Examining the Effects of Text-Only and Text-and-Visual Instructional ...

Examining the Effects of Text-Only and
Text-and-Visual Instructional Materials on the
Achievement of Field-Dependent and
Field-Independent Learners During
Problem-Solving with Modeling Software
Charoula Angeli
Nicos Valanides
Sixty-five undergraduates were classified into
field-dependent, field-mixed, and
field-independent learners, and were randomly
assigned to two groups: text-only and
text-and-visual. Participants in the text-only
group received a description of a model in
textual format, whereas participants in the
other group received the same description in
textual-and-visual format. Participants were
then asked to individually explore a computer
model, test hypotheses, and solve a problem
related to immigration policies. Their
problem-solving performance was analyzed
using a 3 × 2 analysis of variance (ANOVA).
Results showed that the text-and-visual group
outperformed the text-only group, that
performance was significantly related to
field-dependence–independence, and that there
was a significant interaction effect.
Specifically, field-independent learners in the
text-and-visual group outperformed
field-dependent and field-mixed learners in
both groups, and field-independent learners in
the text-only group. The findings indicate that
adding visuals to textual explanations can
enhance understanding, and that the
functional role of visuals depends on cognitive
differences.
The results of national and international assessments,
such as the National Assessment of Educational
Progress (National Center for Education
Statistics, 2000), the Third International Mathematics
and Science Study (Schmidt, McKnight, Cogan,
Jakwerth, & Houang, 1999), the Program for International
Student Assessment (Organization for
Economic Co-Operation and Development, 2000),
and the Science and Scientists project (Sjoberg,
2002), clearly indicate that students are not well
equipped to think and communicate effectively,
learn individually, or solve complex problems that
they may face in real life. Nonetheless, learning
how to think, communicate, and problem solve
can be taught, and greatly depends on classroom
practices (Bruer, 1993).
New interactive and computer-based technologies,
such as electronic communication systems,
visualization and dynamic systems
modeling tools, simulations, and networked
multimedia environments, can, for example, be
integrated in the classroom to scaffold and
amplify student thinking and learning
(Bransford, Brown, & Cocking, 2001). Several
researchers (e.g., Glass & Mackey, 1988; Haken,
1981; Jonassen & Reeves, 1996; Penner,
2000/2001) have asserted that dynamic systems
modeling tools are, perhaps, the most intellectually
demanding technologies that “enhance the
cognitive powers of human beings during thinking,
problem solving, and learning” (Jonassen &
Reeves, p. 693). Vygotsky (1978) pointed out that
the tools we use shape our experience and, consequently,
our thinking. Siding with this view,
ETR&D, Vol. 52, No. 4, 2004, pp. 23–36 ISSN 1042–1629 23

24 ETR&D, Vol. 52, No. 4
Brown, Collins, and Duguid (1989) stated that
knowledge is not objective but contextually situated,
and is fundamentally influenced by the
activity, context, and culture in which it is used.
A central implication of this situated view of
learning for the design of technology-enhanced
learning environments is that knowledge building
and understanding can be viewed as the
appropriation of tools allowing learners to build
on their initial conceptions, while being engaged
in a problem-solving activity (Jonassen & Land,
2000; Land & Hannafin, 1997; Rogoff, 1990).
In view of recognizing the importance of
understanding how computer-modeling tools
assist the learning process, it seems useful to
study the effects that different instructional
materials, textual or visual, may have on learner
performance during problem solving with these
tools. Undoubtedly, the phrase “one picture is
worth a thousand words is a core idea of visualization
and modeling” (Kali, 2002, p. 305) is true,
because appropriate visualizations can improve
learner perception of the objects or the ideas the
pictures represent. Neverthelesss, learning from
either textual or visual information is also
directly associated with representational preferences
and cognitive controls or cognitive styles.
As Salomon (1994) argued, “It would be impossible
to consider the interactions of media, cognition,
and learning without taking into account
the variety of ways in which individuals construct
meaning and acquire knowledge” (p.
xxii). Chinien and Boutin (1992/1993) also
asserted that individual differences become
important for researchers to consider when
studying the performance of individuals interacting
with technology to accomplish a task.
Individual differences may refer to differences
in cognitive ability, or cognitive control 1 (style)
representing patterns of thinking that regulate
and control the way individuals process and
reason about information (Jonassen &
Grabowski, 1993).
A well-documented and popular source of
cognitive difference is the construct of fielddependence–independence
(FD-I) (Dillon &
1. The words cognitive control and cognitive style are used
interchangeably in this article, as synonyms, without
differentiating between them.
Gabbard, 1998). FD-I is generally considered to
represent differences in learner visual perception,
or comprehension of information, due to
the effects of the encompassing field, or instructional
context, related to the complexity of the
problem-solving task and the instructional
materials (Morgan, 1997; Reiff, 1996; Witkin,
Moore, Goodenough, & Cox, 1977). FD-I
describes learners along a continuum such that
individuals at one end are considered to be fielddependent
(FD), and individuals at the other
end field-independent (FI). Individuals who fall
in the middle of the continuum are characterized
as field-mixed (FM) (Liu & Reed, 1994).
The key difference between FD and FI learners
is visual perceptiveness. FD learners who are
asked to identify a simple geometric figure that
is embedded in a complex figure will take longer
to identify the simple figure than FI learners, or
FD learners may not be able to do it at all. FD
learners are, thus, not visually perceptive and
have more difficulty in abstracting relevant
information from visual (or even textual)
instructional materials supporting more difficult
learning tasks (Canelos, Taylor, & Gates, 1980;
Liu & Reed, 1994; Lyons-Lawrence, 1994). Obviously,
FD learners are more influenced by the
prevailing field, and, thus, often fail to isolate
target information, because other information
tends to camouflage what they are looking for
(Jonassen & Grabowski, 1993).
The characteristics of FD and FI learners
appear to have important implications for
instructional design (Chinien & Boutin,
1992/1993). FI learners are more successful in
isolating target information from a complex
whole, and can process information with more
accurate performance on visual search tasks,
analyze ideas into their constituent parts, and
reorganize ideas into new configurations (Davis,
1991; Goodenough & Karp, 1961; Snowman &
Biehler, 2003). On the contrary, FD learners are
global, factually oriented, and traditional in
their thinking (Lambert, 1981; Tannenbaum,
1982). The ramifications, however, of FD-I on the
performance of learners interacting with computers
to accomplish a task are not well established,
and the results of research studies are still
inconclusive (Davis, 1991; Dillon & Gabbard,
1998), and, at times, contradictory.

