Full text of DOI:10.1007/BF02505022

The Effects of Argumentation Scaffolds on
Argumentation and Problem Solving
Kyoo-Lak Cho
David H. Jonassen
An important skill in solving problems,
especially ill-structured problems, is the
production of coherent arguments to justify
solutions and actions. Because direct
instruction in argumentation has produced
inconsistent results and cannot effectively
support online learning, we examined the use
of online argumentation scaffolds to engage
and support coherent argumentation. In this
study, we showed that providing a
Problem solving requires argumentation
(Cerbin, 1988). Argumentation is a process of
making claims and providing justification for
the claims using evidence (Carr, 1999; Toulmin,
1958). Argumentation is an essential kind of in-
formal reasoning that is central to the intellec-
tual ability involved in solving problems,
making judgments and decisions, and formulat-
ing ideas and beliefs (Kuhn, 1991). Argumenta-
tion requires problem solvers to identify various
alternative perspectives, views, and opinions;
develop and select a preferred, reasonable solu-
tion; and support the solution with data and
evidence (Voss, Lawrence, & Engle, 1991). Ar-
gumentation is a variable that significantly
predicts student performance in both well-struc-
tured and ill-structured problems (Hong, Jonas-
sen, & McGee, in press).
constraint-based argumentation scaffold
during group problem-solving activities
increased the generation of coherent
arguments. The same scaffold further resulted
in significantly more problem-solving actions
during collaborative group discussions. The
effects of the scaffold varied for problem type.
Groups that solved ill-structured problems
produced more extensive arguments. When
solving ill-structured problems, students need
more argumentation support because of the
importance of generating and supporting
alternative solutions. The close relationship
between argumentation and problem solving,
especially ill-structured problem solving, is
significant. The effects of the argument scaffold
consistently transferred to the production of
arguments during individual problem solving.
Students used the familiar argumentation
scripts while solving problems individually.
Despite its importance, students are not
adept at constructing cogent arguments (Cerbin,
988). Specifically, they do not connect evidence
1
to claims via warrants (claim = tax cut will in-
crease savings; warrant = marginal propensity
to consume; evidence = decrease in savings last
year and increase in sales tax revenue), but
reasoning from claims to evidence is essential
for problem solving. Bell and Linn (2000) sug-
gest that conjecturing with warrants, as opposed
to descriptions, in order to support arguments
indicates that students are making scientific con-
jectures, which enables them to generate better
problem solutions.
How then do we facilitate learner develop-
ment of argumentation skills? Cerbin (1988)
proposed direct instruction of reasoning skills
based on an explicit model of argumentation.
That approach has been the standard method for
years. Leeman (1987) and Saunders (1994) advo-
ETR&D, Vol. 50, No. 3, 2002, pp. 5–22 ISSN 1042–1629
5

6
ETR&D, Vol. 50, No. 3
cated using Toulmin’s (1958) model of argument
in law education class to develop argumentation
skills. Several researchers have also advocated
direct instruction on the structure and notation
of argumentation (Knudson, 1991; Sanders,
Wiseman, & Gass, 1994; Yeh, 1998). However,
research findings show inconsistent results:
Direct instruction does not always improve ar-
gumentation skills as expected. Some research
indicates that direct instruction enhances ar-
gumentation skills (Sanders et al., 1994), where-
as other research demonstrates no positive
effects for direct instruction on improving ar-
gumentation skills (Knudson, 1991).
requiring them to preclassify their contributions
to an argumentative discussion. Many other
cognitive activities are also constraint based. For
example, reading researchers have explored
syntactic and lexical constraints on meaning
generation while parsing sentences. In problem-
solving research, constraints are the set of pos-
sible combinations of values between variables
in the problem that must be progressively
restricted (satisfied) during problem solving
(Darses, 1991; Richard, Poitrenaud, & Tijus,
1
993). Chi, Slotta, and de Leeuw (1994)
described constraint-based interactions between
values in the system. Those interactions can be
defined canonically, which is what CSCA sys-
tems do. Any conceptual system (from a simple
sentence to a complex conceptual domain) con-
sists of attributes with values that interact. Those
interactions impose constraints on the
psychological processes required to operate on
that system. Those constraints must be satisfied
or eliminated in order for the processes to be
completed. In this study, we sought to constrain
the nature of interactions among learners in an
online discussion forum by requiring them to
predetermine the nature of the message they
were contributing to the forum in argumentative
terms. The constraints were the predetermined
message types that modeled a form of argumen-
tation.
An alternative approach to developing ar-
gumentation skills is to scaffold argumentation
skills through the use of cognitive tools. Lajoie
and Lesgold (1992) demonstrated the cognitive
effects of Sherlock, a computer-based argumen-
tation generation tool in avionics troubleshoot-
ing, showing that the group that used Sherlock
out-performed the control group in terms of the
number of troubleshooting problems solved and
in the quality of processes. A qualitative analysis
of the treatment group revealed that the quan-
tity and quality of troubleshooting steps became
more expert-like as a function of experience with
Sherlock (Lajoie, 1993).
A new class of cognitive tools, computer-sup-
ported collaborative argumentation (CSCA)
software, has emerged for the purpose of scaf-
folding student’s seeking of warrants and
evidence for supporting claims. Scaffolding, for
purposes of this study, represents any
manipulation by the system of the task itself
So, constraint-based CSCA conversation scaf-
folds are prestructured forms of conversation
systems that impose different conversational on-
tologies onto the discussion. These ontologies
make explicit the constraints involved in the
conversation. They supply the explicit state-
ments of the interactions among the attributes in
the domain. In this study, we preclassified the
types of comments that students could make
while solving problems, which constrained the
nature of verbal interactions among conver-
sants. The Belvedere environment used in this
study, for example (see Figure 1), provides four
predefined argumentation constraints: (a)
hypothesis, (b) data, (c) principles, and (d) un-
specified; and three links: (a) for, (b) against, and
(
Jonassen, 1999). When scaffolding perfor-
mance, the system performs part of the task for
the student, supplants the student’s ability to
perform some part of the task by changing the
nature of the task or imposing the use of cogni-
tive tools that help the learner perform, or ad-
justs the nature or difficulty of the task. Scaffolds
are temporary frameworks to support student
performance beyond their capacities, in this
case, constructing arguments.
The rationale for using CSCA to scaffold ar-
gumentation during conversations is in the
theory of constraints. That is, CSCA scaffolds ar-
gumentation by constraining the nature of stu-
dent interactions between group members,
(
c) and (Suthers, 1998). These constraints form
the relations between the ideas that conversants
produce.
Limited research has examined the use of

