Computer-Supported Collaborative Learning in
Higher Education: Scripts for Argumentative
Knowledge Construction in Distributed Groups
Knowledge Media Research Center,
Department for Applied Cognitive
Psychology and Media Psychology,
University of Tübingen
Abstract. Learners rarely know how to construct knowledge together in argumentation. This
experimental study analyzes two computer-supported collaboration scripts, which should facilitate
processes and outcomes of argumentative knowledge construction. One script aims to support the
construction of single arguments and the other script aims to support the construction of
argumentation sequences. Both scripts were varied independently in a 2×2-factorial design. 120
students of Educational Science participated in the study in groups of three. Results show that the
computer-supported scripts facilitate specific processes and outcomes of argumentative
knowledge construction. Learners with scripts argued better and acquired more knowledge on
argumentation than learners without scripts.
Keywords: argumentative knowledge construction, computer-supported collaboration scripts
University students are supposed to become experts within a specific domain. In this regard, students are meant
to be able to both understand and participate in argumentative discourse in their field. Even though knowledge
on argumentation start to develop from an early age (Stein & Bernas, 1999), studies showed that adults’
knowledge on argumentation are often suboptimal (e.g., Kuhn, 1991). Adults hardly base their claims on
grounds and rarely consider counterarguments. Even though students may in general acquire domain-specific
knowledge, they hardly seem to learn how to argue based on this knowledge within their domain.
An important opportunity for the development of knowledge on argumentation is the active participation in
high-quality argumentative discourse in instructional settings (Kuhn, 1991). High-quality argumentative
discourse in instructional settings means that collaborative learners construct formally and content adequate
arguments while jointly working on a learning task. Computer-supported collaborative learning (CSCL) may
provide an ideal context for this kind of discourse (Marttunen & Laurinen, 2001). During CSCL, students may
construct and exchange arguments online that can be examined and evaluated by learning partners for longer
periods of time than in face to face situations. Collaborative learners may thus elaborate the learning material by
constructing arguments themselves to promote their perspective on one hand and on the other integrate
arguments of their learning partners in their own perspective. In this way, learners may lead a high-quality
online argumentative discourse with regard to formal aspects and contents and acquire domain-specific
knowledge as well as knowledge on argumentation (see Andriessen, Baker, & Suthers, 2003; Weinberger &
Fischer, in press).
The goal of this study is to investigate how processes as well as outcomes of argumentative knowledge
construction can be facilitated by means of computer-supported scripts within a CSCL environment.
ARGUMENTATIVE KNOWLEDGE CONSTRUCTION
Argumentative knowledge construction means that learners construct arguments within a specific domain with
the goal to acquire knowledge (Weinberger & Fischer, in press). First, we will portray potential outcomes of
argumentative knowledge construction. Second, we will describe the processes of argumentative knowledge
construction and how they may facilitate specific outcomes.
Argumentative knowledge construction aims to foster at least two different outcomes, namely domain-specific
knowledge as well as knowledge on argumentation (Andriessen et al., 2003).
Knowledge on argumentation comprises knowledge on how to construct formally complete arguments with
the components claim, ground and qualifier (knowledge on the construction of single arguments) and the
knowledge on how to construct specific sequences of arguments consisting of arguments, counterarguments and
integrations (knowledge on the construction of argumentation sequences).
Domain-specific knowledge in the context of this study means to be able to apply concepts from a specific
theory that is to be learned. Learners constructing formally and content adequate arguments activate their prior
knowledge, elaborate the given learning material, and thus construct new domain-specific knowledge
(Andriessen et al., 2003).
The processes of argumentative knowledge construction are allocated on at least two dimensions, namely the
formal argumentative dimension, regarding the formal structure of arguments and argumentation sequences, and
the epistemic dimension, regarding the contents of the single arguments (Weinberger & Fischer, in press).
On the formal argumentative dimension, the construction of single arguments and the construction of
argumentation sequences consisting of more than one single argument can be differentiated.
A single argument has been regarded as a claim which can be supported by grounds and/or specified by
qualifier (Toulmin, 1958). Grounds may justify the claim through a warrant. The qualifier limits the validity of
the claim. Constructing arguments with these elements facilitates self-explanation of the learning material
(Baker, 2003). Self-explanation is supposed to facilitate the integration of new knowledge into existing
cognitive structures. Learners prompted to give self-explanations acquired more knowledge than unsupported
learners (Chi, DeLeeuw, Chiu, & LaVancher, 1994).
