Learning from text with diagrams: Promoting mental model development and inference generation. Journal of Educational Psychology, 98, 182-197

Article (PDF Available)inJournal of Educational Psychology 98:182-197 · February 2006with 1,156 Reads
DOI: 10.1037/0022-0663.98.1.182
Cite this publication
Two experiments investigated learning outcomes and comprehension processes when students learned about the heart and circulatory system using (a) text only, (b) text with simplified diagrams designed to highlight important structural relations, or (c) text with more detailed diagrams reflecting a more accurate representation. Experiment 1 found that both types of diagrams supported mental model development, but simplified diagrams best supported factual learning. Experiment 2 replicated learning effects from Experiment 1 and tested the influence of diagrams on novices' comprehension processes. Protocol analyses indicated that both types of diagrams supported inference generation and reduced comprehension errors, but simplified diagrams most strongly supported information integration during learning. Visual representations appear to be most effective when they are designed to support the cognitive processes necessary for deep comprehension.
Learning From Text With Diagrams: Promoting Mental Model
Development and Inference Generation
Kirsten R. Butcher
University of Colorado at Boulder
Two experiments investigated learning outcomes and comprehension processes when students learned
about the heart and circulatory system using (a) text only, (b) text with simplified diagrams designed to
highlight important structural relations, or (c) text with more detailed diagrams reflecting a more accurate
representation. Experiment 1 found that both types of diagrams supported mental model development,
but simplified diagrams best supported factual learning. Experiment 2 replicated learning effects from
Experiment 1 and tested the influence of diagrams on novices’ comprehension processes. Protocol
analyses indicated that both types of diagrams supported inference generation and reduced comprehen-
sion errors, but simplified diagrams most strongly supported information integration during learning.
Visual representations appear to be most effective when they are designed to support the cognitive
processes necessary for deep comprehension.
Keywords: learning, diagrams, self-explanation, mental models, comprehension processes
As multimedia technology becomes increasingly popular in
formal and informal educational settings, the importance of re-
search investigating learning with visual and verbal materials takes
on added value. Understanding the ways in which visual materials
influence learning will be essential to developing multimedia tools
with consistent and predictable benefits. Advancing technology
has meant that multimedia often now includes complex forms of
interactive and computationally intensive presentations; however,
multimedia can be more simply defined as any presentation that
includes verbal and visual information (Mayer, 2001).
In practice, basic types of multimedia—such as pictures and
text—still appear to be frequently used. Currently, many digital
and print materials use pictures, diagrams, and text as their primary
communication format. But how does the visual representation of
information influence learning? Can changes in comprehension
processes account for the impact of diagrams on learning? The
purpose of this research was to investigate potential effects of
different diagram representations on students’ learning outcomes
and comprehension processes when diagrams were added to a
science text.
Learning With Text and Diagrams
Early research on pictures and text consistently demonstrated
that students learned more after reading illustrated versus nonil-
lustrated text (for a review, see Levie & Lentz, 1982). These
studies predominantly administered memory measures for the
source materials, including multiple-choice and fill-in-the-blank
tests. But a long history of cognitive research has distinguished
between rote memorization and deeper understanding (e.g., Brans-
ford & Franks, 1971; Hilgard, Irvine, & Whipple, 1953; Olander,
1941; Tyler, 1959; see Kintsch, 1998, for a discussion). Deeper
learning— evidenced by measures that assess application and
transfer of information—was not widely tested in early research on
illustrations. However, one early set of studies (see Dwyer, 1967,
1968, 1975) did test both memory and deep comprehension for an
illustrated text. Dwyer (1967, 1968, 1975) found benefits for
illustrated text when testing visual production (having participants
draw a diagram and label diagram components after learning),
component identification, and multiple choice but found no com-
prehension advantage when comparing the illustrated text with
text-only conditions.
The lack of a comprehension advantage with diagrams in
Dwyer’s (1967, 1968, 1975) early studies can be contrasted with
the more recent body of research on multimedia comprehension
developed by Mayer and his colleagues (for a summary, see
Mayer, 2001). In a large number of studies investigating multime-
dia effects on learning, this research has found a general (although
not completely unanimous) advantage for memory (e.g., Mayer,
1989; Mayer & Gallini, 1990) and an overwhelming and consistent
advantage for deep comprehension (e.g., Mayer & Anderson,
1992; Mayer & Gallini, 1990) when students receive a multimedia
(verbal and visual) presentation as opposed to a text-only presen-
tation. Mayer (2001, 2003) referred to these benefits as the mul-
timedia effect. In these studies, memory effects are indicated by
performance on recall or recognition questions, and comprehen-
sion performance is measured by transfer questions that draw on
the learning material but whose answers are not explicitly ad-
Kirsten R. Butcher, Department of Psychology, University of Colorado
at Boulder.
This work was supported in part by the National Science Foundation
Information Technology Research Grant EIA-0121201 and by a research
grant from the Institute of Cognitive Science, University of Colorado at
Boulder. This work is based on the 2003 doctoral dissertation of Kirsten R.