ARGUMENTATION SCAFFOLDS
7
Figure 1
Illustration of Belvedere argument being constructed.
constraint-based argumentation scaffolds. Tan
and Allaire Forum, an asynchronous bulletin
board system (BBS). The content of the group
discussions was characterized in terms of their
constructive and argumentative contributions
and their focus on the meaning and application
of concepts. They found that Belvedere discus-
sions were the most argumentative because Bel-
vedere stimulates students to check and counter
each other’s information most frequently. This
finding implies that constraint-based tools are
able to scaffold argument development during
group problem-solving activities, however the
residual effects on content knowledge remain
unclear. Buckingham-Shum, MacLean, Bellotti,
and Hammond (1997) conducted a series of
qualitative studies in order to identify how QOC
(Questions, Options, and Criteria), a constraint-
based structure and notation system, supports
particular kinds of design problem solving and
hinders others. They reported that QOC
provided the most support when designers
dealt with poorly understood (ill-structured)
(
2000) reported that students who used Quest-
Map™, a constraint-based argumentation tool,
performed significantly better in stating
“
grounds” in their argumentation. Grounds are
one of the argument components of Toulmin’s
model of argument (Toulmin, Rieke, & Janik,
1
984). However, the research results did not
show significant effects of QuestMap™ on prob-
lem-solving skills. Suthers and Hundhausen
(
2001) found no differences between constraint-
based systems conveyed in graphical node-link
system used in this study versus a tabular layout
of the information. Using discourse analysis,
Veerman, Andriessen, and Kanselaar (1999)
compared and analyzed the content of small
group online collaborative discussions carried
out by college students. Three online learning
tools were provided as a means of communica-
tion and discussion, including Netmeeting, an
unstructured synchronous online chatting tool,
Belvedere, a structured synchronous CSCA tool,

8
ETR&D, Vol. 50, No. 3
design problems, unlike the findings of Tan
is, problem solvers must defend their chosen
solution by providing their beliefs or
opinions about the problem and presenting
data, warrants, and evidence in support of
their decision (Meacham & Emont, 1989).
Therefore, we predicted that constraint-
based groups solving ill-structured problems
would generate more argumentation and
problem-solving conversational elements
than those in the unconstrained discussion
groups.
(
2000). But the same QOC also distracted the
designers’ activities when they dealt with well-
understood (well-structured) design problems.
These research results are not consistent, in part,
because their structures and functions vary. Re-
search is needed to clarify these differences.
Purpose of Study
There are many issues that are left unresolved
by the limited amount of research on argumen-
tation scaffolds and problem solving. Those is-
sues are restated here in terms of four research
questions:
4. Do the effects of constraint-based scaffolds
transfer to individual argumentation perfor-
mance and to individual problem-solving
performance following treatment? That is,
does the quality of students’ arguments and
the quality of their problem solutions differ
in a transfer task as a function of the prior use
of a constraint-based argumentation scaf-
fold? No one has researched the transfer ef-
fects of argumentation on problem solving.
We believe that the efficacy of any scaffold is
determined by how well students can per-
form a task when the scaffold has been faded.
This question has not been tested in the re-
search. We predicted that the effects of
quality of argumentation and problem solv-
ing would transfer to individual problem-
solving performance.
1
. How do constraint-based scaffolds affect the
nature of argumentation during a problem-
solving activity when compared with an un-
constrained discussion board? How is the
quality of student arguments in the discus-
sion forum affected by the constraint-based
tool? We predicted that discussions in the
constraint-based groups would include more
argumentation than those in the uncon-
strained discussion groups.
2
. How do constraint-based scaffolds affect the
nature of problem-solving activity during
discussions? While the relationship of ar-
gumentation to problem solving is well es-
tablished conceptually, students’ arguments
have never been analyzed in terms of their
problem-solving activities. We predicted that
students in the constraint-based groups
would generate more argumentative conver-
sational elements than those in the uncon-
strained discussion groups.
Method
Participants
Participants were college students registered for
an undergraduate introductory economics
course at a major university in the eastern
United States. A total of 69 volunteers agreed to
participate in the beginning of the study. How-
ever, missing data from some participants
reduced the number in the final analyses to 60
participants. Among the participants, 32 were
male and 28 were female. There were 4 fresh-
men, 28 sophomores, 24 juniors, and 4 seniors.
This activity was a scheduled course require-
ment. Participation was voluntary. Students
could opt to complete an alternative assignment,
however, none did. Participants were assured of
the confidentiality of their responses.
3
. Do constraint -based discussions differential-
ly affect the nature of argumentation and
problem solving while solving well-struc-
tured versus ill-structured problems? Al-
though we believe that the role of
argumentation is more important in solving
ill-structured problems than in solving well-
structured ones, the very small amount of re-
search has produced mixed results. Because
ill-structured problems do not have a single,
preferred solution, problem solvers are more
obliged to produce arguments for a preferred
solution and counterarguments that rebut al-
ternative ones (Kuhn, 1991; Voss, 1988). That

ARGUMENTATION SCAFFOLDS
9
Research Design
States). What are the costs and benefits of a national
money? Does the relative political stability of the two
countries have any role in this decision? Explain.
A 2 × 2 completely randomized factorial design
was used for the study. All participants were
randomly assigned to work in groups of three
students. These small groups were also random-
ly assigned into one of four treatment groups ac-
cording to the two independent variables. These
between-subject variables included problem
type and type of discussion forum used by the
groups.
A total of 10 three-person groups solved the
well-structured problem during online group
discussions while 10 groups solved the ill-struc-
tured problem.
All students in all groups corresponded
during the problem-solving sessions through a
BBS that provided a hierarchically structured,
threaded discussion. Each participant con-
tributed comments to a discussion board estab-
lished for each group for each problem. The
threaded discussion showed the list of all the
messages with headings, so learners did not
have to search through old messages unrelated
to the discussion topic. In threaded discussions,
student messages are organized around topics
and subtopics (Klemm & Snell, 1996) that
emerge in the discussion. That is, none of the
discussions in the BBS was prestructured.
Rather, the structure emerged as students
responded to each other’s messages.
Participants solved two different types of
problems. Thirty participants in groups of three
solved well-structured problems, while the
other 30 solved ill-structured problems. In order
to distinguish between problem types, we used
a modification of McGrath’s (1984) task clas-
sification scheme, called the Task Circumplex, as
a theoretical framework for selecting two tasks,
a well-structured problem (referred to in the
Task Circumplex as an intellectivetask) and an ill-
structured problem (referred to in the Task Cir-
cumplex as
a decision-making task). Well-
structured problems have demonstrably correct
answers. Each problem required students to in-
vent, or select, and compute the correct answer.
The criterion for successful group performance
is the correct answer. The correct answers for
well-structured tasks are based either on cul-
tural norms and beliefs, logic and broadly
known factors, or on expert consensus (Laugh-
lin, 1980; McGrath, 1984). An example of a well-
structured (intellective) task used in the study is:
Half of the three-person groups (5 solving
well-structured and 5 solving ill-structured
problems) used only the BBS to collaboratively
solve their problems, while the other 10 three-
person groups (5 solving well-structured and 5
solving ill-structured problems) were provided
with a constraint-based tool, Belvedere, to struc-
ture their discussions while solving their
problems. Belvedere was developed by the
Learning Research and Development Center at
the University of Pittsburgh to support students
in the creation of socially constructed arguments
(Learning Research and Development Center,
1996). Belvedere provides a framework for or-
ganizing, displaying, and recording the ar-
gumentation process, so that the students can
easily develop their argument toward solving a
problem as they work with group members.
Specifically, Belvedere provides the four
predefined conversation nodes and three links
described earlier (see Figure 1; Suthers, 1998).
Constraints in this system are determined by the
types of nodes and the links between them.
Users are required to link their comments to an
already existing comment using one of the four
message types. For example, data and principles
are meant to modify hypotheses. Students use
In the country of Vietnam, the velocity of money is
constant. Real [gross domestic product] grows by 5%
per year, the money stock grows by 14% per year, and
the nominal interest rate is 11%. What is the real inter-
est rate?
Ill-structured tasks lack demonstrably correct
answers. Each problem required students to
select and/or resolve by consensus a preferred
alternative. The goal for ill-structured problems
is to achieve consensus on a collective decision.
An example of an ill-structured (decision-
making) task used in the study is:
Suppose you are advising a small country (such as Ber-
muda) on whether to print its own money or to use the
money of its larger neighbor (such as the United