Specific argumentation sequences have been regarded as an indicator for the construction of knowledge
(Leitão, 2000). First, learners construct arguments to justify their position. This construction of arguments
facilitates self-explanation of the learning material (see Baker, 2003). Second, learning partners construct
counterarguments to challenge and reconsider these positions. Counterarguments facilitate meta-cognitive
activities, prompting learners to rethink their initial argument (Leitão, 2000). Finally, learners construct replies
and eventually refine the initial positions. By balancing arguments and counterarguments in order to solve
complex problems, participants may acquire knowledge on argumentation and domain-specific knowledge.
The epistemic dimension regards how learners work on the learning task, what (theoretical) concepts they
consider and how they may construct knowledge. Beyond formal aspects of argument construction, the contents
learners use to construct arguments supposedly play a crucial role in argumentative knowledge construction
(Kuhn, Shaw, & Felton, 1997). It has been found that learners in problem-oriented learning environments need
to apply those theoretical concepts, which they are supposed to learn (application of new knowledge) in order to
acquire domain-specific knowledge (Fischer, Bruhn, Gräsel, & Mandl, 2002; Weinberger, 2003). Beyond
applying new knowledge, application of prior knowledge has been regarded as important to the acquisition of
domain-specific knowledge, e.g., in problem-oriented learning environments of medical education (Schmidt,
1993). The amount of activated prior knowledge is supposed to determine how much new knowledge can be
acquired. Students construct meaning by using their prior knowledge in the sense that new knowledge needs to
be meaningfully related to existing bodies of knowledge (Anderson & Pearson, 1984).
USING COMPUTER-SUPPORTED COLLABORATION SCRIPTS TO FACILITATE
ARGUMENTATIVE KNOWLEDGE CONSTRUCTION
A central topic of CSCL research is how argumentative knowledge construction can be facilitated. Different
approaches are being investigated. One prominent approach is visualization, which uses software tools and
different representations to guide argumentative knowledge construction. Interfaces with different
representational aids such as graphs, matrices or texts were found to have different effects on CSCL (Suthers &
Hundhausen, 2001). Software tools, may visualize the argumentation of learners (Kirschner, Buckingham Shum,
& Carr, 2003). For instance, diagrammatic representations visualize how arguments are related to each other and
thus facilitate and guide learners’ awareness of the argumentative discourse (Hoppe, Gaßner, Mühlenbrock &
Tewissen, 2000). Tools like SenseMaker (Bell, 1997) support learners to represent their arguments by providing
spaces and categories to group arguments and differentiate claims from evidence.
Another approach to facilitate argumentative knowledge construction is to realize computer-supported
collaboration scripts based on O’Donnell’s (1999) scripted cooperation approach. The interface suggests
learners to construct specific arguments by providing prompts learners should use or respond to (Baker & Lund,
1997; Dillenbourg, 2002; Kollar, Fischer, & Hesse, 2003; Nussbaum, Hartley, Sinatra, Reynolds, & Bendixen,
2002; Weinberger, 2003; Weinberger, Ertl, Fischer, & Mandl, in press). In this approach, interfaces may be
designed to specify and sequence and eventually to allocate different learning activities to learners. Studies
show, that computer-supported collaboration scripts may support specific processes and outcomes of
argumentative knowledge construction, but may have “side effects” on others (Dillenbourg, 2002; Weinberger
et al., in press). Kollar and colleagues (2003) investigated computer-supported collaboration scripts, which
provide text spaces for claims and evidence that learners are supposed to fill as well as a specific sequence of
arguments, counterarguments and integrations. Whereas learners acquired domain-specific knowledge
independent of the script support in this study, computer-supported collaboration scripts facilitated knowledge
on argumentation as an outcome of argumentative knowledge construction. Against this background, scripts can
be conceptualized that facilitate the construction of a single argument according to Toulmin’s model (1958) and
scripts that facilitate the construction of argumentation sequences according to Leitão (2000). A script for the
construction of single arguments should facilitate the relative frequency of grounds that support a claim while a
script for the construction of argumentation sequences should foster the relative frequency of counterarguments.
Both scripts should support learners to apply concepts from prior knowledge to problems (application of prior
knowledge) as well as the new theoretical concepts they are supposed to learn (application of new knowledge).
There is little knowledge whether computer-supported collaboration scripts that specifically aim to support the
construction of single arguments and argumentation sequences may foster the formal argumentative and / or the
epistemic dimension of argumentative knowledge construction. Based on this, the following two research
questions are examined:
• To what extent does a script for the construction of single arguments and a script for the construction of
argumentation sequences and their combination, influence the processes of argumentative knowledge
construction on the formal argumentative and the epistemic dimension?
