Contemporary Educational Psychology 38 (2013) 211–224

Contents lists available at SciVerse ScienceDirect

Contemporary Educational Psychology

journal homepage: www.elsevier.com/locate/cedpsych

Self-generated drawings for supporting comprehension of a complex animation

Lucia Mason a,⇑, Richard Lowe b, Maria Caterina Tornatora a
a
Department of Developmental Psychology and Socialization, University of Padova, Via Venezia 8, 35131 Padova, Italy
b
School of Education, Curtin University, Kent Street, Bentley, WA 6102, Australia

a r t i c l e i n f o a b s t r a c t

Article history: The perceptual and cognitive processing demands involved in comprehending complex animations can
Available online 13 April 2013 pose considerable challenges to learners. There is a tendency for learners to extract information that is
highly perceptually salient but neglect less conspicuous information of crucial relevance to the building
Keywords: of a quality mental model. This study investigated the effectiveness of self-generated drawing for learn-
Animation ing from an animation illustrating a scientific phenomenon, the so-called ‘‘Newton’s Cradle.’’ Participants
Self-generated drawing were 199 students in grade seven, randomly assigned to three experimental conditions: self-generated
Drawing-to-learn
drawing, traced/copied drawing, and no drawing. All participants were asked to produce an explanation
Comprehension of animations
Learning from animations
of the animation for both immediate and delayed posttests. The results revealed the superiority of self-
generated drawing in supporting animation comprehension at both testing times compared to the other
two conditions, which did not differ from each other. In addition, comprehension of the animation was
related to the quality of self-generated drawings. Specifically, the depiction of information characterized
by low perceptual salience but high conceptual relevance to the phenomenon predicted comprehension
and retention over time.
Ó 2013 Elsevier Inc. All rights reserved.

1. Introduction
Educational animations have become an increasingly common
feature of learning environments that are supported by technology.
Multimedia learning resources and educational websites make
extensive use of such dynamic graphics with the aim of supporting
students’ comprehension of complex materials, especially those
that deal with scientific phenomena and systems (Ainsworth,
2006). Animations may serve various purposes (Hegarty, Kriz, &
Cate, 2003). They can depict not only visible phenomena but also
those that are invisible, such as changes in pressure or temperature
as shown in an animated weather map (Lowe, 2004). Animations
can also represent more abstract types of content, such as statisti-
cal concepts (Bodemer, Ploetzner, Feuerlein, & Spada, 2004), prob-
lem solving procedures (Wouters, Paas, & van Merriënboer, 2009),
computer algorithms (Narayanan & Hegarty, 2002), or prey and
predator relationships (Ainsworth & Van Labeke, 2004).
Although many animations present their referent subject mat-
ter realistically, others deliberately distort reality. They do this
by approaches such as speeding up some processes and slowing
down others, portraying an object or phenomenon from different
or changing perspectives, augmenting the display to cue viewers’
attention to more relevant parts, or making moving objects leave
⇑ Corresponding author.
E-mail addresses: lucia.mason@unipd.it (L. Mason), R.K.Lowe@curtin.edu.au (R.
Lowe), caterina.tornatora@unipd.it (M.C. Tornatora).
tracks. These very different mappings between referents and dy-
namic visualizations make it difficult to generalize findings from
one type of animation to another (Hegarty, 2004).
Some evidence in the literature indicates that dynamic graphics
may have advantages over static displays, particularly with respect
to student motivation and implicit learning (Rieber, 1991; Rieber,
Tzeng, & Tribble, 2004) or the learning of procedures (Höffler &
Leutner, 2007). However, there is mounting contrary evidence
indicating that the effects of animation on learning are not uni-
formly positive. For example, the review by Tversky, Morrison,
and Bétrancourt (2002) identifies cases where there are no benefits
for animation over static graphics. Further, studies reporting an
advantage for animation may be confounded by the presence of
accompanying verbal information. A compelling case therefore ex-
ists to investigate learning from animations alone in order to pro-
vide a scientific account of how they are comprehended per se,
without help from verbal accompaniments (Boucheix & Lowe,
2010; Hegarty, 2004; Lowe, 2003, 2004).