Butcher at the University of Colorado at Boulder. I thank Walter Kintsch
for helpful advice and critical reviews of this work, Noe¨lle Lavoie for
skilled assistance in data scoring, and Gregory Carey for expert consulta-
tion on statistical methods.
Correspondence concerning this article should be addressed to Kirsten
R. Butcher, who is now at the Pittsburgh Learning Research & Develop-
ment Center, University of Pittsburgh, 3939 O’Hara Street, Room 558,
Pittsburgh, PA 15260. E-mail: kbutcher@pitt.edu
Journal of Educational Psychology Copyright 2006 by the American Psychological Association
2006, Vol. 98, No. 1, 182–197 0022-0663/06/$12.00 DOI: 10.1037/0022-0663.98.1.182
dressed; transfer questions require students to generate new solu-
tions or applications based on their learning.
Consistent comprehension benefits found in recent multimedia
research likely highlight the progress that has been made in iden-
tifying effective forms of visual and verbal information for learn-
ing. A variety of multimedia learning studies by Mayer and his
colleagues (discussed in Mayer, 2001) have demonstrated the
importance of carefully designed multimedia presentations and
have generated a collection of important principles that predict
useful multimedia conditions for learning. These principles include
spatial contiguity, temporal contiguity, coherence, modality, re-
dundancy, and individual differences (see Mayer, 2001). A thor-
ough discussion of these principles is beyond the scope of this
work, but Mayer and his colleagues (e.g., Mayer, 2001) have
repeatedly found improved learning performance when multime-
dia principles are followed rather than violated. As discussed later,
the multimedia principle most relevant to this work is the coher-
ence principle. This principle states that learning materials should
include only relevant multimedia and should avoid irrelevant pic-
tures, sounds, or words (Mayer, 2001; Mayer, Heiser, & Lonn,
2001; Moreno & Mayer, 2000).
Multimedia principles can be used to optimize learning by
maximizing the effectiveness of multimedia design. These princi-
ples help govern the choices of multimedia content—for example,
in choosing relevant rather than irrelevant pictures—as well as its
presentation format. However, principles governing the format of
individual elements in learning materials have not been as strongly
or consistently identified. For example, what type of diagram
representation best supports learning from text?
What Makes Diagrams Difficult to Understand?
Previous research has suggested that different visual represen-
tations are not equally effective for all learners. Hegarty, Carpen-
ter, and Just (1991) proposed that diagrams become more difficult
to interpret as they become increasingly schematic. Schematic
diagrams often depict abstract relationships (as in the case of Venn
diagrams or flowcharts) and, thus, do not preserve the physical
relationships present in the source information. Schematic dia-
grams can be difficult to interpret because the reader or learner
must be able to understand and make use of abstract visual con-
ventions to interpret them correctly. Indeed, previous research has
demonstrated the importance of experience in using abstract rep-
resentations effectively. For example, Petre and Green (1993)
studied the use of electronic circuit schematics— diagrams that
make heavy use of abstract conventions such as symmetry and
proximity for functional association— by novice and expert elec-
tronics designers. Petre and Green’s results overwhelmingly indi-
cated that, unlike experts, novices were unable to make use of
abstract notation such as logical groupings and also were unable to
determine what was important versus irrelevant in the schematic
Compared with schematics, iconic diagrams are less abstract
and usually depict a close correspondence between the diagram
and the concrete object that it is intended to represent (Hegarty et
al., 1991). In iconic diagrams, structural and relational information
between components is central, and, as a result, iconic diagrams
rely less on knowledge conventions for their interpretation. How-
ever, the structural relations in iconic diagrams are not necessarily
transparent to all learners; the correspondence between an iconic
diagram and the physical object it depicts means that increasing
complexity in a physical object requires increasing complexity in
an iconic diagram. Simplification of this complexity may result in
additional abstraction that removes some of the correspondence
between the diagram and its physical object, but it is likely that
additional abstraction will benefit the learner when simplification
makes relevant components of the diagram more clear or easily
Diagram Complexity
Although many learners might assume that a faithful depiction
of an object would be most useful for learning, previous research
has demonstrated that the use of realism is not always a benefit in
iconic diagrams. Parkhurst and Dwyer (1983) found that additional
realism (as depicted in shaded drawings of the human heart)
hindered low- and medium-IQ participants’ learning outcomes.
This effect may be due to the possibility that lower ability learners
have difficulty processing richer and possibly redundant informa-
tion (Winn, 1987). However, an overall conclusion is complicated
by other studies showing that less able students perform better
when some elements of pictorial realism are added to diagrams
(Holliday, Brunner, & Donais, 1977; Moyer, Sowder, Threadgill-
Sowder, & Moyer, 1984).
Discrepancies in research results likely stem from difficulty in
determining the amount of appropriate complexity and realism in
diagrams for lower knowledge learners. To some extent, complex-
ity and realism are confounded in diagrams because adding real-
istic detail almost always increases diagram complexity. Con-
versely, simplifying iconic diagrams often involves introducing
some form of abstraction—in structure or function—to the con-
crete object being represented. As has been suggested by other
researchers (e.g., Koedinger & Anderson, 1990; Larkin & Simon,
1987; Seel & Strittmatter, 1989), the power of iconic representa-
tions may be that they make structural relations explicit, but
learners with varied knowledge may not extract structural relations
from complex iconic diagrams equally effectively. Thus, increased
detail may help lower knowledge learners when they have diffi-
culty extracting relevant diagram information, but the same detail
may hinder them when it is not helpful in locating relevant visual
information or in matching diagram elements to text.