EFFECTS OF INSTRUCTIONAL MATERIALS ON LEARNER ACHIEVEMENT 25
2. In the context of this study, cognitive coupling is the
relationship between the cognitive characteristics of the
learner and the corresponding cognitive demands of the
employed textual and visual materials.
For example, in a study undertaken by
Zehavi (1995), the results showed that FD students
could also benefit from computer-based
instruction when the software was designed
appropriately. Similarly, Liu and Reed (1994)
studied the extent to which field type could have
an effect on learner achievement in learning
English using a hypermedia instructional system
that was appropriately designed to
accommodate preferences of different field
types. It was found that FD and FI learners benefited
equally from the hypermedia instructional
system. On the other hand, there is
research evidence supporting the idea that field
type does have bearing on learning from
hypermedia. For example, Weller, Repman, and
Rooze (1994) found that the effectiveness of
hypermedia is correlated with cognitive control.
In an experiment with four treatments that varied
according to the presence of advance organizers
and navigational cues, FI learners
outperformed FD learners across all treatments.
Lyons-Lawrence (1994) also concluded in her
study that FD students are not well suited to
computerized instruction. Thus, it is still not
clear whether the medium of instruction has an
effect on the performances of FI and FD learners
or whether FI learners will always outperform
FD learners despite the medium of instruction
(Davis, 1991).
Given the equivocal results, more research
effort should be directed toward examining the
extent to which dynamic modeling tools are better
fitted to one type of learner or another.
Undoubtedly, the consideration of individual
differences in human-computer interaction
studies may provide not only guidance on how
computer tools can best be targeted at specific
types of learners in the classroom learning environment,
but also clear indications of how optimal
“cognitive coupling” 2 and better
performance of “joint cognitive systems,” can be
achieved (Dalal & Kasper, 1994; Dillon & Gabbard,
1998).
On the basis of the above rationale, the present
study was designed to investigate whether
different instructional materials, using only textual
or textual-and-visual representations, differentially
affect learner achievement during
problem solving with modeling software,
depending on learner cognitive control (i.e., FD-
I). More specifically, answers to the following
three questions were sought:
1. Do textual-only (T-O) and textual-and-visual
(T-V) instructional materials differentially
affect learner achievement during problem
solving with modeling software?
2. Does problem-solving performance with
modeling software relate to learner FD-I?
3. Do T-O and T-V instructional materials interact
with learner FD-I to differentially affect
their achievement during problem solving
with modeling software?
Participants
METHODOLOGY
Research participants were recruited from the
2001–2002 freshman class of teacher-education
students at a university. Specifically, of the 165
first-year teacher-education students, who at the
time were enrolled in four different sections of
an undergraduate-level technology course, 157
of them volunteered to participate in the study.
This task was one of three possible assignments
from which students could choose to fulfill the
course requirements.
Initially, prospective participants were asked
if they had any prior knowledge relating to
either dynamic systems modeling software or
immigration policies. None of the students
reported any familiarity with immigration policies.
However, two students, a male and a
female, who stated that they had some prior
knowledge with modeling software, were not
allowed to participate in the study. Before forming
the final sample of the study, prospective
participants were also administered the Hidden
Figures Test (HFT) (French, Ekstrom, & Price,
1963). Based on their HFT scores, most of them
were classified as either FD (55) or FM (79), and
only 23 as FI, learners. Therefore, we included
all 23 students who were identified as FI, and

26 ETR&D, Vol. 52, No. 4
randomly selected 24 FM and 24 FD students, in
an attempt to maximize the final sample of the
study without any serious threat to the validity
of the results. Of the 71 students who participated
in the study, data obtained from 6 (2 males
and 4 females) were used to pilot test the
research materials and procedures. Thus, only
the data from the remaining 65 participants (22
FD, 22 FM, and 21 FI learners) were used in the
main study, and the final sample was much
smaller than the initial pool of voluntary participants.
Of the 65 participants, 53 were female and
12 were male, to match the ratio of female to
male in the initial pool of 165 teacher-education
students.
Students from each group of learners (FD,
FM, and FI) were randomly assigned into two
groups, namely, text-only (T-O) and text-andvisual
(T-V), that differed in the type of materials
they received in order to solve a complex
task. The T-O group consisted of 11 students
from each group of FD, FM, and FI learners,
whereas the T-V group had 11 FD and 11 FM
learners, but only 10 FI learners.
Instruments
The HFT, which is one of the 72 tests in the kit of
factor referenced cognitive tests, was used to
determine participant field type (French et al.,
1963). The HFT has 32 questions, and is either
self- or group administered. It consists of two
separate parts; 12 min are allowed for answering
each part. The items require individuals to identify
or determine which one of five simple figures
is embedded in a more complex pattern.
One point is assigned to each correct answer to a
test item; the total test score ranges from 0–32.
The HFT has been used extensively in research,
is reliable, and is highly correlated (r = .67 to .88)
to the Group Embedded Figures Test (Witkin,
Oltman, Raskin, & Karp, 1971).
For the purpose of this study, the HFT was
administered to six groups of elementary-education
students, totaling 157 people. Their average
performance was 14.43 (SD = 5.42). Those who
scored 10 or lower were classified as FD, those
who scored from 11 to 21 as FM, and those who
scored 22 or higher as FI. The cut-off scores and
the resulting classification scheme were based
on the rationale that the FD-I construct describes
learners along a continuum (Liu & Reed, 1994),
and, consequently, studying the performance of
those who fall in the two extreme ends of the
continuum and those who fall in the middle
range would reveal any differences associated
with the construct. The results of the HFT indicated
that only 23 students were FI, whereas the
rest were either FD (55) or FM (79), and, thus, the
final sample of the study was smaller than the
initial pool of volunteers.
Description of Modeling Software
Model-It® is a dynamic systems modeling tool
that has been used with middle school, high
school, and college students with notable success
(Metcalf, Krajcik, & Soloway, 2000; Stratford,
Krajcik, & Soloway, 1998). Model-It was
used in this study to create a model about immigration
dynamics, as shown in Figure 1.
A model usually consists of entities, factors,
and relationships between factors. In the model
in Figure 1, there are two entities, Mexico and
the United States. Each entity has several factors
associated with it. Factors represent measurable
or calculable characteristics of the entities, such
as population, labor force, immigration flow,
immigration rate, and jobs. Finally, factors are
designated as causal or affected depending on
the direction of the relationship between them.
For example, as shown in Figure 2, Mexican
labor force is the affected factor and Mexican
population is the causal factor, because any
increase in the Mexican population will cause an
increase in the Mexican labor force.
After creating a model, the user may run it.
When a model is run, a timer, as shown in Figure
3, counts arbitrarily sized time steps, which
may represent a minute, an hour, or whatever
time interval the user may conceptualize. The
user can also test a model using graphical tools.
One tool, the meter (see Figure 3), displays the
value of the factor at the current time step. If a
factor is considered to be independent, its value
can be adjusted while the model is running.
Thus, the user may test a model at run time and
observe how it changes dynamically. As shown