Despite equivocal research evidence about animation’s effec-
tiveness as a tool for learning (Bétrancourt, 2005), more and more
reliance is being placed on animated graphics. Given that this take-
up of animations seems unstoppable, the question then arises as to
what approaches could be adopted to improve learning from these
representations? Recent research identifies the importance of
using strategies that encourage learners to process the information
provided in animations more deeply (Kombartzky, Ploetzner,
Schlag, & Metz, 2010; Rebetez, Bétrancourt, Sangin, & Dillenbourg,

0361-476X/$ - see front matter Ó 2013 Elsevier Inc. All rights reserved.
http://dx.doi.org/10.1016/j.cedpsych.2013.04.001
212 L. Mason et al. / Contemporary Educational Psychology 38 (2013) 211–224
2010). Such approaches typically use verbal instructions that direct
students to carry out a series of steps as they study an animation.
However, another more visually-oriented strategy that has come
into prominence of late with respect to informational texts is
‘‘drawing-to-learn’’ in which students generate their own graphic
representations to aid their developing understandings. This ap-
proach has been seen as having particular relevance to science
education, with Ainsworth, Prain, and Tytler (2011) suggesting that
learners’ drawing should be valued (along with writing, reading,
and talking) as a crucial component of science learning.
Embedded in the theoretical framework of the Animation Pro-
cessing Model (APM; Lowe & Boucheix, 2008, 2011), the present
study examines the potential of self-generated drawing to invoke
deeper processing of animations by requiring learners to engage
in constructive, transformative manipulation of the presented con-
tent in order to support their extraction of relevant information,
irrespective of its salience. According to the APM, perceptual as-
pects of processing play a key role in learning with animations so
that careful observation of the depiction is absolutely central to
success. Requiring learners to generate explanatory drawings of
the events depicted in an animation should enhance their extrac-
tion of information in general and less perceptually salient infor-
mation in particular. However, comprehension requires that
effective information extraction and internalization is accompa-
nied by appropriate cognitive processing that facilitates the con-
struction of a high quality mental model.
The study reported in this paper extends recent research on the
effects of self-generated drawing on learning from texts by inves-
tigating its effects on the comprehension of a science animation.
The next sections review issues relevant to the use of drawing as
a learning strategy and to the comprehension of animations.
1.1. Drawing as a learning strategy
Learner-generated drawing is the ‘‘construction of an external
visual representation, or picture, of to-be-learned content’’ (Van
Meter, Aleksic, Schwartz, & Garner, 2006, p. 143). This definition
has been developed to include the requirement that learners
should ‘‘maintain responsibility for the final appearance of draw-
ings and the constraint that final drawings are representational’’
(p. 143), that is, ‘‘drawings are intended to show how depicted ob-
jects actually look’’ (p. 143). Research indicates that when learners
create their own visualizations (instead of only interpreting others’
external representations), scientific learning is facilitated (Ains-
worth et al., 2011). The effectiveness of self-generated drawings
in supporting learning from science texts has been attributed to
the benefits that come from externalizing one’s own cognition
(Cox, 1999). Van Meter (2001) compared the effectiveness of three
approaches for supporting fifth and sixth grade students’ learning
from a text about the nervous system: drawing instructions only,
drawing instructions with additional supports, and illustrations
with no drawing instruction. Students who produced their own
drawings recalled more from the text than those who saw illustra-
tions but did no drawing.
Beneficial effects of self-generated drawing were also found
with regard to problem-solving performance, particularly if the
drawing activity was supported (van Meter et al., 2006). In a study
by Schwamborn, Mayer, Thillmann, Leopold, and Leutner (2010) of
ninth graders learning the chemistry of washing, students who re-
ceived drawing instructions with prompts performed better than a
control group with neither drawing instruction nor prompts.
A series of small-scale studies by Ainsworth (2010) investigated
how students could use drawing more effectively when learning
from texts about the cardio–vascular system. Participants were
asked to self-explain the text content before drawing, or to draw
for different audiences (i.e., for themselves or a peer). The results
indicated that learners’ drawing activity can be enhanced by strat-
egies that stimulate their active processing of the learning material
before they represent their new understandings in drawings.
Two recent studies by Leopold and Leutner (2012) compared
the effectiveness of self-generated drawing not only with a control
group, but also with groups using verbal strategies that required
learners to invest effort but via a non-visualization activity (Exper-
iment 1: selection of main ideas; Experiment 2: summarizing). Par-
ticipants were tenth graders learning about water molecules from
a chemistry text. Both studies revealed a clear superiority of draw-
ing over verbal activity. In addition, the cognitive advantages of
drawing were found for transfer tests but not for multiple-choice
tests of factual knowledge. The finding that self-generated drawing
is particularly beneficial for deeper learning is consistent with the
outcomes of previous studies (Van Meter et al., 2006).