Because comprehension of text with diagrams requires the se-
lection of relevant components, the integration of related verbal
and visual information, and the representation of such an integra-
tion (e.g., Kalyuga, Chandler, & Sweller, 2000; Mayer, 2001), it is
likely that the representational complexity of diagrams plays an
important role during comprehension. Thus, it was hypothesized
that making a diagram’s structural relations more explicit (by
abstracting structural information in the service of a functional
explanation) would support learning in lower knowledge students
better than a fully explicit diagram.
Reduction of a diagram’s representational complexity can be
seen as an extension of the multimedia principle of coherence
proposed by Mayer and his colleagues (Mayer et al., 2001; Moreno
& Mayer, 2000). The coherence effect summarizes research find-
ings showing that student learning is supported when extraneous
materials are removed from multimedia lessons. Extraneous ma-
terials include interesting but informationally irrelevant words,
pictures, sounds, and music. For example, video clips of a light-
ning storm embedded in a lesson on how lightning forms hinders
student learning rather than supporting it (Mayer et al., 2001).
Related to seductive details, irrelevant materials can be anything
that students find interesting but that is not necessary to the
cause-and-effect explanation at hand.
The coherence effect can be extended to include those aspects of
a diagram that are not directly necessary to understanding an
explanation. Applying the coherence effect to elements of a dia-
gram would prompt the removal of extraneous information, so that
the diagram includes only explanation-relevant components. How-
ever, simply removing elements of the diagram that are not ex-
plicitly necessary to an explanation often will not make structural
relations more explicit. Relevant diagrams with inherent complex-
ity likely tax working memory resources during learning. When
diagram elements must be processed interactively, rather than
serially and in isolation, working memory load is increased (Carl-
son, Chandler, & Sweller, 2003; Sweller, Chandler, Tierney, &
Cooper, 1990). Simplifying the structural information in a diagram
may reduce working memory load in two ways: first, by reducing
visual search difficulty and, second, by emphasizing the key func-
tions of the process at hand through a clear visual representation.
In the current study, structural simplifications to a diagram of the
human heart were used in order to highlight the functional pro-
cesses necessary for a correct understanding of the heart and
circulatory system. Because previous research has suggested that
diagrams can make structural relations explicit and allow learners
to process information simultaneously rather than sequentially
(Koedinger & Anderson, 1990; Larkin & Simon, 1987), it was
hypothesized that diagrams would best support learning when they
simplify the structure of a system in order to highlight important
relationships between component parts. Specifically, simplified
diagrams should best support understanding of the heart and cir-
culatory system as well as memory of the to-be-learned
The current work can be seen as testing an additional prediction
to that suggested by the coherence effect: specifically, that visual
material will be most effective if it has been designed to highlight
the relational information contained in an explanation, even if this
means abstracting and simplifying the structure of the physical
object being represented. Additional abstraction in the service of
highlighting structural relationships should make diagrams eas-
ier—rather than harder—to understand.
The current study varies only the visual portion of the learning
materials in order to pinpoint effects of the visual representations
used in the learning materials. Two types of informationally equiv-
alent iconic diagrams were tested: (a) a set of simplified diagrams
that did not faithfully represent the heart’s anatomy but that were
designed to clearly represent essential structural relations neces-
sary for mental model development and (b) a set of more detailed
diagrams that preserved the anatomy of the heart and circulatory
system (see Figure 1).
In the current experiments, I refer to simplified and detailed
diagrams, but it should be noted that this reflects a choice in
terminology. The reader is cautioned that these labels reflect the
relative difference between the diagrams rather than an absolute
definition of representational complexity. An alternative set of
appropriate terms for the diagrams could be functional versus
structural diagrams, reflecting the type of information that each
diagram emphasizes. Simplified diagrams reflect modified anat-
omy to highlight functional workings of the heart, whereas the
detailed diagrams more faithfully reflect the heart’s anatomical
structure. The diagrams could also be described by terms based on
research using the STEAMER training system (Hollan, Hutchins,
& Weitzman, 1984), where conceptual fidelity reflects the extent
to which a simulation depicts the mental models that are needed to
understand and to use a system, and physical fidelity emphasizes
its actual, physical characteristics. In the current experiments,
simplified diagrams are more tied to conceptual fidelity, whereas
detailed diagrams reflect more physical fidelity.
Diagram Representation and Learner Background
Although simplifying the representation of structural relations in
a diagram should simplify its interpretation, it is not always the
case that easier or simpler materials always benefit learners. The
body of research concerning background knowledge and learning
materials suggests that a diagram’s influence on learning may
depend on both the complexity of the diagram and the prior
knowledge of the learner.