• To what extent does a script for the construction of single arguments and a script for the construction of
argumentation sequences and their combination, facilitate the outcomes of argumentative knowledge
construction, namely domain-specific knowledge and knowledge on argumentation?
Sample and Design
One hundred twenty students of educational psychology participated in this study. The experimental learning
environment was part of a regular curriculum. The students, who were attending a mandatory introduction
course, participated in an online learning session as a substitute for one regular face to face session of the
course. Participation was required in order to receive a course credit at the end of the semester. The learning
outcomes of the experimental session, however, were not accounted for in students’ overall performance. The
participants were separated into groups of three and each group was randomly assigned to one of the four
experimental conditions in a 2×2 factorial design. We varied (1) the script for the construction of single
arguments (without vs. with) and (2) the script for the construction of argumentation sequences (without vs.
with). Time on task was held constant in all four conditions.
Learning environment in the different experimental conditions
The subject matter of the learning environment was Weiner’s attribution theory (1985). A three-page description
of this theory was handed out to the students. Three learning cases were used as a central component of the
learning environment. Each case was authentic and complex and allowed learners to construct different
arguments based on theoretical concepts of the attribution theory. One case, for instance, asked to interpret
school performance differences between Asian and American/European students with the attribution theory.
Three students worked separately in one of three different laboratory rooms. The learners’ task was to
analyze together the three cases in an 80 minute collaboration phase and to provide a joint solution of the case.
A problem-oriented learning environment, developed for asynchronous, text-based collaboration was used. The
implemented newsgroup tool was used to exchange email-like text messages. In addition, the environment
allowed for implementing different types of computer-supported collaboration scripts.
(1) The control group received no additional support in solving the three problem-cases.
(2) The script for the construction of single arguments is implemented in the interface as a given text
structure within the individual messages and aims to support learners in the formation of single arguments. The
script, based on Toulmin’s model (1958), differentiates between claim, ground with warrant and qualifier and is
realized by text windows in the interface of the CSCL environment (see figure 1). The learners were asked to fill
out each text window of the interface to construct a complete single argument. After building the argument, the
single argument would be added with a click to the message body. Non-argumentative parts of the message, like
questions, could be added directly to the message body, without using the argument construction interface.
Figure 1. The interface of the script for the construction of single arguments.
(3) The script for the construction of argumentation sequences aimed to facilitate a specific argumentation
sequence of argument-counterargument-integration (following Leitão, 2000). The subject of the posted message
was automatically pre-set, depending on the position in the cascading discussion thread. Each first message of a
discussion thread was labelled “argument”. The answer to an argument was automatically labelled as
counterargument and a reply to a counterargument was labelled as integration. The next message was again
labelled counterargument, then integration and so on. In this way, there was a default path through the
discussion according to the Leitão model (see figure 2). The learners could change the subject of their message
(4) In the combined condition, the learners are supported with both scripts during collaboration. The
interface contains the three fields for argument construction and subjects of the messages are pre-set
automatically by the script for the construction of argumentation sequences.
Figure 2. Discussion thread guided by the script for the construction of argumentation sequences.
First, pre-tests served to determine prior domain-specific knowledge, knowledge on argumentation and
experience with CSCL environments. The pre-tests were used to control randomization. Subsequently, the
participants could individually study the three-page description of the attribution theory for 20 minutes. Learners
were then introduced to the learning environment. Afterwards, the learners collaborated for 80 minutes in
groups of three to work on the learning cases and to agree on case analyses. In the final phase (about 45
minutes), the students took individual post-tests on domain-specific application-oriented knowledge regarding
the attribution theory and knowledge on argumentation tests.
Data sources and instruments
Processes and outcomes of argumentative knowledge construction have been analyzed with an instrument
described in Weinberger and Fischer (in press). Trained coders segmented the discourse corpora into
propositions and rated the segments on the epistemic dimension with regard to application of prior knowledge
and application of new knowledge and on the formal process dimension of argumentative knowledge
construction with regard to the construction of single arguments and the construction of argumentation
sequences. With respect to segmentation, the coders achieved an agreement of 83%. The median of the Kappa
values for categorizing the epistemic dimension was sufficiently high with κ = .72 as well as for the formal
argumentative dimensionl (κ = .61).