In all the above studies, self-generated drawing was examined
with respect to learning from text (albeit, sometimes illustrated
text). To our knowledge, only two studies have been carried out
to investigate the effects of drawing on learning from a dynamic
visualization. The content for both of these studies was the chem-
ical reactions that occur in hydrogen fuel cell cars. In the first,
Zhang and Linn (2011) compared drawing with giving students
more time to explore the visualization. Eighth graders who gener-
ated their own drawings to interpret the animation integrated
more ideas than peers who explored the learning material for a
longer period. In the second study, Zhang (2010) compared the
effectiveness of having high school students generate drawings
with having them produce a critique of pre-existing drawings.
Those who generated drawings were better able to distinguish
amongst different ideas and to acknowledge the flaws in their rep-
resentations than students who produced the critique.
Why is drawing-to-learn effective? Van Meter et al. (2006) de-
scribe learner-generated drawing in broad terms as an ‘‘elabora-
tive, strategic’’ activity that ‘‘should lead to the construction of a
mental model’’ (p. 144). More specifically, the Generative Theory
of Drawing Construction (GTDC) proposed by Van Meter and Gar-
ner (2005) concerns the role of drawing in learning from text
accompanied by illustrations. It identifies a number of important
processes that could be facilitated if learners generate drawings
while studying such materials. We will suggest that with one
exception, the processes identified in the GTDC should also be
applicable to learning from animation as investigated in the pres-
ent research. That exception is integration because the animation
involved is not accompanied by any form of text so learners are
not required to integrate information from two different types of
representation. As will be discussed below, the remaining pro-
cesses fostered during drawing of (i) selecting key elements from
a provided external representation, (ii) organizing those key ele-
ments into groups, and (iii) constructing an internal representation
appear likely to benefit learning from animation.
The beneficial effects of drawing on learning from illustrated
text are attributed to its role as a recursive activity in which there
are mutual influences amongst the various internal and external
representations available to the learner (Van Meter et al., 2006).
The recursions involved support learner processes such as self-
monitoring that can expose misunderstandings and redirect atten-
tion to hitherto neglected key aspects of external representations.
Although this account targets illustrated text, in principle it ap-
pears possible that the learner’s repeated cycling back and forth
between an animation and an evolving drawing could serve similar
functions. However, this would require the activities supported by
drawing (i.e., selection, organization, and construction) to be of va-
lue not only to learning from illustrated text, but also to learning
from animations. In the next section we explore this possibility
by considering some challenges to learning from animation, the
likely origin of these challenges, and the extent to which the
 L. Mason et al. / Contemporary Educational Psychology 38 (2013) 211–224
processes identified above as benefitting from self-generated
drawing might apply to animation rather than illustrated text.
1.2. Comprehending animations
Although instructional designers and educational practitioners
assume graphics to be more educationally beneficial if they are dy-
namic rather than static, this assumption has been disputed by
researchers (Bétrancourt & Tversky, 2000). For example, a study
by Hegarty et al. (2003) showed that students did not learn more
from an animation of a flushing cistern action than from three sta-
tic diagrams.
Why are animations sometimes not particularly effective? To ad-
dress this question, it is necessary to consider how animations are
comprehended in their own right (i.e., without any accompanying
explanatory text). Although the term ‘‘comprehension’’ is mostly
equated with text-comprehension, it is important to keep in mind
that this term does not refer exclusively to the understanding of
written or spoken text (see Schnotz & Kulhavy, 1994). In this paper,
we adopt a more inclusive view – that comprehension applies to rep-
resentations in general, not solely to those that are text-based. At
this very broad level, comprehension can be regarded as the forma-
tion of an effective internal representation of the referent subject
matter from an externally represented information set. Such inter-
nal representations are described as mental models (e.g., Betran-
court & Chassot, 2008; Schnotz, 2005).
A learner who comprehends an animation has used the infor-
mation it presents to construct a high quality mental model (i.e.,
appropriate, accurate, and comprehensive) of the depicted content.
Conversely, a lack of comprehension or misunderstandings of the
content occur when the constructed mental model is deficient.