Kalyuga et al. (2000) found that the form of multimedia repre-
sentation that is most useful for learners can change as participants
become more knowledgeable in a domain. In fact, Kalyuga, Ayres,
Chandler, and Sweller (2003) found that expert students benefited
most from the same presentation format that was most unhelpful to
novice students (and vice versa); highly trained students benefited
most from a visual-only presentation and least from a presentation
with ample text, whereas novices showed the opposite pattern.
Kalyuga et al. (2003) referred to the changing pattern of optimal
materials based on a learner’s knowledge as the expertise reversal
The finding that the optimal format of visual material may
depend on a learner’s background knowledge is consistent with
findings from text comprehension research, where a learner’s prior
knowledge has been found to have clear impact on learning from
different text materials. Text comprehension research has shown
that readers with adequate domain knowledge benefit from more
difficult texts; whereas novices need clear and explicit materials,
more expert students often need additional challenges to fully
engage in the learning task. For example, unlike low-knowledge
learners, high-knowledge readers learn more (as indicated by sit-
uation model measures) from low-coherence texts (McNamara,
Kintsch, Songer, & Kintsch, 1996; McNamara & Kintsch, 1996).
Figure 1. a: A simplified diagram, depicting blood flow from the atria
into the ventricles. b: A detailed diagram, depicting the same process.
The body of previous research on background knowledge and
the optimal format of learning materials suggests that the useful-
ness of certain diagrams may depend on a learner’s background
knowledge, with higher knowledge learners requiring more diffi-
cult materials in order to optimize their learning, and lower knowl-
edge learners requiring materials that are more explicit and clear.
Thus, it was hypothesized that high-knowledge participants would
benefit most from the detailed diagrams and that low-knowledge
participants would benefit most from the simplified diagrams. In
the current research, student background knowledge is measured
by the accuracy of students’ existing mental models of the heart
and circulatory system. Experiment 1 investigates the effects of
diagram representation and learner background knowledge on
learning outcomes, including mental model development, memory
for information, and transfer of knowledge.
Whereas Experiment 1 seeks to explore aspects of the multime-
dia effect, Experiment 2 delves deeper into cognitive reasons
underlying the multimedia effect by assessing potential changes in
comprehension processes resulting from the use of diagrams dur-
ing text learning. Specifically, Experiment 2 was designed to
determine how diagrams might affect the processes that students
perform during learning. Currently, little is known about the ways
in which diagrams affect comprehension processes and whether a
diagram’s representation may influence these processes. In order
to directly measure the influence of diagrams on cognitive pro-
cesses, Experiment 2 uses self-explaining methodology (e.g., Chi,
2000; Chi & VanLehn, 1991) and verbal protocol analyses to
assess the influence of diagrams on student learning.
Self-explaining has been used extensively by Chi and her col-
leagues (for a summary, see Chi, 1997) and refers to a type of
verbal protocol in which participants explain the content of a text
during learning. The goal of the explanations is for students
actively to make sense of what they are learning (Chi, 2000). To
clarify terminology, the term self-explaining is used to refer to the
process of generating statements during learning by explaining the
content of learning materials to oneself, self-explanation is used to
refer to a single statement generated when self-explaining, and
self-explanation inferences refer to knowledge inferences made
when students are self-explaining (Chi, 2000). The content of
verbal protocols generated during self-explaining represents a
learner’s knowledge in the context of that learner’s mental repre-
sentation of the problem or domain. For example, by analyzing the
content of self-explanations generated during learning, Chi and
VanLehn (1991) were able to analyze what knowledge gave stu-
dents the ability to solve physics problems and to understand the
knowledge from which students generated self-explanations. Thus,
self-explanations can create a picture of what knowledge is im-
portant in comprehension and can indicate how and when infer-
ences are derived.
It is the characterization of prior knowledge, the use of such
knowledge during learning, and the ability to represent compre-
hension processes during learning that makes self-explaining an
ideal method for testing the cognitive impact of diagrams during
text comprehension. Analysis of verbal protocols generated during
self-explaining can provide a window on the comprehension pro-
cesses in which students engage while learning. Tracking the
generation of self-explanation inferences, in particular, offers a
unique method for determining whether diagrams promote infer-
ence generation during learning and for investigating the potential
influence of diagram representation on these processes.
Experiment 1
The primary goal of Experiment 1 was to investigate whether
diagram representation would interact with an individual’s back-
ground knowledge to predict learning outcomes. A second goal of
Experiment 1 was to establish a multimedia effect using the
current materials consistent with results from prior research (e.g.,
Mayer, 2001, 2003).
Experiment 1 consisted of three stages: (a) assessment of par-
ticipants’ existing knowledge using two techniques; students drew
and explained what they knew about the heart and circulatory
system and then completed a written pretest assessing general
(factual) knowledge of the domain. (b) Participants learned about
the domain using one of three possible types of online materials
presented as a series of Web pages: (1) text only, (2) text with
simplified diagrams, or (3) text with more detailed diagrams. (c)
Assessment of participants’ learning outcomes in four areas: Stu-
dents drew and explained what they knew about the heart and
circulatory system, completed a written posttest of general knowl-
edge (same as pretest), answered memory questions about the text,
and completed inferences questions related to the text.