1
0
ETR&D, Vol. 50, No. 3
these predefined boxes and links to develop
their arguments during a problem-solving ses-
sion (see Figure 1 for an example of a problem
being solved during the study). Students used
the tool to organize their ideas and the messages
that they posted to the BBS.
the BBS database for each of three problems
were analyzed for which components of ar-
gumentation were present in each posting using
a coding scheme adapted from Toulmin’s model
of argument. The model (Toulmin et al., 1984)
identifies five major components of argument,
including (a) claims, (b) grounds, (c) warrants,
This argumentation support is structurally
congruent with Toulmin’s model of argument
(
d) backings, and (e) rebuttals. The total num-
bers of messages coded in each category for each
group comprised one of the dependent vari-
ables.
(
(
Toulmin et al., 1984). Hypothesis in Belvedere
“We will have more inflation next year,” in Fig-
ure 1) corresponds to claim in Toulmin’s model,
because stating a hypothesis is like stating a
claim (i.e., making a prediction); data (“The
deficit is increasing,” in Figure 1) to ground, be-
cause stating measurement, observation, or cal-
culation provides evidence (grounds) for the
possible answer; principle (“Large deficit causes
inflation,” in Figure 1) to warrant, because it
describes why some data (grounds) are for or
against the possible answer (claim). Backing (im-
plying support for an argument) and rebuttal
Each message produced by every group was
classified by two coders without knowing the
identity of the group in order to determine the
number (frequency of occurrence) of each com-
ponent of argumentation according to the
coding scheme. Cohen’s (1968) kappa (0.70) was
calculated to establish inter-rater reliability of
the coding. The assessment was determined
based on consensus between coders. Decisions
about messages with different codes were recon-
ciled in discussions between the coders.
(
implying rejection of an argument) can be ex-
pressed with the links. If a statement is inconsis-
tent with another statement, learners use the
against. If a statement is consistent with another
statement, they use for. Learners can also use the
and relation to link two or more statements that
together support or contradict another state-
ment. The links are used to show how state-
ments are related to each other (Learning
Research and Development Center, 1996).
Assessment of the problem-solving process during
group discussion. All of the posted messages for
each group for all problems stored in the BBS
database were reanalyzed for their problem-
solving function by classifying each message
based on a coding scheme adapted from Poole
and Holmes’s (1995) Decision Function Coding
System (DFCS). The DFCS consists of seven
categories, including (a) problem definition
(PD), (b) orientation (OT), (c) criteria develop-
ment (CD), (d) solution development (SD), (e)
solution approval (SA), (f) solution critique (SC),
and (g) nontask statement (NS). Jonassen and
Kwon (2001) used a similar coding scheme to
compare problem-solving processes used in
computer-mediated communication versus
face-to-face communication. They reported
greater use of nontask, simple agreement (cor-
responding to solution approval), and simple
disagreement (corresponding to solution criti-
que) categories for both well-structured and ill-
structured tasks in the computer-mediated
group relative to the face-to-face group.
Instruments
Every message from the discussion boards was
saved while the students collaboratively solved
problems. These messages provided the data
used as dependent variables in the first three re-
search questions described earlier. Following
the collaborative group problem-solving ac-
tivities, each participant individually completed
a problem-solving activity (an essay describing a
solution and justification). These individual es-
says provided the data used as the dependent
variable in the final research question described
before. The following assessments were made of
the data generated in those two activities.
Each message was coded by two coders
without knowing the identity of the group in
order to determine the number of messages (fre-
quency of occurrence) in each category during
Assessment of quantity of student arguments during
group discussion. The posted messages stored in

ARGUMENTATION SCAFFOLDS
11
the problem-solving process. Inter-rater
reliability of the coding was fairly high (k = .714).
Decisions about messages with different codes
were reconciled between the coders.
money). The rubric for the ill-structured
problems analyzed student agreements or dis-
agreements with the position taken in the prob-
lem. The professor, a grader, was looking for
student interpretations of the claims, reasons to
believe those interpretations, and a prediction of
whether inflation is expected or unexpected. The
interpretations had to be backed by warrants
Assessment of the quality of argumentation. F ol -
lowing the experiment, each participant com-
pleted an individual problem-solving assign-
ment. Participants solved new problems that
were similar in structure and function to the
problems they solved during the treatment
period (either well-structured or ill-structured).
During this assignment, students wrote an in-
dividual essay to describe how they would solve
the problem, what their solutions were, and why
they used the problem-solving approach that
they did. During this assessment, students did
not have access to the online learning tools. This
assignment was used as a measure of transfer of
argumentation and problem-solving skills
developed during group problem-solving ac-
tivities. The individual written reports were
scored by two assessors without knowing the
identity of the participant, using the scoring
rubric in Table 1 in order to determine the
quality of argumentation based on Toulmin’s
model of argument (Toulmin et al., 1984). The
individual scores were achieved by summing
the number of points achieved in each argumen-
tation category (claims, grounds, warrants,
backings, and rebuttal). Coefficient alpha was
calculated (0.99) to establish interassessor
reliability of the score. Alpha was used because
the data were ordinal. The assessment was
determined based on consensus between asses-
sors. Decisions about messages with different
codes were made in discussions between the as-
sessors.
(
e.g., Inflation is a result of the government’s
refusal to pay its spending; Inflation decreases
buying power; Inflation rejects value of money).
Each individual written report was scored by
the professor and the researcher, without know-
ing the identity of the participant, using these
rubrics (alpha = .964). The assessment was deter-
mined based on consensus between assessors.
Decisions about messages with different codes
were made in discussions between the assessors.
Procedure
During the study, students in small groups
worked together to perform the well-structured
or ill-structured problem-solving tasks. Group
members met online. Although students were
not required to use university laboratories to
complete these activities, most did. Participants
in the threaded discussion groups used only the
BBS, while students in the Belvedere group con-
structed solutions and arguments that they then
corresponded to each other using the BBS. Each
week for three weeks, the groups solved one
economics problem. During the study, each
group solved three problems. The week after the
treatment period, all students completed the in-
dividual problem-solving activity in class.
During class time while the experiment was
conducted, the instructor lectured as normal but
did not intervene in student group work. The
professor answered some questions that stu-
dents raised, but only if he believed that the
questions were not directly related to the perfor-
mance on the task.
Assessment of problem-solving performance. E a c h
individual written report was also analyzed for
problem-solving performance. The rubric for the
well-structured task assessed student analyses
of different items of value introduced in the
problem for their economic function (store of
value, unit of account, or medium of exchange)
backed by a warrant (e.g., You can accumulate
debt with it; It is accepted in exchange for goods
and services; It is a property (fortune) that you
can sell to make money and purchase to spend
Analysis
To analyze the components of student argu-
ments during group discussion, multivariate
analyses of variance (MANOVAs) and uni-
variate analyses of variance (ANOVAs) were