Unfortunately, the very characteristic of animations that gives
them potential benefits compared with static graphics (i.e., their
capacity to depict change over time directly rather than indirectly)
can also have associated processing costs for learners that may
compromise mental model quality.
The dynamic character of an animation can impose very chal-
lenging perceptual and cognitive processing demands on learners,
particularly if the depicted subject matter is complex or unfamiliar
(Ayres & Paas, 2007; Lowe, 1999; Schnotz & Lowe, 2008; Sweller,
2005). Narayanan and Hegarty (2002) identify three main sources
of problems in the processing of an animation in relation to how
spatial and temporal changes are depicted: (i) if the changes are
represented as concurrent, they may conflict with mental anima-
tions that are performed serially; (ii) if the depicted changes are
too complex so that too many components vary, processing may
be impeded; (iii) if the changes are illustrated at a speed that is
incompatible with the viewer’s perceptual and cognitive process-
ing capacity, relevant information may be neglected. Information
that is of central relevance to the theme of an animation may also
be neglected because it is insufficiently conspicuous. The distinc-
tion between salience and relevance, which is well established in
the literature on learning from animation (e.g., Schnotz & Lowe,
2008), is particularly pertinent to our study. Salience refers to
how noticeable an item of information is due to its perceptual fea-
tures, whereas relevance is about how important the information
is with respect to the content (Hegarty, Canhan, & Fabrikant, 2010).
With respect to making meteorological predictions from weath-
er maps, Lowe (1999, 2004) found that learners were more likely to
extract perceptually salient information while neglecting more
visually subtle but conceptually relevant aspects. In other words,
their attention was preferentially directed to what was more visu-
ally conspicuous to the detriment of what was most important for
learning. As a consequence of the failure to extract these key as-
pects, the quality of the mental model constructed for the referent
from the external representation was compromised. This biasing of
213
information extraction towards perceptually salient aspects of a
display has also been found for animations of other types of con-
tent such as a piano mechanism (Lowe & Boucheix, 2011).
If a fundamental problem in learning from animation is inade-
quate processing of the presented information, it seems reasonable
that activities for increasing the general depth of processing could
be of assistance (cf. Kombartzky et al., 2010; Rebetez et al., 2010).
Further, because the visually-oriented approach involved with
learners generating their own drawings is consistent with the vi-
sual nature of animations, a drawing-to-learn strategy offers a pos-
sible way to support learner extraction of key information that
they may otherwise miss. However, in order to be successful, the
selection, organization, and construction processes that the GTDC
(Van Meter & Garner, 2005) posits are fostered by such a drawing
strategy would need to address the specific types of processing is-
sues that arise during learning from animations. In the following
analysis, we show that there is considerable correspondence be-
tween processes facilitated by self-generated drawing and those
required to build a mental model from an animation.
The Animation Processing Model (Lowe & Boucheix, 2008,
2011) characterizes learning from animation as an iterative, cumu-
lative activity involving five broad phases in which there is interac-
tion between various bottom-up and top-down processes (Fig. 1).
In Phase 1 processing, the continuous flux of the dynamic infor-
mation comprising the animation is parsed into event units (enti-
ties plus their associated behaviors). To be beneficial for ultimate
mental model building activity, this initial processing requires
learners to select out these event units from the animation during
the time course of its presentation. Accordingly, the fostering of
selection that the GTCD posits to occur during self-generated
drawing should enhance Phase 1 animation processing. In Phase
2 processing, the event units that have been previously selected
as a result of parsing are combined at a local level (e.g., event units
that are spatially and temporally adjacent) to form broader struc-
tures termed dynamic micro-chunks. This is essentially an early
form of organizing multiple event units to construct higher order
components from which the mental model will ultimately be com-
posed. Because the GTDC posits that organization and construction
are fostered during self-generated drawing, this activity should
also enhance Phase 2 animation processing. In Phase 3 processing,
the previously formed dynamic micro-chunks are interconnected
by relational bridges that allow super-ordinate organizing struc-
tures such as domain general causal chains to be constructed. These
higher-level Phase 3 processes should also benefit from the organi-
zation and construction fostered by self-generated drawing. Phases
4 and 5 of the APM are more top-down aspects of processing con-
cerned especially with domain-specific background knowledge.