Participants were 74 undergraduates from the University of Colorado at
Boulder. All were native English speakers and received partial credit in an
introductory psychology class for their participation.
Learning Materials
All participants read the simplest text about the heart and circulatory
system used by Wolfe et al. (1998); this text (excerpted from Silverstein &
Silverstein, 1983) was written at an elementary level and consisted of 1,616
words. For this research, the text was broken into 43 HTML pages
presented in Internet Explorer, typically with a 1– 4 sentences on each
page. Learners controlled the pace of the presentation by clicking an arrow
at the bottom of each screen to continue to the next page. In the diagram
conditions, 32 pages included a relevant diagram adjacent to the text. Each
diagram depicted the text information that it accompanied; diagrams were
modeled on the “How Your Heart Works (Lower Elementary): The Heart”
handout distributed by the American Heart Association (1988) Schoolsite
Program. A series of simplified and detailed (see Figure 1) iconic diagrams
were produced that depicted concepts from the experimental text. The
detailed diagrams preserved the anatomical structure of the original dia-
gram, but the simplified diagrams modified the heart’s anatomy in order to
more clearly depict the functional workings of the heart and the relation-
ships between heart components.
Learning Outcome Measures
Mental model improvement: Proportion of possible gain. This mea-
sure represents the degree to which participants in each experimental
condition were able to improve their existing mental models of the heart
and circulatory system. Five mental model categories (explained below)
were used in this experiment, and each was assigned a score from 0 to 5
corresponding to its complexity and accuracy: 0 the least accurate no
loop model and 5 the complete and accurate double loop 2 model.
Proportion of possible mental model improvement was calculated as the
difference between the student’s mental model scores pre- and postlearning
divided by the maximum possible increase in mental model score.
Categorization of mental models. Participants were asked to draw a
picture of what they knew about how the heart and circulatory system
works, explaining as they drew. Drawings were categorized according to
the mental model (Chi, 1997; Chi, de Leeuw, Chiu, & LaVancher, 1994)
of the heart and circulatory system that they depicted. Drawings were
completed before and after learning; premodels refer to students’ mental
models diagnosed before learning, and postmodels refer to students’ mental
models diagnosed after learning. Two raters compared students’ mental
model drawings and verbal explanations with the list of necessary features
for each mental model developed by Chi et al. (1994). Initial rater agree-
ment calculated by weighted Kappa (
was very good, but disagreements were resolved through discussion until
100% agreement was reached.
The models identified by Chi et al. (1994), from least to most advanced,
are as follows: (a) no loop, (b) ebb and flow, (c) single loop, (d) single loop
with lungs, (e) double loop 1, and (f) double loop 2. Mental models are
briefly described here (see Chi et al., 1994, for a detailed discussion) and
are illustrated by student drawings (see Figure 2).
The critical feature of the no loop model is that blood flows from the
heart to the body but does not return to the heart. The participant who drew
the no loop model in Figure 2 described blood as “gushing” out of the heart
to the body and decided that used blood does not return to the heart.
The ebb and flow model (see Figure 2) reflects knowledge of blood
vessels as well as the fact that blood returns to the heart from the body.
However, the crucial feature of the ebb and flow model is that blood travels
to and from the heart via the same blood vessel(s).
The single loop and the single loop with lungs (see Figure 2) models
share a key feature: Blood makes one and only one continuous loop with
the heart and the body. For the single loop with lungs model, the contin-
uous loop includes a path through the lungs to oxygenate the blood before
or after traveling through the rest of the body.
The double loop 1 model (see Figure 2) correctly reflects two loops of
blood flowing from the heart— one loop flowing to and from the body and
another loop flowing to and from the lungs— but the path of blood flow in
this model is erroneous. For example, the double loop 1 model in Figure 2
depicts blood returning from the body to the bottom left of the heart
(instead of the top right) and blood flowing across the heart in two
directions (instead of from top to bottom).
Finally, the double loop 2 model (see Figure 2) reflects a complete and
accurate understanding of the path of blood through the heart and circu-
latory system. Medical conventions dictate that left and right in diagrams
are reversed from the viewer’s perspective (as if looking at a person facing
the viewer). The double loop 2 drawing in Figure 2 depicts the accurate
path of blood: Blood exits the left ventricle (lower chamber), flows to the
body, and then returns to the right atrium (upper chamber). From the right
atrium, blood flows to the right ventricle and then to the lungs where
carbon dioxide in the blood is exchanged for oxygen. Oxygen-rich blood is
returned to the left atrium, flows to the left ventricle, and, again, flows to
the body through the aorta. Blood always flows from top to bottom in the
heart and cannot cross sides within the heart. Each of these elements was
necessary for a participant’s drawing and explanation to be categorized as
a double loop 2 mental model.
Two students produced a no loop premodel, and 2 students produced an
ebb and flow premodel; these students were not included in analyses
because their premodels were so infrequent. In addition, 3 participants
produced a double loop 2 premodel and were excluded because they could
not improve their mental models. Thus, data from 67 of the 74 participants
were analyzed.