On the formal argumentative process dimension of argumentative knowledge construction, grounds as well as
counterarguments have been coded. Grounds are reasons given in support of a claim. Grounds can come in form
of facts, statistics, expert opinions, examples, explanations and logical reasoning. In the context of this study,
learners may support claims with case information or concepts from the given attribution theory. Indicators for
grounds that support claims are prepositions such as “because”, “due to the fact” etc. even though learners may
not always explicitly connect grounds to the respective claims. For instance, if the claim, “Asian attribution
patterns are typically superior” is based on the ground “Asians typically ascribe failure to a lack of efforts rather
than a lack of talent”, this last phrase has been coded as one ground. The percentage of grounds has been
calculated in comparison to other components of single arguments (simple claims, qualifiers, and non-
argumentative moves such as questions). A high share of grounds indicates high-quality argumentative
discourse with respect to the construction of single arguments.
Regarding the construction of argumentation sequences, the percentage of counterarguments was calculated
in comparison to other argumentative moves within an argumentation sequence (arguments, integrations, and
non-argumentative moves). Counterarguments are arguments that oppose another argument. The opposition of
arguments has been assessed on the basis of differences of claims. If one claim contradicts a preceding claim,
the later claim is being coded as counterargument. For instance, the argument “The teacher is supporting his
pupils in adjusting the task difficulty to their individual skill levels” can be countered by the argument “The
teacher is not supporting the pupils in adjusting the task difficulty (because adjusting task difficulty can be based
on a dysfunctional attribution of the teacher)”. Counterarguments are typically expressed by another learner than
the one who made the initial claim. Learners may, however, also construct counterarguments to their own
On the epistemic dimension, both the application of new knowledge and the application of prior knowledge
have been focused on. With regard to the application of new knowledge, any unit of analysis has been coded that
contains a relation of a theoretical concept from the given attribution theory to case information. For instance,
the case information “Michael says he is not talented for maths” is explained with the following theoretical
concept in the phrase “this indicates that Michael attributes his failure in maths to internal stable causes”. When
learners explain case information with concepts that do not stem from the given attribution theory, they apply
prior knowledge to case information, e.g., the case information “Michael says he is not talented for maths” is
considered in “Michael is just plain lazy – he needs to acquire learning strategies and discipline”.
The processes on the formal argumentative and the epistemic dimension will be illustrated in a single case
study based on a segment of a discussion thread that has been supported with the script for the construction of
argumentation sequences. The segment will indicate the single messages, their titles, authors (with fictional
names), and their position in the cascading discussion thread. Each message will be analyzed for the above
process categories on the two dimensions, namely with regard to grounds, counterarguments, application of new
knowledge and application of prior knowledge.
In order to measure domain-specific knowledge, participants had to individually analyze a new case. The written
analyses of the participants were segmented into propositions and coded with respect to adequate applications of
theoretical concepts of the attribution theory. The number of these propositions that the individual learners were
able to construct was counted by five trained coders (κ = .72) and served as indicator for the acquisition of
In the knowledge on argumentation test the participants had to recall components of single arguments and
argumentation sequences. Furthermore, participants were asked to formulate arguments about “smoking” in the
knowledge on argumentation test. The arguments that learners built were analyzed with respect to the
components of single arguments (claim, ground, and qualifier). The argumentation sequences that learners built
were analyzed with respect their function as argument, counterargument, and integration. Thus, knowledge on
the construction of single arguments and knowledge on the construction of argumentation sequences were
differentiated. Two trained coders rated the knowledge on argumentation test (κ = .83).
Research Question 1 on processes of argumentative knowledge construction
First of all, the effects of the two computer-supported collaboration scripts and their combination on the
processes of argumentative knowledge construction were examined. This includes the effects of the two scripts
on the formal argumentative dimension and the effects of the scripts on the epistemic dimension.
With respect to the formal argumentative dimension the scripts produced the following specific effects on
the relative frequency of grounds (see table 1 for percentage of grounds). The script for the construction of
single arguments increases the percentage of arguments based on grounds substantially and strongly (F(1, 36) =
21.24; p < .05; η2 = .37). The script for the construction of argumentation sequences shows no effect on the
percentage of grounds (F(1, 36) = 0.02; n.s.). No interaction effect of both scripts could be found (F(1, 36) = 0.91;
Both scripts influenced the percentage of counterarguments (see table 1). The script for the construction of
argumentation sequences strongly affected the percentage of counterarguments (F(1, 36) = 9.08; p < .05; η2 = .20)
positively, as did the script for the construction of single arguments (F(1, 36) = 7.14; p < .05; η2 = .17). The two
scripts did not interact with regard to the percentage of counterarguments (F(1, 36) = 1.23; n.s.).
Table 1. Formal argumentative dimension by experimental group: Mean percentages and standard deviations of
grounds and counterarguments.
