, pp. 85–126.
Stein, Dieter, Wright, Susan (Eds.), 1995. Subjectivity and Subjectivisation. Linguistic Perspectives. Cambridge
University Press, Cambridge.
Stukker, Ninke, 2005. Causality marking across levels of language structure. A cognitive semantic analysis of causal verbs
and causal connectives in Dutch. Dissertation. LOT, Utrecht University, Utrecht (Accessible online via: http://igitur-
archive.library.uu.nl/dissertations/2006-0428-200107/UUindex.html).
Stukker, Ninke, Sanders, Ted, Verhagen, Arie, in preparation. Categories of subjectivity in Dutch causal connectives: A
usage-based analysis. Utrecht University, Leiden University.
Sweetser, Eve E., 1990. From Etymology to Pragmatics. Metaphorical and Cultural Aspects of Semantic Structure.
Cambridge University Press, Cambridge.
Talmy, Leonard, 1976. Semantic causative types. In: Shibatani, M. (Ed.), Syntax and Semantics. The Grammar of
Causative Constructions. Academic Press, New York, pp. 43–116.
Talmy, Leonard, 1988a. Force dynamics in language and cognition. Cognitive Science 12, 49–100.
Talmy, Leonard, 1988b. The relation of grammar to cognition. In: Rudzka-Ostyn, B. (Ed.), Topics in Cognitive
Linguistics. John Benjamins, Amsterdam, pp. 165–205.
Talmy, Leonard, 2000. Toward A Cognitive Semantics, vol. I. MIT Press, Cambridge.
Taylor, John R., 1995. Linguistic Categorization. Prototypes in Linguistic Theory. Clarendon Press, Oxford.
Thompson, Sandra A., 1985. Grammar and written discourse: initial and final purpose clauses in English. Text 5, 55–
84.
Van Maaren, Arvid, 2002. Willens en wetens. Onderzoek naar de conceptuele overeenkomsten tussen causaliteitsmar-
keringen. [‘‘Willfully and knowingly’’. A study on conceptual parallels in causality markers’]. Unpublished Master’s
thesis, Utrecht University.
Verhagen, Arie, 1995. Subjectification, syntax, and communication. In: Stein, D., Wright, S. (Eds.), Subjectivity and
Subjectivisation. Linguistic Perspectives. Cambridge University Press, Cambridge, pp. 103–128.
Verhagen, Arie, 1997. Context, meaning, and interpretation, in a practical approach to linguistics. In: Lentz, L., Pander
Maat, H.L.W. (Eds.), Discourse analysis and discourse evaluation. Rodopi, Amsterdam, pp. 7–39.
Verhagen, Arie, 2000. Interpreting usage: construing the history of Dutch causal verbs. In: Barlow, M., Kemmer, S.
(Eds.), Usage Based Models of Language. CSLI Publications, Stanford, pp. 261–286.
Verhagen, Arie, 2005. Constructions of Intersubjectivity. Discourse, Syntax, and Cognition. Oxford University Press,
Oxford.
Verhagen, Arie, Kemmer, Suzanne, 1992. Interactie en oorzakelijkheid: een cognitieve benadering van causatief-
constructies in het Nederlands. Gramma/TTT 1, 1–20.
Verhagen, Arie, Kemmer, Suzanne, 1997. Interaction and causation: causative constructions in modern standard Dutch.
Journal of Pragmatics 27, 61–82.
Wierzbicka, Anna, 1988. The Semantics of Grammar. John Benjamins, Amsterdam.
N. Stukker et al. / Journal of Pragmatics 40 (2008) 1296–1322 1321
Author's personal copy
Wolff, Phillip, Song, Grace, 2003. Models of causation in the semantics of causal verbs. Cognitive Psychology 47, 276–
332.
Ninke Stukker is post doctoral research fellow in the project Causality and Subjectivity in Discourse and Cognition,
Utrecht University. In 2005 she defended her dissertation Causality marking across levels of language structure. A
cognitive semantic analysis of causal verbs and causal connectives in Dutch (Utrecht: LOT).
Ted Sanders is professor of Discourse Studies and Dutch language use at Utrecht University, and heads the research
project Causality and Subjectivity in Discourse and Cognition, funded by the Dutch organization for scientific research
(NWO). He co-edited special issues of Cognitive Linguistics and Discourse Processes, and a book volume on Text
Representation: Linguistic and Psycholinguistic Approaches (Benjamins, 2001).
Arie Verhagen has been the chair of Dutch Linguistics at the University of Leiden since 1998. He heads the research
project Stylistics of Dutch, funded by the Dutch organization for scientific research (NWO). He is the author of Linguistic
Theory and the Function of Word Order in Dutch (Foris/Mouton de Gruyter, 1986) and Constructions of Intersubjectivity
(Oxford University Press, 2005), and co-edited Usage-Based Approaches to Dutch (LOT, 2003).
N. Stukker et al. / Journal of Pragmatics 40 (2008) 1296–13221322