As stated above, research has revealed that a significant prob-
lem in learning from animation is that students may extract infor-
mation that is perceptually salient and yet neglect more subtle but
nevertheless conceptually relevant information. The resulting fail-
ure to internalize key aspects of the content compromises their
comprehension. Self-generation of drawings has the potential to
ameliorate this problem by providing opportunities for the learner
to process the animation more deeply overall through transform-
ing and manipulating the presented information. However, for
self-generated drawing to be effective in improving learning from
animation, it is crucial that learners go beyond the superficial as-
pects of the animation (those that are most readily perceptible).
If they are to build a coherent and stable high quality mental mod-
el, learners must extract not only high relevance information that
is highly salient but also high relevance information with a low le-
vel of perceptual salience. Learners whose self-generated drawings
include both types of information should therefore have superior
comprehension of the content to those whose drawings are limited
to only high perceptual salience aspects.
214 L. Mason et al. / Contemporary Educational Psychology 38 (2013) 211–224
Fig. 1. Animation Processing Model. Adapted from Lowe and Boucheix (2011, p. 651). Copyright 2011 by Elsevier.
We assume that the constructive mental manipulation and
transformation of the subject matter a learner engages in during
the drawing process is crucial to comprehension. The more the
content to be learned is manipulated and transformed (with its
attendant self-monitoring and elaboration), the more likely it is
to be understood and remembered. The effectiveness of activities
that require learners to constructively transform content has also
been empirically demonstrated in the writing-to-learn approach
in science (Gunel, Hand, & McDermott, 2009; Klein, 2000; Mason,
2001).
We emphasize that it is not the physical activity of drawing per
se – the mere act of pushing a pencil or pen around to mark a piece
of paper – that is likely to support comprehension. Simply tracing
over given static frames and reproducing them would be active but
not constructive (Chi, 2009). Rather, it is the intense inspection of
the animation and deep level of generative cognitive processing re-
quired to create original drawings that would foster high quality
mental model construction.
2. Research questions and hypotheses
To extend current research on learning from animations, this
investigation was set in the context of both research on self-gener-
ated drawing for learning in science and research on learning from
animated illustrations. Our interest was in whether such drawing
activity would foster key perceptual and cognitive processes upon
which animation comprehension relies. More specifically, would it
foster comprehension by supporting the extraction and internali-
zation of not only the most perceptually salient information, but
also less salient information with high thematic relevance? An
underlying assumption was that self-generated drawing could
benefit learning from animation by stimulating extensive con-
structive transformational activity during the creation of graphical
representation depicting the events portrayed in the animation.
The following research questions guided the study:
1. Does self-generated drawing result in higher immediate and
delayed posttest scores for comprehension of a physics phe-
nomenon animation (Newton’s Cradle) than merely reproduc-
ing provided drawings or doing no drawing?
2. Are immediate and delayed animation comprehension scores
related to the quality of self-generated drawing produced (after
controlling for science achievement and prior knowledge)?
3. Which individual aspects of the self-generated depiction pre-
dict both immediate and delayed animation comprehension?
For the first research question, we hypothesized that at both
testing times comprehension by participants in the self-generated
drawing condition would be superior to comprehension from those
in the other two conditions, as revealed by their explanations of
the animated phenomenon. Requiring students to create original
graphic representations should stimulate deeper processing of
the presented material. In particular, the careful observation and
elaboration processes involved in locating relevant information,
choosing which items to draw, deciding how those items should
be arranged, devising ways of showing what happens, etc. would
help students develop and retain a better mental representation
of the animation. In contrast, just tracing over provided drawings
in order to reproduce them would not require this beneficial inter-
rogation and mental manipulation of the learning content.
Some clarification is needed here. Comprehension of the anima-
tion was assessed by the quality of participants’ explanation. Be-
cause our participants were seventh graders, it was not realistic
to expect them to produce an advanced and abstract formal phys-
ics explanation of the depicted phenomenon. Rather, our focus in
this study was on extraction and internalization of crucial informa-
tion that could be considered the precursor to developing a more
sophisticated understanding. Thus we did not expect a full appre-
ciation of the underlying principles of physics but only the acqui-
sition of conceptually relevant information upon which deeper
comprehension could subsequently be built.
For the second research question, we hypothesized that the
overall quality of the self-generated drawings would predict the
learners’ explanation performance. The rationale here was that
higher quality drawings would be produced by those who engaged
in deeper processing of the stimulus materials and therefore would
produce superior mental representations of the content.