Experimental group M SD M SD
Control group 12.08 % 11.48 2.46 % 3.67
Script for the construction of
single arguments 33.80 % 11.19 5.36 % 8.07
Script for the construction of
argumentation sequences 16.36 % 17.78 5.99 % 3.95
Combined condition 30.64 % 6.10 13.00 % 6.59
The computer-supported collaboration scripts also affected the epistemic dimension (see table 2). With
regard to the application of new knowledge, the script for the construction of single arguments produced a
negative effect (F(1, 36) = 5.47; p < .05; η2 = .13). Neither a main effect of the script for the construction of
argumentation sequences (F(1, 36) = 1.91; n.s.) nor an interaction effect of both scripts on application of new
knowledge could be found (F(1, 36) = 0.00; n.s.).
The script for the construction of argumentation sequences significantly and strongly increases the
application of prior knowledge (F(1, 36) = 11.24; p < .05; η2 = .24). Neither a main effect of the script for the
construction of single arguments (F(1, 36) = 0.00; n.s.) nor an interaction effect of both scripts could be found (F(1,
36) = 0.90; n.s.) with respect to application of prior knowledge.
Table 2. Epistemic dimension by experimental group: Means and standard deviations of application of new
knowledge and application of prior knowledge.
Experimental group M SD M SD
Control group 7.97 3.45 13.80 7.93
Script for the construction of
single arguments 5.07 3.75 16.80 8.52
Script for the construction of
argumentation sequences 6.23 4.16 27.00 12.78
Combined condition 3.43 4.03 24.20 8.72
Both scripts successfully facilitated the specific processes of argumentative knowledge construction they
aimed at. Supported with the script for the construction of single arguments, the percentage of grounds doubles,
but still only one third of the claims is supported with grounds. Both scripts double the share of
counterarguments, but typically learners do not construct counterarguments, but argue in favor of one claim
only. Both scripts seem to have specific “side effects” on the epistemic dimension. Learners with the script for
the construction of single arguments less frequently show applications of new knowledge and learners with the
script for the construction of argumentation sequences show applications of prior knowledge more frequently
than learners without scripts. In order to illustrate these effects on the epistemic and formal dimension of
argumentative knowledge construction simultaneously, a discourse segment of a learning group supported by
the script for the construction of argumentation sequences is presented next. This discourse segment has been
selected to illustrate a discourse with a high frequency of application of prior knowledge.
The learning case that the learners work on in this example is about a pupil who is afraid not to make
mathematics class next year, because he thinks he is not talented for this subject (internal stable self-attribution).
His parents equally attribute their son’s failure on lack of talent (internal stable attribution of others), whereas
his teacher ascribes his failures to a lack of efforts (internal variable attribution of others). In the learning group
of three, 32 messages in two threads have been sent to discuss and analyze this case. This means that two
“argumentation” messages have been followed by 17 messages automatically labeled “counterargument” and 13
messages automatically labeled “integration”. The participants did not replace or modify any of the
automatically set subjects of the messages. Eight messages from one of the discussion threads have been
selected for the example illustration.
(1) Karin: Argument
If you think that you are not talented for math, you have little chance for success!
(2) Karin: Counterargument
What do you think? Can you compensate lack of talent by learning?
(3) Ella: Integration
I think you may compensate lack of talent momentarily by learning, than you will forget what
you have learned, because you are happy that you are done with learning.
(4) Karin: Counterargument
But you can learn in order to improve your performance, can’t you?
(5) Ella: Integration
Yes, but math is one of these subjects you can drill and practice, but only with the help of others,
because when you are learning all alone you lack certainty all of a sudden, you get scared and
you don’t know right from wrong anymore.
(6) Karin: Counterargument
That’s right, you might need help for learning, maybe friends who know their stuff better than
you. But you still can learn!
(7) Ella: Integration
Yes, he has to be at home and learn, but certainly only as a “MUST” – he won’t have fun.
(8) Karin: Counterargument
You may be more successful with fun things. But how many things that need to be learned are
fun to do?
Figure 3. Schematic representation of a segment of one of the discussion threads from an example discourse
that has been supported by the script for the construction of argumentation sequences.
In this example (see figure 3), the first message (1) of Karin labeled “Argument” is a short explanation of the
learning case which can be categorized as application of new knowledge, because the attribution theory states
that the attribution for failure on lack of talent decreases chances for success. Ignoring the label
“Counterargument”, the same learning partner, Karin, continues the discussion thread and (2) replies to herself
with a message containing questions that point beyond analyzing the case with the attribution theory. Karin’s
learning partner, Ella, (3) replies to that in a message labeled “Integration” and constructs a claim that is
supported by a ground, but shows application of prior knowledge rather than analyzing the case with adequate
concepts from attribution theory. Karin (4) replies with a message labeled “Counterargument”. Her message
actually contains a counterargument (without grounds and qualifiers) to Ella’s claim that learning compensates
lack of talent only momentarily, but Karin does not return to analyze the case with new knowledge, but rather
discusses other aspects of the learning case and applies prior knowledge. Another (5) message labeled
“Integration” from Ella follows. However, this message is actually rather a counterargument then an integration.