Analysis of verbal protocols prompted the collapse of the single loop
mental model categories in the participant sample. A number of partici-
pants with single loop premodels knew that the lungs oxygenated blood but
believed that the lungs sent oxygen into the heart. Separating participants
with functionally equivalent knowledge about oxygen into separate cate-
gories did not appropriately describe their overall mental model for this
experiment. Indeed, the pattern of results for single loop and single loop
with lungs participants did not differ, and, for simplicity, these participants
were grouped into a single category called single loop premodels.
Pre- and posttest of general (factual) knowledge about the domain.
The General Knowledge Test from Wolfe et al. (1998) was used to assess
each participant’s factual knowledge of general information about the
human heart and circulatory system. As in Wolfe et al., this measure was
scored as a proportion of possible gain from the pre- to the posttest, but
only questions with a visual component (e.g., structure, location) were
analyzed in this experiment, for example, “How many chambers are there
in the heart?” Because one question (“How many continuous, closed
circuits of blood are there from the heart?”) could be answered for a total
of 3 points by giving the correct number of circuits and naming them (two
circuits: pulmonary–lungs and systemic– body), the pre- and posttest of
Figure 2. Mental model categories, as drawn by participants.
  • Article
    Full-text available
    The ability to analyze arguments is critical for higher-level reasoning, yet previous research suggests that standard university education provides only modest improvements in students’ analytical-reasoning abilities. What pedagogical approaches are most effective for cultivating these skills? We investigated the effectiveness of a 12-week undergraduate seminar in which students practiced a software-based technique for visualizing the logical structures implicit in argumentative texts. Seminar students met weekly to analyze excerpts from contemporary analytic philosophy papers, completed argument visualization problem sets, and received individualized feedback on a weekly basis. We found that seminar students improved substantially more on LSAT Logical Reasoning test forms than did control students (d = 0.71, 95% CI: [0.37, 1.04], p < 0.001), suggesting that learning how to visualize arguments in the seminar led to large generalized improvements in students’ analytical-reasoning skills. Moreover, blind scoring of final essays from seminar students and control students, drawn from a parallel lecture course, revealed large differences in favor of seminar students (d = 0.87, 95% CI: [0.26, 1.48], p = 0.005). Seminar students understood the arguments better, and their essays were more accurate and effectively structured. Taken together, these findings deepen our understanding of how visualizations support logical reasoning and provide a model for improving analytical-reasoning pedagogy.
  • Article
    Text‐picture integration is one of the most important cognitive processes when reading illustrated text. There is empirical evidence that text‐picture integration takes place when learning with pictures combined with single sentences. The present experiment investigated whether text‐picture integration also takes place when the single sentences are embedded into longer text segments and hence when materials become more complex. In a within‐subjects design, 43 participants read an illustrated story, in which the different combinations of general and specific sentences and pictures, respectively, were embedded. In line with previous findings, participants were more likely to falsely recognize specific versions of the sentences after having studied their general versions combined with specific pictures. Thus, the experiment shows that text‐picture integration also occurs when learners have to read longer text passages combined with pictures.
  • Preprint
    Teaching and learning biodiversity and classification is highly dependent upon visual information and Graphical Representations (GRs) are an essential component of the study materials discussing these topics. However, GRs are double-edged swords; if used appropriately it reduces the mental effort experienced while learning (cognitive load). Meanwhile, the illustrations loosely connected with the text or complex in its visual properties can hinder the learning significantly. The present study compared the usage of GRs in the lesson on biodiversity and classification given in the IXth standard biology textbooks followed by five different states in India namely Chhattisgarh, Karnataka, Mizoram, Rajasthan, and Telangana and its contribution to the cognitive load. Our result revealed that though these textbooks followed similar contents the properties of the GRs used were noticeably dissimilar. Divergence was also noticed in the way in which the GRs were integrated with the textual information (conceptual integration) and the degree of interaction between such conceptually integrated groups. Influence of such dissimilarity in frequency, form, function, and physical and conceptual integration of GRs on various dimensions of the mental effort required for the comprehension of these vital concepts in environmental science is discussed under the light of the cognitive load theory.
  • Article
    The study explores students’ production of interactive visualized stories with visual analytics (VA). The aim is to understand emerging interactions in classrooms of grades 7–9 students when visual storytelling methods are playing a part in producing social science content. The dual aspects of visual literacy, information retrievement paired with the creation of interactive visualized stories, are crucial. Video captures of students working in groups and of what happens on their screens are conducted. The results show that students can handle the technical aspects of a VA application, but interpretation of visualized statistics is challenging. The study suggests that VA has potential to strengthen students’ ability to handle huge amounts of data and increase the possibilities for young people to take part in society.(Keywords: digital technology, digital skills, democracy, civic knowledge, social science education, visual analytics, visual storytelling, visual literacy, digital literacy)
  • Chapter
    Learning, development, and response to instruction often involve changes in the strategies that learners use to solve problems. In this chapter, our focus is on mathematical problem solving in both children and adults. We offer a selective review of research on three classes of factors that may influence processes of strategy change in mathematical problem solving: contextual factors, individual factors, and metacognitive factors. Contextual factors involve information that learners encounter in the learning context, such as feedback about prior strategies and examples of alternative strategies. Individual factors involve the abilities, dispositions, and knowledge that learners bring to the learning context. Metacognitive factors involve knowledge about strategies and factors that affect the application of strategies—including perceptions of problem difficulty, confidence in the strategies one already knows, and judgments about the qualities of alternative strategies. These factors operate both independently and in combination to influence learners’ behavior. Therefore, we argue that scientific progress in understanding strategy change will require comprehensive conceptual models that specify how different factors come together to explain behavior. We discuss several such models, including vulnerability–trigger models, cumulative risk models, and dynamic systems models. Research guided by such models will contribute to greater progress in understanding processes of strategy use and strategy change.