EFFECTS OF INSTRUCTIONAL MATERIALS ON LEARNER ACHIEVEMENT 27
Figure 1
A model on immigration dynamics.
in Figure 3, there is also another tool, the simulation
graph, which presents a line graph displaying
how factors change over a series of time
steps.
Model-It supports relationships (Jackson,
Stratford, Krajcik, & Soloway, 1996) that can
model immediate effects in the value of the
affected factor due to a change in the value of the
causal factor that preceded it. As shown in Figure
2, Model-It also supports a qualitative, verbal
description of relationships, because changes
in a relationship may be defined in terms of two
orientations (i.e., increases or decreases) and different
variations (i.e., about the same, a lot, a little,
more and more, less and less).
Instructional Task
Participants had to individually explore the
model for solving a problem about immigration
policy, shown in Figure 1. In order to solve the
problem, students had to understand the underlying
structure of the model, that is, its entities,
the related factors, and the relationships among
them. Thereafter, they were to form hypotheses,
test them, evaluate their consequences, and
decide on the optimal solution to the problem.
The model presented the immigration situation
created by a gap in unemployment rates
between the United States and Mexico, and
opportunities for employment in the United
States. It showed cause-and-effect relationships
between several factors affecting the unemployment
rate in each country, such as population,
labor force, and job export. For example, the
model showed how an increase in Mexican population
would cause an increase in the Mexican
labor force, and, consequently, an increase in the
Mexican unemployment rate. The model also
showed how an increase in the Mexican unemployment
rate would cause an increase in the
movement of Mexicans to the United States, and
how this increase in immigration flow to the
United States would finally cause an increase in
the unemployment rate of the United States.
Participants were given four possible policies

28 ETR&D, Vol. 52, No. 4
Figure 2
Defining relationships between factors.
Figure 3
Running a model.

EFFECTS OF INSTRUCTIONAL MATERIALS ON LEARNER ACHIEVEMENT 29
to explore in Model-It: (a) Open Border, (b)
Closed Border, (c) Job Export, and (d) Immigration.
An open border policy encourages the
immigration of Mexican people to the United
States, and allows the free movement of businesses
and jobs from the United States to Mexico.
A closed border policy discourages the
immigration of Mexicans to the United States
and the movement of industries and jobs from
the United States to Mexico. A job export policy
creates disincentives for American businesses in
order to discourage their movement from the
United States to Mexico. In addition, this policy
implements trade barriers, so that Mexican
goods become more expensive to sell in the
United States. Lastly, an immigration policy
does not allow the immigration of Mexicans to
the United States, but it takes no action about the
movement of businesses and jobs from the
United States to Mexico. Students were asked to
form hypotheses based on these policies, and
test them using Model-It. Then, they were asked
to evaluate the results, and propose the policy
that should be adopted for the purpose of regulating,
as optimally as possible, the situation at
the Mexico–United States border.
Materials
Two sets of materials were used. Both sets
instructed the participants to assume the role of
a chief immigration officer responsible for
studying all matters concerning the Mexican-
United States immigration problem. The materials
explained that a team of research staff
prepared the model in Figure 1 to help the chief
immigration officer better understand the
dynamics of immigration policy. Both sets also
included Figure 1, followed by a description and
instructions for opening the file with the model
in Model-It. Lastly, the four immigration policies
were explained, and students in both
groups were asked to examine the model in
Model-It, conduct experiments, and write in the
space provided which policy they would
assume to better manage the situation at the
Mexico–United States border.
In the T-O set, the model was described only
in narrative (textual) form. In particular, the textual
description explained all cause-and-effect
relationships in the model, and, in particular,
how an increase in Mexican population would
cause an increase in the Mexican labor force and,
finally, an increase in the Mexican unemployment
rate. Accordingly, an increase in the Mexican
unemployment rate would cause an
increase in the movement of Mexicans to the
United States, and, ultimately, an increase in the
U.S. population, labor force, and unemployment
rate. In turn, an increase in the number of jobs
available in the United States would cause a
decrease in the U.S. unemployment rate. In contrast,
an increase in job exports from the United
States to Mexico would cause an increase in the
U.S. unemployment rate. Finally, an increase in
the movement of American businesses to Mexico
would cause a decrease in the Mexican
unemployment rate.
In the T-V set, the model was decomposed
into four smaller diagrams of the same form.
Each of the smaller diagrams was presented
along with a description in narrative (textual)
form explaining all cause-and-effect relationships
depicted in the diagram. For example, one
of the diagrams showed the relationships
among Mexican population, Mexican labor
force, Mexican unemployment rate, and Mexican
immigration flow to the United States. A textual
description of all cause-and-effect
relationships illustrated in the diagram followed.
Thus, the two sets of materials differed
only in how the underlying structure of the
model was explained. In the T-V set, the description
of the model was presented gradually,
using four diagrams (visuals) along with their
corresponding descriptions (textuals) in alternate
form, whereas, in the T-O set, the model
was described only in narrative (textual) form.
The specific textual and visual materials in
the two sets were assumed to be informationally
equivalent representations, because every information
item, which could be taken from Figure 1
and its description in narrative form (textual),
could also be taken from the four diagrams
(visuals) and their corresponding descriptions
(textuals). “Two representations are (in a taskspecific
sense) informationally equivalent if both
allow the extraction of the same information
required to solve the specific tasks” (Schnotz,