On the epistemic dimension, Ella does not apply new knowledge, but yet again introduces new aspects
(application of prior knowledge), namely instructional approaches towards the subject mathematics. Karin’s (6)
message “Counterargument” again actually contains a counterargument. On the epistemic dimension, she also
applies prior knowledge. Ella then (7) turns to other motivational approaches (application of prior knowledge) to
make her point that learning is of little help in this case, but again does not refer to the theory which is to be
learned. Her “Integration” message can be coded as counterargument. Karin finally (8) notes that motivation is
important for learning, but not sufficient to explain performance differences in different subjects (application of
prior knowledge). In opposing Ella, Karin constructs a counterargument.
First of all, it can be noted that learners apply an argument-counterargument sequence. The learners do not
always respond to the given labels of their messages in the intended manner, e.g., they construct a
counterargument even if their message has been automatically labeled “integration”. Learners do not always
follow the prescriptions of the script for the construction of argumentation sequences. But as the results show,
the computer-supported collaboration scripts still affect the processes of argumentative knowledge construction
in the intended direction.
With regard to the formal argumentative dimension, Ella claims that learning may not improve performance,
which she supports with various grounds (messages 3, 5, and 7). Karin constructs the counterargument that
learning may improve performance (messages 4, 6, and 8). With regard to the epistemic dimension, the
participants appear to wander off the actual task to analyze the case with the help of concepts from attribution
theory (application of new knowledge). Instead, learners apply prior knowledge starting with the second
message of this discussion thread. Karin is asking the question which leads learners to discuss their
epistemological beliefs on the efficacy of learning.
Research Question 2 on outcomes of argumentative knowledge construction
In order to answer research question 2, the influence of the two computer-supported collaboration scripts on the
outcomes of argumentative knowledge construction, namely domain-specific knowledge and knowledge on
argumentation was examined.
Neither the script for the construction of single arguments (F(1, 36) = 0.33; n.s.), nor the script for the
construction of argumentation sequences (F(1, 36) = 0.08; n.s.), nor the interaction of both scripts (F(1, 36) = 1.27;
n.s.) facilitated the acquisition of domain-specific knowledge. Learners of all four experimental conditions did
not differ with respect to the acquisition of domain-specific knowledge.
Knowledge on argumentation could be specifically facilitated with the scripts.
Table 4. Outcomes of argumentative knowledge construction by experimental group: Means and standard
deviations of domain-specific knowledge and knowledge on argumentation.
Knowledge on the construction of
knowledge single arguments argumentation sequences
Experimental group M SD M SD M SD
Control group 4.33 2.16 3.08 1.08 2.23 1.65
Script for the construction of
single arguments 4.70 1.49 4.17 1.55 2.03 1.69
Script for the construction of
argumentation sequences 4.90 2.52 2.70 1.21 5.25 1.05
Combined condition 3.77 2.12 4.78 0.75 4.55 0.85
The script for the construction of single arguments strongly facilitated knowledge on the construction of
single arguments (F(1, 36) = 17.97; p < .05; η2 = .33), whereas no effect of the script for construction of
argumentation sequences (F(1, 36) = 0.10; n.s.) nor an interaction effect of both scripts could be found (F(1, 36) =
The script for the construction of argumentation sequences strongly facilitated knowledge on the
construction of argumentation sequences (F(1, 36) = 41.50; p < .05; η2 = .54), whereas no effect of the script for
the construction of single arguments (F(1, 36) = 1.10; n.s.) nor an interaction effect of both scripts could be found
(F(1, 36) = 0.39; n.s.).
Although all experimental groups acquired a similar amount of domain-specific knowledge, both scripts
successfully facilitated the acquisition of knowledge on the construction of single arguments or the construction
of argumentation sequences. The learners were able to construct single arguments and argument sequences
depending on what the computer-supported collaboration script aimed at. The scripts did not interact and can be
combined to foster knowledge on the construction of single arguments as well as knowledge on the construction
of argumentation sequences at the same time.