1
2
ETR&D, Vol. 50, No. 3
Table 1
Rubrics for assessing quality of argumentation in student essays
Claims
Quality
Criteria
6
4
The writer states generalizations that are related to the proposition and which are clear and complete.
The writer states generalizations that are related to the propositions, but the assertions are not complete.
Enough information is available to figure out the writer’s intent, but much is left to the reader to
determine.
2
0
The writer makes generalizations that are related to the proposition, but the assertions lack specificity
or offer unclear referents. The writer leaves much for the reader to infer in order to determine the impact
of the claim.
No claim related to the proposition or unclear assertions.
Grounds
Quality
Criteria
6
4
The supporting data are compete, accurate, and relevant to the claim.
The data offered are relevant but not complete. The writer leaves much for the reader to infer from the
data. The writer may have offered the data without the complete citation, which would allow the reader
to determine the reliability of the data as evidence. The writer may offer data, which are not complete
enough to allow the reader to determine their significance.
2
0
The data or evidence are weak, inaccurate, or incomplete. E.g. a) an attempt at using a general principle
without establishing the truth of the principle; b) the use of examples from personal experience which
are not generalizable; c) the citation of data when no source is identified; and d) the use of obviously
biased or outdated material.
No supporting data are offered or the data are not related to the claim.
Warrants
Quality
Criteria
6
4
2
The writer explains the data in such a way that it is clear how they support the claim.
The writer explains the data in some way, but the explanation is not linked specifically to the claim.
The writer recognizes a need to connect the data to the claim and states some elaboration of data, but
the writer fails to make the connection. Or most rules and principles are not valid or relevant.
0
No rules and principles are offered.
Backings
Quality
Criteria
6
4
2
0
The writer states correct, relevant, and specific sources of warrants.
The writer states correct, relevant sources of warrants but the sources are very general, not specific.
The writer states incorrect, irrelevant sources of warrants.
No sources of warrants are given.
Rebuttals
Quality
Criteria
6
4
2
0
The writer states complete and systematic identification of constraints of solutions.
The writer identifies constraints of solutions but the constraints are not sufficient.
The writer offers few constraints of solutions but the constraints are not elaborated.
No recognition of constraints of solutions.
Note. Based on Toulmin’s model of argument (Toulmin, Rieke, & Janik, 1984).

ARGUMENTATION SCAFFOLDS
13
conducted. The mean frequency per group per
problem of the number of messages classified
into each category (claims, grounds, warrants,
backings, and rebuttals) was treated as depend-
ent variables. This analysis was conducted to
test the first and third research questions.
of argumentation based on Toulmin’s model of
argumentation (Toulmin et al., 1984) were
summed. The categories included claims,
grounds, warrants, backings, and rebuttals. In
order to test the first research question, a 2 × 2
MANOVA was conducted treating the five ar-
gument components as dependent variables.
The independent variables were the presence or
absence of the argumentation scaffold and the
kind of problem-solving task.
To analyze the components of the problem-
solving process during group discussion,
MANOVAs and ANOVAs were calculated.
Seven different variables from the coding
scheme adapted from Poole and Holmes (1995)
discussed earlier were treated as dependent
variables: (a) PD, (b) OT, (c) CD, (d) SD, (e) SA,
Descriptive statistics for student arguments
during group discussions are presented in Table
2
. The MANOVA showed main effects for the
(
f) SC, and (g) NS. The mean frequency of each
use of the scaffold and the type of problem, but
no significant interaction occurred. The
MANOVA showed that students using the ar-
gumentation scaffold, Belvedere, produced sig-
nificantly more argument components during
group discussions Λ = .769; F(5, 52) = 3.132, p =
variable was the unit of analysis. This analysis
was conducted to test the second and third re-
search questions.
To examine the quality of argumentation,
ANOVAs were calculated. The unit of analysis
was the score for argumentation quality, as-
sessed using a rubric (scale = 0–30). This analysis
was conducted to test the fourth research ques-
tion.
.
015, ES =.231) than subjects in the unscaffolded
discussion groups. Follow-up ANOVAs showed
that the groups using the scaffold produced sig-
nificantly more claims (F(1) = 4.888, MSE = 9.22,
p = .031, ES = .08) and grounds (F(1) = 9.326,
MSE = 6.43, p = .003, ES=0.14) than groups that
did not have access to the scaffold. No other
variables differed significantly between groups.
Almost no messages stating backings or rebut-
tals were generated by any of the groups, and
very few warrants were provided either. This
analysis indicates that the constraint-based scaf-
fold can affect student argumentation during
group discussions. Students using the scaffold
provided increased claims about how to solve
the problem and backed them up with more
grounds.
To investigate the performance in problem
solving, ANOVAs were performed. The unit of
analysis was the score for individual perfor-
mance in problem solving, assessed according to
the assessment rubric (possible six points per
question). The rubrics were specific to the ques-
tion. For the well-structured problems, the
rubrics focused on the store of value, unit of ac-
count, and medium of change. For the ill-struc-
tured problems, the rubrics focused on reasoned
agreement-disagreement with a solution look-
ing for specific economic principles employed,
the justification for those principles, and
whether inflation is expected or unexpected.
This analysis was conducted to test the fourth
research question.
The MANOVA also showed a main effect for
type of problem. We wanted to know if the type
of problem (well-structured vs. ill-structured)
affected the number of student arguments
during group discussions. We believed that dif-
ferences in the processes required to solve ill-
structured problems would engage more
argumentation than the solution of well-struc-
tured problems. The main effect showed that
groups solving ill-structured problems did in-
deed produce more arguments during group
discussion Λ = .774; F(5, 52) = 3.045, p = .017, ES
= .27) than students solving well-structured
problems. The only significant follow-up
RESULTS
Components of Argumentation During
Group Problem Solving
The first analysis was of the effects of the Bel-
vedere scaffold on group argumentation. In
order to perform this analysis, the contents of
the group discussions were classified and the
frequency of occurrence of the five components