30 ETR&D, Vol. 52, No. 4
2002, p. 104). These representations were also
considered to be computationally equivalent,
because any task-specific information could be
retrieved from the descriptive representations
(text) as easily as from the depictive representations
(visuals) (Larkin & Simon, 1987). Text and
visuals are external representations, and their
contribution to learning is understood when
learners construct internal representations of the
content described in the text or shown in visuals.
about the problem. Every 15 min, research participants
were prompted to write on the last
page of their materials the current time and, next
to it, the word materials, if they were using the
materials to study the model, or the word Model-
It, if they were using the computer model in
Model-It. It was, thus, possible to collect, indirectly,
approximate data relating to the time that
the different groups of students spent studying the
model, the time spent using the model to solve the
given problem, and their total time on task.
Research Procedures
Data were collected in the fall semester of 2002.
The researchers administered the HFT during
three different 24-min sessions, scheduled at
times convenient for all parties. Subsequently,
participants were classified as FD, FM, or FI
learners, and a sample of 71 (23 FI, 24 FM, and 24
FD) participants was selected. Two students
from each field type were used to pilot test the
research materials and procedures. Data for the
pilot study were collected within five days, and
data collection for the actual study started seven
days after the pilot study ended. During the
seven-day elapsed time, the researchers revised
their materials and procedures based on the outcomes
of the pilot study.
Data for the actual study were collected in six
different sessions, scheduled at times convenient
for the researchers and the participants, over a
period of three weeks. In addition, participants
who were coming from the same class were
scheduled to participate in the same research
session in order to avoid diffusion of information
related to the study. During each two-hour
session, the researchers initially demonstrated
Model-It for 20 min and showed, using a different
model, how to run and test a model in
Model-It. Each participant was then given the
appropriate set of materials (T-O or T-V). Students
were instructed to use their instructional
materials and the computer model in Model-It
individually, to think about immigration
dynamics, investigate the effects of each policy,
and suggest which of the four policies constituted
the optimum solution to the United States–
Mexico border problem. Students could use only
the materials and the computer model to think
Time on Task
RESULTS
Initially, the collected data related to student
time on task were analyzed using a 3 (FD, FM,
FI) × 2 (T-O, T-V) multivariate analysis of variance.
However, no significant differences were
identified between the T-O and T-V groups or
between FD, FM, and FI groups in time spent to
study the model, to solve the problem using the
model, and the total time on task. Student total
average time on task in minutes for the T-O and
T-V groups was 73.48 (SD = 18.73) and 65.62 (SD
= 17.77), respectively, and for the three subgroups
of FD, FM, and FI learners, 68.86 (SD =
18.32), 70.68 (SD = 16.50), and 69.28 (SD = 21.46),
respectively. The results showed that students
completed their tasks well before the scheduled
two-hour session ended. Students in the T-V sessions
tended to spend less time completing the
task, but there were no significant differences
between the groups in time on task. Thus, no differences
in student problem-solving performance
could be attributed to differences in how
much time was spent on task, because all groups
of learners spent approximately the same
amount of time studying and using the model to
solve the immigration problem.
Problem-solving Performance
A rubric was constructed to evaluate student
problem-solving performances. The instrument
was constructed inductively using the constant
comparative analysis method (Glaser & Strauss,
1967; Strauss & Corbin, 1990), and was based on

EFFECTS OF INSTRUCTIONAL MATERIALS ON LEARNER ACHIEVEMENT 31
participant solutions of how to best regulate the
situation at the Mexico–United States border.
The goal of the constant comparative method is
to classify a participant’s answer into an appropriate
level. Initially, each answer is coded into
as many levels of analysis as possible. Gradually,
as each answer is constantly compared
with all other answers, the levels of the rubric, as
well as the properties of each level, start to
develop. The rubric that we developed to score
participant answers had three mutually exclusive
levels, and is shown in Table 1.
Based on the rubric, participant scores could
range from 1 (low performance) to 3 (high performance).
Learner performance was holistically
evaluated depending on three criteria: (a) if they
took into account and correctly interpreted the
simulated outcomes of the model; (b) if they
examined both the pros and cons of each policy;
and (c) if they considered the long-term effects of
each policy, and recognized that ramifying may
take a long time. For example, some learners
suggested as the best solution, the immigration
policy, because the simulated outcomes showed
that adoption of this policy could regulate the
uneven unemployment rates between Mexico
and the United States. In reality though, this policy
cannot be the best one to adopt because in the
long run, it will cause high unemployment rates
in the United States.
The researchers initially explained to two
graduate students the process of grading the
answers using the rubric in Table 1, and provided
appropriate explanations to their questions.
Then, the two raters independently
graded all answers, and the inter-rater reliability,
a Pearson r, between the two ratings was
found to be .87. The two raters and the researchers
discussed the observed differences between
the two raters and resolved, after discussion, the
existing differences.
Table 2 shows student mean problem-solving
performances for the T-O and T-V groups, and
for the three subgroups of FD, FM, and FI learners.
The average number of participants in each
treatment group was 11.
The results in Table 2 indicate that participants
in the T-V group scored, in general, higher
than those in the T-O group, but the effect attributed
to the textual-and-visual materials seems to
be dependent on learner field types. Specifically,
FI students seemed to outperform the other two
groups of learners in the T-V group, whereas
such differences did not seem to exist in the T-O
group. A 3 (FD, FM, FI) × 2 (T-O, T-V) ANOVA
was subsequently performed to identify any differences
related to the instructional materials or
the classifying variable, and their possible inter-
Table 1
Problem-solving performance scoring rubric.
3
a. Reaches a decision by correctly interpreting the simulated outcomes of the model.
b. Examines the consequences of all policies and identifies pros and cons of each policy.
c. Considers possible long-term effects of the full impact of each policy and recognizes that ramifying
may take a long time.
2
a. Reaches a decision by correctly interpreting the simulated outcomes of the model.
b. Examines the consequences of all policies and identifies pros and cons of each policy.
c. Does not consider possible long-term effects of the full impact of each policy and does not
recognize that ramifying may take a long time.
1
a. Reaches a decision, which is not based on accurate interpretations of the simulated outcomes of
the model.
b. Does not consider pros and cons of each policy and shows biased thinking.
c. Does not consider possible long-term effects of the full impact of each policy and does not recognize
that ramifying may take a long time.