Computer-supported collaborative learning can be realized in the curriculum of university studies and facilitated
with computer-supported collaboration scripts. Potential settings for CSCL in university lectures could be that
learners build small groups and work on problems together via the internet. Computer-supported collaboration
scripts can facilitate specific processes and outcomes of argumentative knowledge construction of students in
higher education. The analysis of the formal argumentative dimension of the discourse within the learning
groups of the control condition showed in line with other studies (Kuhn, 1991; Kuhn et al., 1997), that learners
hardly base their claims on grounds and hardly construct counterarguments. The computer-supported
collaboration scripts under investigation showed that they can improve the argumentative discourse quality of
students. Scripts could be integrated into a CSCL environment and proved to facilitate the percentage of grounds
and counterarguments that learners construct in argumentative discourse. Thus, the scripts improved the formal
argumentative dimension, but the scripts also influenced the epistemic dimension of argumentative knowledge
construction. Learners with the script for the construction of single arguments did not as frequently engage in
the application of new knowledge as learners without the script. The script for the construction of argumentation
sequences facilitated the application of prior knowledge. Computer-supported collaboration scripts may activate
prior knowledge and facilitate learners to come up with alternative explanations. An explanation for the pattern
of results with respect to the processes of argumentative knowledge construction is that the scripts provided a
structure that defined the activities of the learners with respect to the formal argumentative dimension, but
shifted the focus of learners away from the question with what kind of content this structure is supposed to be
filled. Thus, learners may have been more concerned to satisfy the affordances on the formal argumentative
dimension than on the epistemic dimension. Learners were challenged to find grounds and counterarguments,
but not supported with respect to the question on what contents these grounds and counterarguments should be
based on. Therefore, the prior knowledge may have been more easily available to learners to apply than the new
knowledge concepts that were to be learned.
In line with other studies (e.g., Kollar et al., 2003), the scripts facilitated knowledge on argumentation on the
specific levels they were aiming at, but did not facilitate domain-specific knowledge. Assumptions that the
construction of arguments also leads to the acquisition of domain-specific knowledge through elaboration of the
learning material cannot be fortified (Baker, 2003). An explanation for this is that learners supported with the
scripts focused on the construction of arguments, but may have based their arguments rarely on new knowledge,
which has been found to be related to acquisition of domain-specific knowledge (Weinberger, 2003). It can be
assumed that parts of these results can be traced back to the specific effects of the scripts on the processes of
argumentative knowledge construction. Therefore, future scripts might need to foster both the formal
argumentative and the epistemic dimension of argumentative knowledge construction in order to facilitate
students to learn to argue on a general level as well as within their domain. There are indications, however, that
knowledge on argumentation may foster acquisition of domain-specific knowledge in the long run (see
EUROSCALE project at http://www.euroscale.net; Kollar et al., 2003). In reference to this prior work, the
scripts applied in this study may be an apt instructional method to foster knowledge on argumentation in CSCL
environments and future argumentative knowledge construction scenarios.
Knowledge on argumentation is important for lifelong learning and should be further developed (Andriessen
et. al., 2003; Linn & Slotta, 2000). Based on this study, consequences for practitioners as well as researchers can
be drawn. In the educational practice of universities, specific scripts in problem-oriented environments may
endorse argumentation trainings. Teachers or coaches can integrate computer-supported scripts into ongoing
collaboration processes with little additional effort. With regard to future CSCL research, there is a lack of
studies on computer-supported collaboration scripts in field settings like classrooms or university lectures in
different domains. Additionally, we need to further investigate the combination of script components with
different goal dimensions, e.g., scripts that also facilitate the epistemic dimension of argumentative knowledge
construction (Weinberger et al., in press). We therefore suggest systematizing the effects of computer-supported
collaboration scripts in universities. An important step in making scripts available and applicable in different
university departments is to formalise script components that aim at specific aspects of argumentative
knowledge construction. Computer-supported collaboration scripts may thus support specific forms of
argumentative discourse within different domains and CSCL may become an endorsement to argumentative
knowledge construction in higher education.
The study has been funded by the DFG.
Anderson, R.C., & Pearson, P.D. (1984). A schema- theoretic view of basic processes in reading
comprehension. In P. D. Pearson (Ed.), Handbook of reading research (255-291). New York: Longman.
Andriessen, J. E. B., Baker, M., & Suthers, D. (Eds.). (2003). Arguing to learn. Confronting cognitions in
computer-supported collaborative learning environments. Dordrecht: Kluwer.
Baker, M. (2003). Computer-mediated argumentative interactions for the co-elaboration of scientific notions. In
J. Andriessen & M. Baker & D. Suthers (Eds.), Arguing to learn: confronting cognitions in computer-
supported collaborative learning environments (Vol. 1, pp. 1-25). Dordrecht: Kluwer.