1
4
ETR&D, Vol. 50, No. 3
ANOVA was the frequency of stating rebuttals
F(1) = 9.453, MSE = .51, p = .003, ES = .14).
lem-solving process during group discussions,
the contents of group discussions were classified
as one of seven components of the problem-solv-
ing process (Poole & Holmes, 1995). Table 3
shows the descriptive statistics for the frequency
of components of the problem-solving process
used during group discussions. A MANOVA
and follow-up ANOVAs were conducted for the
seven dependent variables.
(
Groups solving ill-structured tasks produced
more rebuttals than those solving well-struc-
tured problems. While the difference is statisti-
cally significant, it is not substantive, as neither
group produced many backings or rebuttals.
However, this result provides some evidence
that the type of problem affects the nature of stu-
dent argumentation during group discussion,
specifically in terms of their willingness to rebut
each other’s claims, which is an essential charac-
teristic of solving ill-structured problems.
Next, we tested the effects of the scaffold on
the creation of problem-oriented comments
during group discussions. The MANOVA
showed a main effect for the constraint-based
scaffold (Λ = .495; F(7, 50) = 7.285, p = .000, ES =
.
51). The groups using the scaffold produced
Components of Problem Solving During
Group Problem Solving
more problem-oriented comments during their
discussion. Specifically, follow-up ANOVAs
showed significantly more PD comments (F(1) =
10.048, MSE = 1.49, p = .002, ES = .15), OT com-
In order to determine how the argumentation
scaffold affected the components of the prob-
Table 2
Descriptive statistics for the components of student arguments during group
discussion
Threaded
SD
Scaffolded
SD
Total
SD
M
n
M
n
M
n
Well-structuredProblems
Claims
6.200
2.200
1.867
0.200
0.200
2.705
2.042
2.066
0.561
0.561
15
15
15
15
15
6.733
4.000
1.000
0.000
0.467
3.105
3.684
1.309
0.000
0.640
15
15
15
15
15
6.467
3.100
1.433
0.100
0.333
2.874
30
30
30
30
30
Grounds
Warrants
Backings
Rebuttals
3.067
1.756
0.403
0.607
Ill-structured Problems
Claims
4.667
3.067
2.067
0.333
0.800
2.992
2.017
1.710
0.617
0.676
15
15
15
15
15
7.600
5.267
1.467
1.400
1.000
3.312
1.981
1.598
3.135
0.926
15
15
15
15
15
6.133
4.167
1.767
0.867
0.900
3.442
2.260
1.654
2.285
0.803
30
30
30
30
30
Grounds
Warrants
Backings
Rebuttals
Total
Claims
5.433
2.633
1.967
0.267
0.500
3.105
2.042
1.866
0.583
0.682
30
30
30
30
30
7.167
4.633
1.233
0.700
0.733
3.185
2.977
1.455
2.292
0.828
30
30
30
30
30
6.300
3.633
1.600
0.483
0.617
3.148
2.724
1.699
1.672
0.761
60
60
60
60
60
Grounds
Warrants
Backings
Rebuttals
Note. M = Mean; SD = Standard Deviation; n = Number of participants

ARGUMENTATION SCAFFOLDS
15
Table 3
Descriptive statistics for the components of the problem-solving process during group
discussion
Argumentation Scaffold
Scaffolded
Threaded
SD
Total
SD
M
n
M
SD
n
M
n
Well-structuredProblems
PD
OT
CD
SD
SA
SC
NS
1.267
6.800
1.000
4.933
4.733
0.867
11.00
1.530
2.450
1.460
2.930
2.540
1.120
5.920
15
15
15
15
15
15
15
1.333
9.800
2.867
4.933
4.267
1.333
4.467
1.04
3.29
2.09
2.96
1.62
1.29
2.77
15
15
15
15
15
15
15
1.30
8.30
1.93
4.93
4.50
1.10
7.73
1.2
3.2
2.0
2.8
2.1
1.2
5.6
30
30
30
30
30
30
30
Ill-structured Problems
PD
OT
CD
SD
SA
SC
NS
0.800
0.775
1.930
0.737
2.330
1.790
0.743
2.170
15
15
15
15
15
15
15
2.733
10.130
1.533
6.067
7.133
2.600
8.333
1.38
3.72
1.50
1.83
3.35
1.63
3.61
15
15
15
15
15
15
15
1.76
7.16
1.06
4.63
4.90
1.86
5.76
1.4
4.1
1.2
2.5
3.4
1.4
3.9
30
30
30
30
30
30
30
4.200
0.600
3.200
2.667
1.133
3.200
Total
PD
OT
CD
SD
1.033
5.500
0.800
4.067
3.700
1.000
7.100
1.210
2.540
1.150
2.750
2.400
0.947
5.910
30
30
30
30
30
30
30
2.033
9.967
2.200
5.500
5.700
1.967
6.400
1.40
3.45
1.91
2.48
2.97
1.58
3.72
30
30
30
30
30
30
30
1.53
7.73
1.50
4.78
4.70
1.48
6.75
1.3
3.7
1.7
2.7
2.8
1.3
4.9
60
60
60
60
60
60
60
SA
SC
NS
Note: M = Mean; SD = Standard Deviation; n = Number of participants; PD = Problem Definition; OT = Orientation;
CD = Criteria Development; SD = Solution Development; SA = Solution Approval; SC = Solution Critique; NS = Nontask
Statement.
ments (F(1) = 34.702, MSE = 8.62, p = .000, ES =
38), CD comments (F(1) = 12.562, MSE = 2.34, p
produced more problem-solving communica-
tions than did the threaded discussion groups.
Not only did Belvedere affect the nature of ar-
gumentation in groups, it also appeared to have
functioned as an affordance for every problem-
solving process. The connection between ar-
gumentation and problem solving appears quite
strong in these results.
.
=
=
.001, ES = .18), SD comments (F(1) = 4.70, MSE
6.56, p = .034, ES = .08), SA comments (F(1) =
1
0.153, p = .002, ES = .15), and SC comments
(F(1) = 9.085, MSE = 5.91, p = .004, ES = .14)
produced by the groups using the argumenta-
tion scaffold. In fact, the scaffolded groups
produced more problem solving comments in
every category except NS. That is, in every
productive category of problem-solving be-
havior, the argumentation-scaffolded groups
Next, we tested the effects of problem type on
problem-solving processes manifest in the
group discussions. The MANOVA showed a
main effect for problem type. That is, the type of