32 ETR&D, Vol. 52, No. 4
action effect. The results of the ANOVA indicated
that students in the T-V group outperformed
those in the T-O group, F (1, 59) =
5.253, p = .025; that performance was significantly
related to FD-I, F (2, 59) = 5.658, p = .006;
and that there was also a significant interaction
effect between the instructional materials
groups and FD-I, F (2, 59) = 3.938, p = .025. The
interaction effect between instructional materials
and field type is shown in Figure 4.
Pairwise comparisons using t tests were then
conducted comparing the performance of FD,
FM, and FI learners in the T-O and T-V groups,
and FD, FM, and FI learners within each group
of T-O and T-V. The comparisons showed that
there were no significant differences between
FD (t = .310, p = .760) and FM (t = 0.000, p = 1.00)
learners in the two groups, but that FI learners in
the T-V group had significantly better performance
than FI learners in the T-O group (t =
3.542, p = .002). FI learners in the T-V group had
a significantly higher performance than both FD
(t = 3.135, p = .005) and FM (t = 3.880, p = .001)
learners in the T-V group, but there were no significant
differences between FI and FD (t = .349,
p = .731), FM and FD (t = 0.000, p = 1.00), and FI
and FM (t = .408, p = .687) learners in the T-O
group.
The magnitude of the superior performance
of FI learners in the T-V group in comparison
with FI learners in the T-O group can be estimated
using the effect size, which is the degree
of mean difference between the first and second
group of learners relative to, or divided by, the
standard deviation of the second group (Glass,
McGaw, & Smith, 1984). The magnitude of this
effect size (ES = +1.8) was very high, indicating
that the average FI learner in the T-V group was
at 1.8 standard deviations above the mean of FI
learners in the T-O group. Similarly, FI learners
in the T-V group had a statistically significant
higher performance than FD and FM learners in
the same group. The advantage for FI learners in
the T-V group over the mean performance of FD
learners in the same group was associated with a
large effect size of +1.38. When effect size is calculated,
departures from normality should
always be taken into consideration, especially
when sample sizes are small, as in the present
study (Feingold, 1992; Wilcox, 1995). Nevertheless,
the high magnitude of both effect sizes
points out that the diagrams inserted in the T-V
materials had a facilitating effect only for FI
learners, who outperformed all other groups of
learners in both the T-O and the T-V groups.
DISCUSSION AND IMPLICATIONS
The results strongly suggest that the effectiveness
of instructional materials depends on the
FD-I style of learners. There were no significant
differences in time spent to study the model,
time to solve the problem using the model, and
total time between the T-O and T-V groups or
between any pair of FD, FM, and FI groups.
However, the results showed that the average FI
learner in the T-V group scored 1.8 standard
deviations higher than the mean of FI learners in
the T-O group, and 1.38 standard deviations
higher than the mean of FD learners in the T-V
group. These results indicate that the visuals
inserted in the T-V materials had a strong
advantage on FI learners who outperformed FD
and FM learners in both groups, and FI learners
in the T-O group. Thus, the evidence suggests
that adding visuals in a spatial and timely coordination
with the textual information can
Table 2 Descriptive statistics of problem-solving achievement scores of students (n = 65).
Classification Based on HFT Scores
FD FM FI Total
M [SD] M [SD] M [SD] M [SD]
Intervention
Text Only 1.45 [.69] 1.45 [.52] 1.55 [.52] 1.48 [.57]
Text-and-Visual 1.55 [.69] 1.45 [.52] 2.50 [.71] 1.81 [.78]
Total 1.50 [.67] 1.45 [.51] 2.00 [.77] 1.65 [.69]

EFFECTS OF INSTRUCTIONAL MATERIALS ON LEARNER ACHIEVEMENT 33
Figure 4
Interaction effect between Field Dependence–Independence and type of
instructional materials.
enhance understanding, and that their functional
role depends on the field types of learners.
Dual coding theory (Clark & Paivio, 1991;
Paivio, 1986) attributes a facilitating effect of
visuals on learning, because words and sentences
are usually processed and encoded in the
verbal system. Visuals, however, are processed
and encoded in two cognitive subsystems, the
imagery and the verbal. Thus, the facilitating
effect of visuals is ascribed to the advantage of
dual coding as compared to single coding in
memory. The conjoint processing theory
(Kulhavy, Stock, & Kealy, 1993) emphasizes that
the simultaneous availability of textual and
visual information in working memory makes it
easier to make cross-connections between text
and visuals, and facilitates later retrieval of
information. Thus, visual displays can contribute
to learning for two reasons. First, storing
information in two codes, linguistic and visual,
may increase memory of that information
because it provides two paths to retrieve it from
long-term memory, and, second, visual representations
can be accessed as a whole and processed
in a simultaneous manner, whereas linguistic
representations are hierarchically
organized and processed sequentially.
The results of the study do not provide, however,
unequivocal support for the aforementioned
theoretical positions and the potential of
visual information to promote learning. “Visual
displays are considered tools for communication,
thinking, and learning that require specific
individual prerequisites. . . . in order to be used
effectively” (Schnotz, 2002, p. 102). Field type,
for example, represents learner preferential
modes of perceiving and processing information.
Thus, some individuals may fail to master
an instructional task when they encounter tasks
that require processing information in a way
that they are unable to accomplish, simply,
because they lack the information-processing
capabilities demanded by the task. Thus, FI
learners appeared to outperform FD and FM
learners in the T-V group, because the cognitive
style of FD and FM learners inhibited the func-