Baker, M., & Lund, K. (1997). Promoting reflective interactions in a CSCL environment. Journal of Computer
Assisted Learning, 13, 175-193.
Bell, P. (1997). Using argument representations to make thinking visible for individuals and groups. In R. Hall,
N. Miyake, & N. Enyedy (Eds.), Proceedings of CSCL '97: The Second International Conference on
Computer Support for Collaborative Learning, (pp. 10-19). Toronto: University of Toronto Press.
Chi, M. T. H., DeLeeuw, n., Chiu, M. H., & LaVancher, C. (1994). Eliciting self-explanations improves
understanding. Cognitive Science, 18, 439-477.
Dillenbourg, P. (2002). Over-scripting CSCL: The risks of blending collaborative learning with instructional
design. In P. A. Kirschner (Ed.), Three worlds of CSCL: Can we support CSCL? (pp. 61-91). Heerlen:
Open Universiteit Nederland.
Fischer, F., Bruhn, J., Gräsel, C., & Mandl, H. (2002). Fostering collaborative knowledge construction with
visualization tools. Learning and Instruction, 12, 213-232.
Hoppe, H. U., Gaßner, K., Mühlenbrock, M., & Tewissen, F. (2000). Distributed visual language environments
for cooperation and learning. Journal Group Decision and Negotiation, 9, 3, 205-220.
Kirschner, P., Buckingham Shum, S. J., & Carr, C. S. (Eds.). (2003). Visualizing argumentation. Software tools
for collaborative and educational sense making. Dordrecht: Kluwer.
Kollar, I., Fischer, F., & Hesse, F. (2003). Cooperation scripts for computer-supported collaborative learning. In
B. Wasson & R. Baggetun & U. Hoppe & S. Ludvigsen (Eds.), Proceedings of the International
Conference on Computer Support for Collaborative Learning - CSCL 2003, COMMUNITY EVENTS -
Communication and Interaction (pp. 59-60). Bergen, Norway: InterMedia.
Kuhn, D. (1991). The skills of argument. Cambridge: Cambridge University Press.
Kuhn, D., Shaw, V., & Felton, M. (1997). Effects of dyadic interaction on argumentative reasoning. Cognition
and Instruction, 15, 3, 287-315.
Leitão, S. (2000). The potential of argument in knowledge building. Human Development, 43, 332-360.
Linn, M. C., & Slotta, J. D. (2000). WISE science. Educational Leadership, 29-32.
Marttunen, M., & Laurinen, L. (2001). Learning of argumentation skills in networked and face-to-face
environments. Instructional Science, 29, 127-153.
Nussbaum, E. M., Hartley, K., Sinatra, G. M., Reynolds, R. E., & Bendixen, L. D. (2002, April). Enhancing the
quality of on-line discussions. Paper presented at the Annual meeting of the American Educational
Research Association, New Orleans, LA.
O'Donnell, A. M. (1999). Structuring dyadic interaction through scripted cooperation. In A. M. O'Donnell & A.
King (Eds.), Cognitive perspectives on peer learning (pp. 179-196). Mahwah, NJ: Erlbaum.
Schmidt, H. G. (1993). Foundations of problem-based learning: some explanatory notes. Medical Education, 27,
Stein, N. L., & Bernas, R. S. (1999). The early emergence of argumentative knowledge and skill. In J.
Andriessen & P. Coirier (Eds.), Foundations of argumentative text processing (pp. 97-116). Amsterdam:
Amsterdam University Press.
Suthers, D. D., & Hundhausen, C. D. (2001). Learning by constructing collaborative representations: An
empirical comparison of three alternatives. In P. Dillenbourg & A. Eurelings & K. Hakkarainen (Eds.),
European perspectives on computer-supported collaborative learning (pp. 577-592). Maastricht, NL:
University of Maastricht.
Toulmin, S. (1958). The uses of argument. Cambridge: Cambridge University Press.
Weinberger, A. (2003). Scripts for computer-supported collaborative learning. Effects of social and epistemic
cooperation scripts on collaborative knowledge construction. Ludwig-Maximilian University, Munich.
Available at: http://edoc.ub.uni-muenchen.de/archive/00001120/01/Weinberger_Armin.pdf.
Weinberger, A. & Fischer, F. (in press). A framework to analyze argumentative knowledge construction in
computer-supported collaborative learning. Computers & Education.
Weinberger, A., Ertl, B., Fischer, F., & Mandl, H. (in press). Epistemic and social scripts in computer-supported
collaborative learning. Instructional Science.
Weiner, B. (1985). An attributional theory of achievement motivation and emotion. Psychological Review, 92,