1
6
ETR&D, Vol. 50, No. 3
problem significantly affected the nature of the
problem-solving process during group discus-
sion (Λ = .694; F(7, 50) = 3.153, p = .008, ES = .31).
Those groups solving ill-structured problems
made significantly more OT comments during
group discussions than those solving well-struc-
tured problems. Follow-up ANOVAs showed
that groups solving ill-structured problems
made significantly more SC comments (F(1) =
tured problems. Generally speaking, argumen-
tation scaffolding appears to affect ill- struc-
tured–problem
solving
more
than
well-structured–problemsolving.
Individual Argumentation and Problem
Solving on Transfer Task
5
.715, MSE = 5.91, p = .020, ES = .09), while those
The fourth research question focused on transfer
effects of treatments to individual problem-solv-
ing tasks. We wanted to know if students would
transfer the argumentation methods that had
been scaffolded during the group problem-solv-
ing activities to their individual problem-solving
activities. In order to answer that question, we
created individual well-structured and ill-struc-
tured problem-solving activities that students
completed following the group activities. We
analyzed the student essays both for problem-
solving performance and quality of argumenta-
tion. Each individual assignment was analyzed
according to a rubric (scale: 0–30) to assess the
quality of argumentation in the student essays.
Descriptive statistics for the quality of argumen-
tation are summarized in Table 4.
solving well-structured problems made sig-
nificantly more CD comments (F(1) = 4.814, MSE
=
standably more important to those resolving the
solutions to ill-structured problems, while those
solving well-structured problems focused more
on the CD for determining the correct answer.
2.34, p = .032, ES = .08). SCs were under-
The MANOVA also indicated a significant
treatment-by-task interaction for the com-
ponents of the problem-solving process during
group discussion between argumentation scaf-
fold and the type of problem (Λ =.429, F(7, 50) =
9
=
.521, p = .000). The frequency of stating PD (F(1)
8.753, p = .005, ES = .14), SD (F(1) = 4.700, p =
.034, ES = .08), SA (F(1) = 15.444, p = .000, ES =
.
22), and NS all showed significant interactions
(
F(1) = 33.649, p = .000, ES = .38) in ANOVAs.
A one-way ANOVA showed that students
who had used the constraint-based scaffold, Bel-
vedere, created higher quality argumentation in
their individual problem-solving exercises than
students who had used the threaded discussion
board (F(1, 56) = 4.797, MSE = .437, p = .033).
This result provides evidence that the argumen-
tation scaffold has a transfer effect on the quality
of argumentation during individual problem
solving.
For the first three variables, the effects of scaf-
folding were much more pronounced in the
groups solving ill-structured problems. For the
non-task comments, a disordinal interaction oc-
curred, with the groups without argumentation
scaffolding engaging in more non-task interac-
tions while solving well-structured problems
and the scaffolded groups engaging in more
non-task interactions while solving ill-struc-
Table 4
Descriptive statistics for the quality of argumentation in individual problem-solving
activities
Threaded
SD
Scaffolded
SD
Total
SD
M
n
M
n
M
n
Well-structured
8.267
4.71
6.03
6.23
15
8.933
5.11
3.64
6.96
15
8.600
4.84
5.50
6.70
30
problems
Ill-structured
problems
14.660
11.460
15
30
19.600
14.260
15
30
17.130
12.860
30
60
Total
Note. M = Mean; SD = Standard Deviation; n = Number of participants

ARGUMENTATION SCAFFOLDS
17
Another one-way ANOVA compared the
performance between those solving well-struc-
tured and ill-structured tasks. It showed that
students faced with ill-structured problems
created higher quality arguments in individual
problem solving (M = 4.53, SD = 1.57) than stu-
dents facing well-structured problems (M = 3.87,
SD = 1.01, F(1, 56) = 3.93, MSE = .030, p = .05).
Finally, we compared students’ problem-
solving performance on their individual prob-
lem-solving tasks using a rubric (scale: 0–6,
described earlier). Means are shown in Table 5.
A one-way ANOVA showed that students who
had used the argumentation scaffold, Belvedere,
did not perform significantly better (p =.08) on
their individual problem-solving tasks than stu-
dents who did not have access to the scaffold.
This result indicates that there is no conclusive
evidence that argumentation scaffold transfers
to individual problem-solving performance as
determined by the performance criteria we used.
claims produced by the students while solving
problems include:
Prices of the goods produced in the year 2010 in-
creased by 52 percent compared to the prices that the
goods in the economy sold for in 2000.
According to the traditional view, a debt-financed tax
cut will decrease public saving, private saving, and na-
tional saving.
The Ricardian view better predicts what will actually
happen in the economy than the traditional one when
tax cuts.
Examples of grounds produced by the stu-
dents while solving well-structured problems
include:
The implicit price deflator for GDP was calculated, and
the value is 1.52.
The tax cut stimulates consumer spending, and
government revenues from taxes have decreased.
DISCUSSION
Even though the government cuts the tax, it does not
mean that we have less tax burden. It just reschedules
the tax burden to the future.
Effects of Scaffolds on Argumentation
and Problem Solving
The results of this study showed that using a
constraint-based argumentation scaffold posi-
tively affected the ability of groups to collabora-
tively construct arguments in an online
environment. Although only claims and
grounds (evidence) were affected, these are two
of the most important components in argument
construction and also components of argumen-
tation in which students are weak. Examples of
Surprisingly, the unscaffolded group
produced more warrants (e.g., “The GDP can
give an estimate of how the economy is doing.
Thus, if GDP increases, the state of the economy
has gotten better, and if it decreases, the
economy has gotten worse”) than the scaffolded
group, perhaps because the scaffolded groups
were focusing on supporting their claims with
evidence. Our results were consistent with those
Table 5
Descriptive statistics for performance in problem solving
Threaded
SD
Scaffolded
SD
Total
SD
M
n
M
n
M
n
Well-structured
3.67
.89
1.58
1.30
15
4.067
1.100
1.490
1.358
15
3.87
1.08
1.59
1.35
30
problems
Ill-structured
problems
4.13
3.90
15
30
4.930
4.500
15
30
4.53
4.20
30
60
Total
Note. M = Mean; SD = Standard Deviation; n = Number of participants

1
8
ETR&D, Vol. 50, No. 3
of Tan (2000), who reported that a CSCA group
performed better than a control group in stating
grounds. It is also likely that performance in
generating warrants was mediated by the in-
structor in this study, whose lectures focused on
the principles that were being studied in the
unit. Rebuttals (rejecting others’ arguments) are
so foreign to college classrooms that students are
reluctant to produce them, even in computer-
mediated environments.
Are these postings hypotheses? I saw you all selected
hypotheses. Some are hypotheses and others are not.
Check them out.
An example of SD by the students while solv-
ing problems is:
Between the years 2000 and 2010, the prices have risen
by sixty percent, rising from 1 to 1.6 percentage points.
Economists call a price index with a fixed basket of a
good Laspeyres index. This can be measured by the
We also believe that scaffolded students
focused more on claims and grounds because
those are the most important elements of the
solution to the problems. That is, answers to
problems are typically stated in the form of
claims and justified by grounds, while warrants,
backings, and rebuttals play only a supportive
role. Premises and evidence are the primary en-
tities in argument construction.
[
consumer price index] because the CPI does not allow
the basket of goods to change.
Examples of SA produced by the students
while solving problems include:
I got the same results as you.
It’s basically what A’s answers were since we all
While this study confirmed the effects of the
constraint-based argumentation scaffold on the
components of argumentation during problem
solving, the more important finding related to
the effects on components of problem solving.
Among seven components of the problem-solv-
ing process that were measured, only one com-
ponent, NS, was not affected by the scaffold.
Groups supported by the scaffold produced
more PD, OT, CD, SD, SA, and SC. Examples of
PD produced by the students while solving
problems include:
agreed.
An example of SC produced by a student
while solving a problem is:
C, I saw what you got for this question but I disagree
with part D.
Yoking problem solving outcomes to formal
argumentation is a significant finding, especially
showing effects of an argumentation scaffold on
all forms of problem solving. Jonassen and
Kwon (2001) showed effects of computer con-
ferencing only for solution approvals and criti-
ques. However, in this study, the constraint-
based argumentation scaffold affected all of the
relevant components of problem solving, in-
dicating a more generalized effect of argumenta-
tion on problem solving.
The first task is a calculation question regarding GDP.
Usually, we calculated to solve them, but this task is a
little bit different from the tasks we did the last two
weeks.
Examples of OT produced by the students
while solving problems include:
Effects of Problem Type on
What do we begin with? I guess it’s GDP.
Argumentation and Problem Solving
How about solving step by step? That is, we deal with
the traditional view first and then with the Ricardian
one.
An extensive research base confirms the dif-
ferential effects of instruction on different tasks.
In this study, the groups solving ill-structured
problems produced more arguments than the
groups solving well-structured problems both
during group discussion and in the quality of ar-
gumentation in individual problem solving.
Examples of CD produced by the students
while solving include:
Why don’t each of us calculate them and compare each
other?