34 ETR&D, Vol. 52, No. 4
tioning of the appropriate information-processing
technique.
Content analysis of students’ work and their
arguments in support of their solutions corroborate
this conclusion. The majority of FI learners
in the T-V group articulated that they considered
the smaller diagrams as representing
graphically the relationships that were
described in the text, and that the model in Figure
1 comprised all the diagrams that had been
shown and explained gradually. As some FI
learners in the T-V group stated, “The same diagrams
were finally synthesized into one model,”
and “it was then easier to identify the existing
relationships and isolate relevant information
despite its complexity.” Similarly, other FI learners
thought that the diagrams provided hints or
scaffolds, which helped them to figure out easily
the constituent parts and relationships in the
model or to identify relevant target information.
FD learners in the T-V group found the task
complicated, and stated that too many causeand-effect
relationships were depicted in the
diagrams and described in the text, making it
difficult for them to extract what was relevant
for solving the problem. These ideas indicate
that FI learners are less influenced by the prevailing
field and can more easily extract information
from a complex field.
Thus, the results clearly suggest that instructional
treatments heavily depend on the FD-I
style learners. The interaction effect between
instructional materials and FD-I supports the
notion of cognitive coupling (Fitter & Sime,
1980). Proper cognitive coupling significantly
facilitates the interaction of the learner with the
instructional task (Moffat, Hampson, &
Hatzipantelis, 1998), which, in this study
entailed a higher degree of immersion in the
problem-solving task, assisted by Model-It.
Deeper immersion in model exploration
resulted in deeper cognitive processing of information,
and consequently, better problem-solving
performance.
From this perspective, “overall system effectiveness
is maximized when the human and the
intelligent computer-partner are conceived,
designed, analyzed, and evaluated as components
of a joint cognitive system” (Dalal &
Kasper, 1994, p. 678). Further research may
highlight the ways of achieving optimal cognitive
coupling and better performance of joint
cognitive systems. For example, learner cognitive
style and other cognitive factors need to be
taken into consideration, because they may
interfere with the desirable effects expected
from learning with dynamic modeling tools.
Even though cognitive style “has sparked the
interest of researchers concerned with instructional
development” (Greco & McClung, 1979,
p. 97), researchers have not yet addressed
adequately its implications for technologyenhanced
learning environments. The development
of new technologies constitutes a specific
challenge for the use of descriptive and depictive
representations. Learning from visual (and
textual) information seems to be associated with
individual representational preferences (for
example individuals tend to be verbalizers or
visualizers) and cognitive control, such as FD-I.
It is, however, premature to conclude that
matching learner cognitive control (style) will
result in better learning. It rather remains an
open question whether adapting instructional
materials, textual or visual, to aptitude-treatment
interaction effects (Cronbach & Snow,
1981) could be beneficial for thinking and learning.
Accommodating learner cognitive styles in
instruction may have not only cognitive benefits
but also cognitive costs. The problems arise not
only because “no matter how you try to make an
instructional treatment better for someone, you
will make it worse for someone else” (Snow,
1976, p. 292), but also because “no matter how
you try to make an instructional treatment better
in regard to one outcome, you will make it
worse for some other outcomes” (Messick, 1976,
p. 266). Clearly, instructional treatments guided
by aptitude-treatment interaction effects may
have beneficial effects on specific and predetermined
outcomes, but they may not be beneficial
if, in the long run, they do not allow learners to
experience other modes of cognitive functioning.
As Chinien and Boutin (1992/93) stated,
building cognitive style in the instructional
design process can be a promising approach for
accommodating individual differences, because
of cognitive functioning, but “it is equally
important to ascertain the impact of accommodating
for cognitive style on students of varying