  • Article
    This paper discusses teachers' interpretations of physics diagrammes. The study based on 55 science and non-science teachers, where a qualitative approach was adopted. First, 12 fundamental physics diagrammes were revealed to the teachers, who were asked to think aloud about them. Science teachers and non-science teachers gave similar answers. It has been observed that only the science teacher cohort extended their explanations. When interpreting the diagrammes, the participants in both groups did not notice certain elements that it was expected they would see. The result of this study can inform how teachers interpret physics diagrammes. This paper contributes to the growing interest in international literature as well as national literature regarding the use of diagrammes for teacher training, because interpreting diagrammes is a comprehensive process, which contains certain elements, such as lines, arrows, curves, colour, and objects with boundaries.
  • Technical Report
    Full-text available
    This Project Report includes published articles and a book chapter that describe the evolution and outcomes of the MyST project. The articles describe the process used to develop spoken dialog sessions, the results of the summative evaluations, and analyses of the performance of the spoken dialog system, and feedback from teachers and students. The Project Report also includes some of the proposals, reviews, and final reports submitted to the granting agencies.
  • Article
    This paper investigates whether exposure to explanatory diagrams can affect a major financial decision. In a controlled experiment, participants were given Pension Benefit Statements with or without one or two diagrams, before answering incentivised questions that measured recall, comprehension and choice of contribution rate. The diagrams had at best a marginal influence on recall or comprehension. Nevertheless, a diagram relating contributions to income projections prompted more participants to advocate higher contributions, while both diagrams influenced the rationale participants gave for decisions. The implication is that although pension products remain hard to understand, diagrams may alter decisions by reinforcing relevant causal thinking.
  • Article
    Full-text available
    Scientific illustrations play an important role in scientific texts, however, young readers show limited ability to use illustration information and integrate it with the text in multimedia learning. The primary goal of the present study was to investigate if strategy instructions for illustrated text reading focused on scientific illustration reading and text-illustration integration can help young readers overcome their deficiencies and change their reading processes, learning outcomes, and subjective perceptions of article difficulty and enjoyment, illustration enjoyment, and self-evaluation of learning. Besides, is subjective perception of the article related to reading behavior? Sixty-two fourth-grade students read an illustrated science text while their eye movements were recorded, and then completed a reading test and questionnaire. The instruction group outperformed the control group on the reading test, but subjective perceptions of the article did not differ between groups. Eye-movements analysis showed that the instruction group spent twice as much time reading the illustrations and made more saccades between relevant text and illustration sections than the control group. These findings indicate that strategy instructions for reading illustrated text promoted reading comprehension and changed learning processes, not subjective perceptions. In addition, the result of this study showed that there was no relationship between subjective perception (article difficulty or illustrations enjoyment) of the article and reading time (total reading time on the illustrated science text and on the science illustrations). This study had empirical and practical contributions. Empirically, this study developed the instruction methods in multimedia learning and further examined their effect on learning processes in young readers. Practically, this study can help elementary school teachers understand the processes used by young readers when reading illustrated texts and provide them with evidence-based instructions to teach science reading effectively.
  • Article
    Background: Health professionals in Australia and New Zealand have used various intrapartum fetal surveillance (IFS) guidelines, with clear differences in how these guidelines present information. Based on clinician feedback, the 2015 Queensland Clinical Guideline on IFS structured the prose-based Royal Australian and New Zealand College of Obstetricians and Gynaecologists (RANZCOG) IFS Guidelines as a traffic-light matrix and represented the categorical terms of unlikely, maybe, possible and likely fetal compromise, as the colours GREEN, BLUE, AMBER, and RED, respectively. Aims: To determine whether the interpretation of the RANZCOG IFS Guidelines in Table Format is more accurate and quicker compared to the current presentation of the RANZCOG Guideline in prose format. Materials and method: Twenty-nine clinicians, naïve to the use of the RANZCOG IFS Guidelines, interpreted ten cardiotocographs (CTGs) using one format and then the alternative format (totalling 580 CTG interpretations). Accuracy and time to decision were recorded as well as a participant questionnaire. A repeated measures analysis of variance was used to compare differences. Results: Compared to prose format, clinicians interpreted CTGs quicker using the table format (P < 0.01), especially CTGs representative of unlikely and maybe fetal compromise. There was a trend toward more accurate interpretation for table format for all clinicians, with significance among medical officers (P = 0.02). Participants responded more favourably to the table format regarding questions about ease of use, determining actions required, and desire to use the system in the future (P < 0.01). Conclusions: Presenting the RANZCOG IFS Guideline in table format as opposed to prose format improved the speed and accuracy of CTG interpretation and is preferred by clinicians.