ARGUMENTATION SCAFFOLDS
19
Groups solving ill-structured problems also
focused more on problem-solving processes,
specifically stating criteria and using those
criteria to critique solutions. Examples of CD
produced by the students while solving ill-struc-
tured problems include:
I definitely agree with B. Don’t you think now is the
time when you make a decision?
We conclude that ill-structured problems are
more affected by argumentation than well-struc-
tured problems. This result is consistent with
many studies contending that when solving an
ill-structured problem, students are required to
choose a preferred solution and reject alternative
ones (Voss, 1988); develop their own argument
Important thing is how we explain and justify well.
Before we make a decision, let’s provide why we
believe so.
(
(
Jonassen, 1997); and defend their own solution
Meacham & Emont, 1989). Because no single,
Examples of SC produced by the students
while solving ill-structured problems include:
correct convergent solution exists in an ill-struc-
tured problem (Kitchener, 1983), a student in a
group problem-solving situation must argue to
persuade group members that his or her claims
are more reasonable and valid than others. Ill-
structured problems provide students with
more opportunities to make arguments than
well-structured problems.
I don’t think the governor can understand what
economists say to him exactly.
I don’t completely agree both because everyone will
save all of the money; some of it will be spent and some
of it will be saved.
Not surprisingly, the effects of problem type
were more pronounced on problem solving than
they were on argumentation. Although solving
ill-structured problems produced more argu-
ments during group discussions than solving
well-structured problems, the only significant
univariate effect was for stating rebuttals. The
use of rebuttals is an requirement for solving ill-
structured problems. The purpose of argumen-
tation in ill-structured problems is to persuade
or convince people to accept one’s claims (Fisher
and Sayles, 1966) and to reject other solution
claims. Successful argumentation thus depends
on the degree of how well one’s argument
makes others’ opinions change. While solving
ill-structured problems, multiple solutions and
solution paths may exist (Kitchener, 1983), and
there is no universal agreement on the ap-
propriate solution (Voss, 1988), so personal
opinions or beliefs about the problem are re-
quired in the process of interpretation
The effects of the argumentation scaffold
were also more pronounced in stating PD, SD,
and SA while solving ill-structured problems.
Examples of PD produced by the students while
solving ill-structured problems include:
This task seems to require us to make a decision. As I
remember, no absolute answer exists.
There are two important conditions we have to take a
close look at. Check them out.
An example of an SD produced by a student
while solving an ill-structured problem is:
With an increase in taxes equal to an increase in
government purchase, the interest rate and investment
will remain the same. The explain is that an increase in
taxes will shift the curve to the right while an increase
in government purchases by the same amount to the
left bringing it back to the equilibrium interest rate and
investment. One would expect the explanation to
depend on the marginal propensity to consume, and it
should, but in this case, when the increase in taxes and
government purchases are the same, they cancel each
other out.
(Meacham & Emont, 1989). It is easy to under-
stand why solving ill-structured problems re-
quires students to use rebuttals more than
solving well-structured problems. The use of
rebuttals in solving well-structured problems is
not as common because there is a single best
solution in well-structured problems (Simon,
Examples of SAs produced by the students
while solving ill-structured problems include:
1
978), so when students in a group problem-
I will accept B’s answer. It also looks OK to me.
solving situation reach an agreement about how

2
0
ETR&D, Vol. 50, No. 3
to solve the problem, no more counterargu-
ments are needed.
available in online environments to support ar-
gumentation. This study has provided evidence
that argumentation can be supported effectively
by online argumentation scaffolds and that the
production of better arguments directly affects
the problem-solving activities that students use.
Belvedere is but one tool for supporting ar-
gumentation. Because the nature of argumenta-
tion varies with the type of problem or task, it is
probable that the nature of the argumentation
support tool should be tailored to the task. The
online environments developed to date (e.g.,
Belvedere) have used the same kind of argu-
ment structure. Jonassen and Remidez (2002)
have described the development of a constraint-
based discussion environment onto which dif-
ferent rhetorical structures can be mapped to
support different kinds of conversations (e.g.,
argumentation, hermeneutic, literary analysis).
This type of environment can be tailored to the
nature of the task that is being supported by con-
versations. To the degree that educators expect
learners to engage in authentic learning in on-
line environments, they must provide learners
with more sensitive and sophisticated tools for
supporting the different kinds of cognitive tasks
implied by those problems.
Transfer Effects to Individual Problem
Solving
We also assessed any potential transfer of the
processes used during group problem solving to
individual problem solving. Students solved
problems individually following their group ex-
periences. While the scaffold clearly affected
quality of argumentation during individual
problem solving, it did not significantly affect
the individual problem-solving performance.
The transfer of argumentation skills from group
to individual performances was understandably
affected by the scaffold. However, because prob-
lem-solving performance was not directly scaf-
folded, it was less likely to appear in individual
performances. The rubrics used to assess prob-
lem-solving performance were more consistent
with the argumentation model than they were
with the Poole and Holmes (1995) problem-solv-
ing model. Finally, it is very likely that in-
dividual problem solving was mediated by
individual differences. Without the support of
peers during group activities, more of the
variance in performance was accounted for by a
variety of individual differences.
Kyoo-Lak Cho is with the Korean Research Institute
for Vocational Education & Training (KRIVET) in
Seoul, South Korea.
David H. Jonassen is Distinguished Professor of
Learning Technologies at the University of Missouri.
CONCLUSION
The situated learning movement in education
presupposes the importance of solving more
authentic problems during learning. Authentic
problems tend to be ill structured, as opposed to
the well-structured problems that are en-
countered in most K–16 education. Because ill-
structured problems are dialectical in nature,
with two or more opposing conceptualizations
of the problem or its solution result, the produc-
tion of arguments to support those differing con-
ceptualizations is essential. Supporting those
activities in online environments has proven
problematic. If it is important that students learn
to articulate the differing assumptions in sup-
port of arguments for whatever solution they
recommend, then online tools should be made
REFERENCES
Bell, P., & Linn, M.C. (2000). Scientific arguments as ar-
tifacts: Designing for learning from the Web with
KIE. International Journal of Science Education, 22(8),
97–817.
Buckingham Shum, S.J., MacLean, A., Bellotti, V.M.E.,
Hammond, N.V. (1997). Graphical argumentation
7
&
and design cognition. Human-Computer Interaction,
12, 267–300.
Carr, C.S. (1999). The effect of computer-supported col-
laborative argumentation (CSCA) on argumentation
skills in second-year law students. Unpublished doc-
toral dissertation. The Pennsylvania State Univer-
sity, Pennsylvania.
Cerbin, B. (1988). The nature and development of informal