EFFECTS OF INSTRUCTIONAL MATERIALS ON LEARNER ACHIEVEMENT 35
degrees of need of accommodation” (p. 308) by
providing the flexibility to attenuate cognitive
style biases in instructional materials.
Lastly, effective learning depends not only on
cognitive processing, but also on affective and
motivational factors. Research on learning from
descriptive and depictive representations “will
have to be conducted not only from a cognitive,
but also from an affective, motivational, and
social perspective to reach adequate educational
decisions” (Schnotz, 2002, p. 118). New generation
learners are exposed to massive information,
and have extensive experience with
electronic media and new kinds of information
presentation, and they may have different
expectations, attitudes, and processing habits,
which might influence their cognitive processing.
Thus, future research studies should be
carefully designed based on memory models
and theories of picture perception in consideration
of individual differences in visual perceptiveness,
so that the cognitive needs of learners
of any field type can be satisfied.
Charoula Angeli and Nicos Valanides are with the
Department of Education at the University of Cyprus.
REFERENCES
Bransford, J. D., Brown, A. L., & Cocking, R. R. (2001).
How people learn: Brain, mind, experience, and school.
Washington, D.C: National Academy Press.
Brown, J. S., Collins, A., & Duguid, P. (1989). Situated
cognition and the culture of learning. Educational
Researcher, 18(1), 32–41.
Bruer, J. T. (1993). Schools for thought: A science of learning
in the classroom. Cambridge, MA: The MIT Press.
Canelos, J., Taylor, W. D., & Gates, R. B. (1980). The
effects of three levels of visual stimulus complexity
on the information processing of field-dependents
and field-independents when acquiring information
for performance on three types of instructional
objectives. Journal of Instructional Psychology, 7, 65–
70.
Chinien, C. A., & Boutin, F. (1992/1993). Cognitive
style FD/I: An important learner characteristic for
educational technologies. Journal of Educational Technology
Systems, 21(4), 303–311.
Clark, J. M., & Paivio, A. (1991). Dual coding theory
and education. Educational Psychology Review, 3, 149–
210.
Cronbach, L. J., & Snow, R. E. (1981). Aptitudes and
instructional methods (2nd ed.). New York, NY:
Irvington.
Dalal, K. P., & Kasper, G. M. (1994). The design of joint
cognitive systems: effect of cognitive coupling on
performance. International Journal of Human-Computer
Studies, 40, 677–702.
Davis, J K. (1991). Educational implications of field
dependence-independence. In S. Wapner & J.
Demick (Eds.), Field dependence-independence: cognitive
styles across the lifespan (pp. 149–175). Hillsdale,
NJ: Lawrence Erlbaum Associates.
Dillon, A., & Gabbard, R. (1998). Hypermedia as an
educational technology: A review of the quantitative
research literature on learner comprehension, control,
and style. Review of Educational Research, 68(3),
322–349.
Feingold, A. (1992). Sex differences in variability in
intellectual abilities: A new look at an old controversy.
Review of Educational Research, 62, 61–84.
Fitter, M. J., & Sime, M. E. (1980). Responsibility and
shared decision making. In A. T. Smith and T. R. G.
Green (Eds.), Human Interaction with Computers (pp.
32–60). London: Academic Press.
French, J. W., Ekstrom, R. B., & Price, L. A. (1963). Kit
of reference tests for cognitive skills. Princeton: Educational
Testing Services.
Glaser, B. G., & Strauss, A. L. (1967). The discovery of
grounded theory: Strategies for qualitative research. Chicago,
IL: Aldine Publications.
Glass, L., & Mackey, M. (1988). From clocks to chaos.
Princeton: Princeton University Press.
Glass, G. V., McGaw, B., & Smith, M. L. (1984). Metaanalysis
in social research. London: Sage Publications.
Goodenough, D. R., & Karp, S. A. (1961). Field dependence
and intellectual functioning. Journal of Abnormal
and Social Psychology, 63, 241–246.
Greco, A. A., & McClung, C. (1979). Interaction
between attention directing and cognitive style. Educational
Communication and Technology Journal, 27(2),
97–102.
Haken, H. (1981). Chaos and order in nature. New York,
NY: Springer-Verlag.
Jackson, S., Stratford, S., Krajcik, J., & Soloway, E.
(1996). Making dynamic modeling accessible to precollege
science students. Interactive Learning Environments,
4, 233–257.
Jonassen, D. H., & Grabowski, B. L. (1993). Handbook of
individual differences, learning and instruction.
Hillsdale, NJ: Lawrence Erlbaum Associates.
Jonassen, D. H., & Land, S. M. (2000). Theoretical foundations
of learning environments. Mahwah, NJ: Lawrence
Erlbaum Associates.
Jonassen, D. H., & Reeves, T. C. (1996). Learning with
technology: Using computers as cognitive tools. In
D. H. Jonassen (Ed.), Handbook of research for educational
communications and technology (pp. 693–719).
New York, NY: Simon & Schuster Macmillan.
Kali, Y. (2002). CILT2002: Visualization and modeling.
Journal of Science and Technology Education, 13, 305–
310.
Kulhavy, R.W., Stock, W. A., & Kealy, W. A. (1993).