  • Chapter
    This chapter discusses about presentation of information by media and its effect on mental models. A predominant function of media consists in the presenting of information associated with the intention to produce knowledge of the world. Learning refers to every alteration of knowledge and can be interpreted as a consequence of changing or transforming the state of existing knowledge. An intelligent system (IS) represents knowledge as ideas or thoughts by means of symbols and codes. Cognitive processes operate on these symbols and codes with the intention of modifying or transforming the knowledge. The commonly used systems of communications are language and pictures. At the mental level they correspond to concepts and images, whereby phrases of natural languages and concepts are discrete units while pictures and images are analogous units of presentation and representation, respectively. The most essential constituents of the knowledge system are abstract data structures called “knowledge bases.” These knowledge bases store and conserve information on objective reality.
  • Article
    In Experiment 1, inexperienced trade apprentices were presented with one of four alternative instructional designs: a diagram with visual text, a diagram with auditory text, a diagram with both visual and auditory text, or the diagram only. An auditory presentation of text proved superior to a visual-only presentation but not when the text was presented in both auditory and visual forms. The diagram-only format was the least intelligible to inexperienced learners. When participants became more experienced in the domain after two specifically designed training sessions, the advantage of a visual diagram-auditory text format disappeared. In Experiment 2, the diagram-only group was compared with the audio-text group after an additional training session. The results were the reverse of those of Experiment 1: The diagram-only group outperformed the audio–text group. Suggestions are made for multimedia instruction that takes learner experience into consideration. (PsycINFO Database Record (c) 2012 APA, all rights reserved)
  • Article
    In three experiments, students read expository passages concerning how scientific devices work, which contained either no illustrations (control), static illustrations of the device with labels for each part (parts), static illustrations of the device with labels for each major action (steps), or dynamic illustrations showing the "off" and "on" states of the device along with labels for each part and each major action (parts-and-steps). Results indicated that the parts-and-steps (but not the other) illustrations consistently improved performance on recall of conceptual (but not nonconceptiual) information and creative problem solving (but not verbatim retention), and these results were obtained mainly for the low prior-knowledge (rather than the high prior-knowledge) students. The cognitive conditions for effective illustrations in scientific text include appropriate text, tests, illustrations, and learners.
  • Article
    In 2 experiments, students studied an animation depicting the operation of a bicycle tire pump or an automobile braking system, along with concurrent oral narration of the steps in the process (concurrent group), successive presentation of animation and narration (by 4 different methods), animation alone, narration alone, or no instruction (control group). On retention tests, the control group performed more poorly than each of the other groups, which did not differ from one another. On problem-solving tests, the concurrent group performed better than each of the other groups, which did not differ from one another. These results are consistent with a dual-coding model in which retention requires the construction of representational connections and problem solving requires the construction of representational and referential connections. An instructional implication is that pictures and words are most effective when they occur contiguously in time or space.
  • Article
    In 2 experiments, students who lacked prior knowledge about car mechanics read a passage about vehicle braking systems that either contained labeled illustrations of the systems, illustrations without labels, labels without illustrations, or no labeled illustrations. Students who received passages that contained labeled illustrations of braking systems recalled more explanative than nonexplanative information as compared to control groups, and performed better on problem solving transfer but not on verbatim recognition as compared to control groups. Results support a model of meaningful learning in which illustrations can help readers to focus their attention on explanative information in text and to reorganize the information into useful mental models.
  • Article
    This experiment was designed to determine: a.) whether students’ level of entering behavior affects their achievement on criterial tests designed to measure their achievement of different educational objectives; b.) whether identical visuals are equally effective in facilitating the achievement of students possessing different levels of entering behavior; and c.) whether all types of visuals are equally effective in facilitating student achievement of identical educational objectives. Five hundred eighty-seven students at The Pennsylvania State University were randomly assigned to one of the nine treatment groups. Each student received a pretest, participated in his respective programmed presentation, and received four individual criterial measures. The three levels of entering behavior—high, medium, and low—were defined by establishing cut-off points one-half standard deviation on each side of the mean which was achieved on the physiology pretest by students in each of the nine instructional treatments. The results of this study indicate that a.) students with high entering behavior consistently achieved equivalent or significantly higher scores on the criterial measures than students with low and medium entering behavior regardless of the type of instructional presentation they received; b.) the use of visualization to complement programmed instruction is an effective instructional technique for reducing differences in achievement on the criterial measures between students with low and medium entering behavior; and c.) the black and white versions of the simple line and detailed, shaded drawing presentations were found to be most effective in reducing differences in achievement among students with high, medium, and low entering behavior.
  • Article
    Eight story problems in a drawn format, eight matching problems in a verbal format, and eight matching problems in a telegraphic format were administered to 854 students in tests at each grade from 3 to 7. Scoring was based on the choice of correct operations to solve the problem. Readers of high ability, as measured by a reading test, chose correct operations more often than low-ability readers. The drawn format was easier than the other two formats. A significant format-by-reading-ability interaction revealed that the advantage of the drawn format was greater for low readers than for high readers.