ARGUMENTATION SCAFFOLDS
21
reasoning skills in college students. (ERIC Document
Reproduction Service No. ED 298 805)
Chi, M.T.H., Slotta, J.D., & de Leeuw, N.A. (1994).
From things to processes: A theory of conceptual
change for learning science concepts. Learning and
Instruction, 4, 27–43.
Cohen, J. (1968). Weighted kappa: Nominal scale
agreement with provisions for scales disagreement
of partial credit. Psychological Bulletin, 70, 213–220.
Darses, F. (1991). The constraint satisfaction approach
Farr & J. Psotka (Eds.), Intelligent instruction by com-
puter: Theory and practice (pp. 15–36). New York, NY:
Taylor & Francis.
Learning Research and Development Center. (1996).
Advanced cognitive tools for learning. [Online]. Avail-
able: http://advlearn/lrdc.pitt.edu/belvedere/in-
dex.html.
Leeman, R.W. (1987). Taking perspectives: Teaching criti-
cal thinking in the argumentation course. (ERIC Docu-
ment Reproduction Service No. ED 292 147)
McCann, T.M. (1989). Student argumentative writing
knowledge and ability at three grade levels. Research
in the Teaching of English, 23(1), 62–76.
to design:
A psychological investigation. Acta
Psychologica, 78, 307–325.
Fisher, W.R., & Sayles, E.M. (1966). The nature and
functions of argument. In G.R. Miller & T.R. Nilsen
McGrath, J.E. (1984). Groups: Interaction and perfor-
mance. Englewood Cliffs, NJ: Prentice-Hall.
(
Eds.), Perspectives on argumentation (pp. 2–22).
Chicago, IL: Scott, Foresman and Company.
Hong, N.S., Jonassen, D.H., & McGee, S. (in press).
Predictors of well-structured and ill-structured
problem solving in an astronomy simulation. Journal
of Research in Science Teaching.
Jonassen, D.H. (1997). Instructional design models for
well-structured and ill-structured problem-solving
learning outcomes. Educational Technology Research
and Development, 45(1), 65–94.
Jonassen, D.H. (1999). Designing constructivist learn-
ing environments. In C.M. Reigeluth, (Ed.). Instruc-
tional design theories and models: A new paradigm of
instructional technology, Vol 2. Mahwah, NJ:
Lawrence Erlbaum Associates.
Jonassen, D.H., & Kwon, H.I. (2001). Communication
patterns in computer-mediated vs. face-to-face
group problem solving. Educational Technology Re-
search and Development, 49(10), 35–52.
Meacham, J.A., & Emont, N.C. (1989). The interper-
sonal basis of everyday problem solving. In J.D. Sin-
nott (Ed.), Everyday problem solving: Theory and
application (pp. 7–23). New York, NY: Praeger.
Poole, M.S., & Holmes, M.E. (1995). Decision develop-
ment in computer-assisted group decision making.
Human Communication Research, 22(1), 90–127.
Richard, J.F., Poitrenaud, S., & Tijus, C. (1993). Prob-
lem-solving restructuration: Elimination of implicit
constraints. Cognitive Science, 17, 497–529.
Sanders, J.A., Wiseman, R.L., & Gass, R.H. (1994). Does
teaching argumentation facilitate critical thinking?
Communication Reports, 7(1), 27–35.
Saunders, K.M. (1994). Law as rhetoric, rhetoric as ar-
gument. Journal of Legal Education, 44(4), 566–578.
Simon, D.P. (1978). Information processing theory of
human problem solving. In D. Estes (Ed.), Handbook
of learning and cognitive processes. Hillsdale, NJ:
Lawrence Erlbaum Associates..
Jonassen, D.H., & Remidez, H. (2002, January). Map-
ping alternative discourse structures onto computer con-
ferences. Paper presented at the annual meeting of
Computer Support for Collaborative Learning,
Boulder, CO.
Suthers, D. (1998). Representations for scaffolding col-
laborative inquiry on ill-structured problems. Paper
presented at the 1998 AERA Annual Meeting, San
Diego, California.
Kitchner, K.S. (1983). Cognition, metacognition, and
epistemic cognition: A three-level model of cogni-
tive processing. Human Development, 26, 222–232.
Klemm, W.R., & Snell, J.R. (1996). Enriching computer-
mediated group learning by coupling construc-
tivism with collaborative learning. Journal of
Instructional Science and Technology, 1(2). [Online].
Suthers, D.D., & Hundhausen, C.D. (2001). Learning
by constructing collaborative representations: An
empirical comparison of three alternatives. In P. Dil-
lenbourg, A. Eurelings, & K. Hakkarainen (Eds.),
European perspectives on computer-supported collabora-
tive learning (pp. 577–584). Universiteit Masstricht,
Maastricht, Netherlands.
Available:
jist/vol1no2/article1.htm.
Knudson, R.E. (1991). Effects of instructional
strategies, grade and sex on students’ persuasive
writing. Journal of Experimental Education, 59(2), 141–
http://www.usq.edu.au/electpub/e-
Tan, S.C. (2000). Supporting collaborative problem solving
through computer-supported collaborative argumenta-
tion. Unpublished doctoral dissertation. The Pen-
nsylvania State University, Pennsylvania.
Toulmin, S. (1958). The uses of argument. Cambridge,
England: Cambridge University Press.
1
52.
Kuhn, D. (1991). The skills of argument. Cambridge, UK:
Cambridge University Press.
Lajoie, S.P. (1993). Computer environments as cogni-
tive tools for enhancing learning. In S.P. Lajoie & S.J.
Derry (Eds.), Computers as cognitive tools (pp. 261–
Toulmin, S.E., Rieke, R.D., & Janik, A. (1984). An intro-
duction to reasoning (2nd Ed.). New York, NY: Mac-
millan.
Veerman, A.L., Andriessen, J.E.B., & Kanselaar, G.
(1999, December). Collaborative learning through com-
puter-mediated argumentation. Paper presented at the
conference on Computer Supported Collaborative
Learning (CSCL 99), San Francisco, California.
Voss, J.F. (1988). Problem solving and reasoning in ill-
2
88). Hillsdale, NJ: Lawrence Erlbaum Associates.
Lajoie, S.P., & Lesgold, A. (1992). Apprenticeship train-
ing in the workplace: A computer coached practice
environment as a new form of apprenticeship. In M.

2
2
ETR&D, Vol. 50, No. 3
structured domains. In C. Antaki (Ed.), Analyzing
everyday explanation: A casebook of methods (pp. 74–
3). London, UK: SAGE Publications.
Voss, J.F., Lawrence, J.A., & Engle, R.A. (1991). From
representation to decision: An analysis of problem
solving in international relations. In R.J. Sternberg &
P.A. French (Eds.), Complex problem solving (pp. 119–
157). Hillsdale, NJ: Lawrence Erlbaum Associates.
Yeh, S.S. (1998). Empowering education: Teaching ar-
gumentative writing to cultural minority middle-
school students. Research in the Teaching of English,
33(1), 49–83.
9