36 ETR&D, Vol. 52, No. 4
How geographic maps increase recall of instructional
text. Educational Technology Research and Development,
41(4), 47–62.
Lambert, T. (1981). Effects of structure in preinstructional
strategies on memory for sentences in fielddependent
individuals (Doctoral dissertation,
University of Southern California), Dissertation
Abstracts International, 42, 1063.
Land, S. M., & Hannafin, M. J. (1997). Patterns of
understanding with open-ended learning environments:
A qualitative study. Educational Technology
Research and Development, 45(2), 47–73.
Larkin, J. H., & Simon, H. A. (1987). Why a diagram is
(sometimes) worth ten thousand words. Cognitive
Science, 11, 65–99.
Liu, M., & Reed, W. M. (1994). The relationship
between the learning strategies and learning styles
in a hypermedia environment. Computers in Human
Behavior, 10(4), 419–434.
Lyons-Lawrence, C. L. (1994). Effect of learning style
on performance in using computer-based instruction
in office systems. The Delta Pi Epsilon Journal,
XXXVI(3), 166–175.
Messick, S. (Ed.). (1976). Individuality in learning: Implications
of cognitive styles and creativity for human development.
San Francisco, CA: Jossey-Bass.
Metcalf, J. S., Krajcik, J., & Soloway, E. (2000). Model-It:
A design retrospective. In M. J. Jacobson & R. B.
Kozma (Eds.), Innovations in science and mathematics
education (pp. 77–115). Mahwah, NJ: Lawrence
Erlbaum Associates.
Moffat, S. D., Hampson, E., & Hatzipantelis, M. (1998).
Navigation in a “virtual” maze: Sex differences and
correlation with psychometric measures of spatial
ability in humans. Evolution and Human Behavior, 19,
73–78.
Morgan, H. (1997). Cognitive styles and classroom learning.
Westport, CT: Praeger.
National Center for Education Statistics, U.S. Department
of Education. (2000). The condition of education
2000. Washington, DC: U.S. Government Printing
Office.
Organization for Economic Co-Operation and Development.
(2000). Knowledge and skills for life: First
results from PISA 2000. Available at:
http://www.oecd.org.
Paivio, A. (1986). Mental representations: A dual coding
approach. Oxford, UK: Oxford University Press.
Penner, D. E. (2000/2001). Cognition, computers, and
synthetic science: Building knowledge and meaning
through modeling. Review of Research in Education,
25, 1–36.
Reiff, J. (1996). At-risk middle level students or field
dependent learners. Clearing House, 69(4), 231–234.
Rogoff, B. (1990). Apprenticeship in thinking: Cognitive
development in social context. New York, NY: Oxford
University Press.
Salomon, G. (1994). Interaction of media, cognition, and
learning. Hillsdale, NJ: Lawrence Erlbaum Associates.
Schmidt, W. H., McKnight, C. C., Cogan, L. S.,
Jakwerth, P. M., & Houang, R. T. (1999). Facing the
consequences: Using TIMMS for a closer look at U.S.
mathematics and science education. Dordrecht, The
Netherlands: Kluwer Academic Publishers.
Schnotz, W. (2002). Towards an integrated view of
learning from text and visual displays. Educational
Psychology Review, 14(1), 101–120.
Sjoberg, S. (2002). Science for the children? Report from the
Science and Scientists project. Oslo, Norway: Acta
Didactica.
Snow, R. E. (1976). Aptitude-treatment interactions
and individualized alternatives in higher education.
In S. Messick (Ed.), Individuality in learning: Implications
of cognitive styles and creativity for human development.
San Francisco, CA: Jossey-Bass.
Snowman, J., & Biehler, R. (2003). Psychology applied to
teaching (10th ed.). Boston, MA: Houghton Mifflin
Company.
Stratford, S. J., Krajcik, J., & Soloway, E. (1998). Secondary
students’ dynamic modeling processes:
Analyzing, reasoning about, synthesizing, and testing
models of stream ecosystems. Journal of Science
Education and Technology, 7(3), 215–234.
Strauss, A. L., & Corbin, J. (1990). Basics of qualitative
research: Grounded theory procedures and techniques.
Newbury Park, CA: Sage.
Tannenbaum, R. K. (1982). An investigation of the relationship(s)
between selected instructional strategies
and identified field dependent and field independent
cognitive styles as evidence among high school
students enrolled in studies of nutrition (Doctoral
dissertation, St. John’s University), Dissertation
Abstracts International, 43, 68.
Vygotsky, L. S. (1978). Mind in society. Cambridge,
MA: Harvard University Press.
Weller, H. G., Repman, J., & Rooze, G. E. (1994). The
relationship of learning, behavior, and cognitive
style in hypermedia based instruction: Implications
for the design of HBI. Computers in the Schools,
10(3/4), 401–420.
Wilcox, R. R. (1995). ANOVA: A paradigm for lower
and misleading measures of effect sizes? Review of
Educational Research, 65, 51–77.
Witkin, H., Oltman, P., Raskin, E., & Karp, S. (1971). A
manual for the Embedded Figures Test. Palo Alto, CA:
Consulting Psychologists Press.
Witkin, H. A., Moore, C. A., Goodenough, D. R., &
Cox, P. W. (1977). Field dependent and field independent
cognitive styles and their educational
implications. Review of Educational Research, 47, 1–64.
Zehavi, N. (1995). Integrating software development
with research and teacher education. Computers in
the Schools, 11(3), 11–24.

Introduction to Part II of the Special Issue:
Design, Development and Implementation of
Electronic Learning Environments for
Collaborative Learning
Paul A. Kirschner
Development in society and business, and
related changes in higher education and lifelong
learning require educators and educational
designers/technologists to rethink education.
Examples of such changes are the growing
importance of achieving complex learning, the
integration of learning and work in education,
and the need for improved flexibility with regard
to time, place and individual needs. These
changes cannot simply be responded to by adding
technological solutions implemented according
to existing educational approaches. Instead, an
integrated view on e-learning is necessary, characterized
by the combination of pedagogical, technical,
social, and organizational factors. The final
four articles this special issue (which began with
articles on the design of and educational
approaches in electronic collaborative learning
environments and the role of authenticity in learning
and assessment, in the previous number:
Gulikers, 2004; Kirschner, 2004; Kirschner, Strijbos,
Kreijns, & Beers, 2004) discuss a theoretical basis
for collaborative learning in work-based settings
and present different aspects of a research and
design agenda for online collaborative learning.
In this number, Collis and Margaryan take e-
learning from the traditional school setting to
the corporate setting. In their article they show
how collaborative learning can take on special
forms in the on-going professional development
of engineers in a multinational corporation as a
tool for capturing experience, reusing it, and creating
new artifacts and solutions for workplace
applications. Reeves, Herrington, and Oliver
present a research agenda for collaborative
learning. In their view traditional “basic to
applied” research methods have provided an
insufficient basis for advancing the design and
implementation of advanced collaborative
learning environments. Instead, most of the significant
progress that has been made has been
accomplished through development research,
design experiments, or formative research. Elen,
the first discussant in this special issue, introduces
the notion of instructional design anchor
points (IDAPs) as the basis for instructional
design, arguing that research on IDAPs can
become more useful and influential when it
meets certain conditions. Finally, Wilson offers
an activity-based perspective on E-learning
environments, resulting in a flexible stance
toward instructional strategies, artifact design,
emergent activity, and learning outcomes.
Paul A. Kirschner [paul.kirschner@ou.nl] is with the
Educational Technology Expertise Center, Open
University of the Netherlands, P.O. Box 2960, 6401
DL Heerlen, The Netherlands. Voice: +31 45 5762361;
Fax: +31 45 5762901.
REFERENCES
Gulikers, J. T. M., Bastiaens, T. J., & Kirschner, P. A.
(2004). A five-dimensional framework for authentic
assessment, Educational Technology Research and
Development, 52(3), 67–86.
Kirschner, P. Introduction to Part I of two-part special
issue: Design, development, and implementation of
electronic learning environments for collaborative
learning, Educational Technology Research and
Development, 52(3), 39–46.
Kirschner, P., Strijbos, J.-W., Kreijns, K., & Beers, P. J.
(2004). Designing electronic collaborative learning
environments, Educational Technology Research
and Development, 52(3), 47–66.
ETR&D, Vol. 52, No. 4, 2004, p. 37 ISSN 1042–1629 37