For the third research question, we hypothesized that compre-
hension at immediate posttest would be related to depiction in
self-generated drawings not only of perceptually salient informa-
tion but also of information with high conceptual relevance, irre-
spective of its salience. Further, the depiction of low salience,
high relevance information would be particularly important for
the on-going comprehension required by the delayed posttest. This
is because participants who successfully extracted such high rele-
vance information would be able to build a more coherent and
 L. Mason et al. / Contemporary Educational Psychology 38 (2013) 211–224
stable mental model of the depicted phenomenon than those who
missed these aspects.
3. Method
3.1. Participants
Two hundred and four 7th grade students attending public low-
er secondary schools in north eastern Italy initially took part on a
voluntary basis with parental consent. Because five participants
were absent for one of the posttests, data reported here concern
199 students (94 girls and 105 boys) with a mean age of
12.07 years (SD = .50). All were Caucasian native speakers of Italian
and shared a homogeneous middle class background. From com-
munication with their teachers, we knew that none of them had
either learning disabilities, or particular learning difficulties. There
were no participants repeating a course at school and their overall
achievement was average or above. Participants were randomly as-
signed to three conditions: no-drawing (n = 68), traced/copied
drawing (n = 64), and self-generated drawing (n = 67). All were in-
volved in the pretest, immediate, and delayed posttest research
design.
3.2. Materials
Participants viewed a realistic animation depicting the behavior
of pendulums in a five-ball Newton’s Cradle. This device consists of
a set of balls that are suspended from cords and free to swing to-
and-fro. The animation was presented alone, without written or
spoken accompaniment. Rather than offering an idealized por-
trayal of Newton’s Cradle behavior in which both total energy
and momentum are conserved, this animation showed the real-
world situation in which degradation occurs. Due to the gradual
redistribution and dispersion of energy, movement within the sys-
tem changes over time, decays, then eventually stops.
In the resting position, all five balls of the Newton’s Cradle are
lined up in a row and in contact. The process depicted in the ani-
mation begins when an end ball (ball 1) is raised and then released
so that it swings down like a pendulum and hits the next ball. This
collision causes energy and momentum to be transmitted through
the subsequent line of balls so that the final ball (ball 5) is pro-
pelled from the opposite end of the row. Once ball 5 reaches its
peak, it then swings back and the process is reversed. These alter-
nating swings are repeated over and over again but with decreas-
ing amplitude of the two end balls (1 and 5) over time due to
energy transfers. As the amplitude of the end balls decreases, the
initially static middle balls gain energy and their movement pro-
gressively increases. This redistribution of energy eventually re-
sults in all five balls swinging in unison, with their amplitude
then gradually decreasing until they finally all come to rest. The
movements involved in the conceptually crucial aspect of energy
redistribution via the middle balls are much less readily percepti-
ble than the initial swings of the end balls or the subsequent joint
swing of all five balls. This description indicates that the operation
of a real-world Newton’s Cradle is a complex scientific phenome-
non. Comprehension of its behavior requires learners to mentally
represent a series of hierarchically interrelated phases which differ
greatly in their relative perceptibility:
Phase 1 – the balls at the end of the row swing alternately;
Phase 2 – motion is transmitted via activity of the middle balls;
Phase 3 – the amplitude of end ball swinging progressive
decreases;
Phase 4 – all balls swing jointly as a unit;
Phase 5 – all motion finally ceases.
215
There were two main reasons that the Newton’s Cradle anima-
tion was chosen for this investigation. First, the device was unfamil-
iar to the participants so that they came to the learning task with
little or no relevant domain-specific background knowledge. This al-
lowed us to study how self-generated drawing might influence
learners’ capacity to construct an appropriate mental model from
a demanding animated depiction. Our particular interest here was
in how well this activity helped them represent the relationships be-
tween the components of the observed system. Although the partic-
ular topic of Newton’s Cradle was unfamiliar to the class, it was
generally consistent with the science curriculum for the grade. Sec-
ond, although Newton’s Cradle is a dynamically complex phenome-
non, it is nevertheless easy to represent graphically because it is
visually simple. The essential information can be portrayed using
only circles and lines. This allows the influence of differences in gra-
phic ability on drawing production to be minimized.
3.3. Conditions
Three treatment conditions were used in this study. In the self-
generated drawing condition, participants provided with blank pa-
per, pencils, erasers, and crayons were asked to produce a set of six
drawings to represent what they had observed in the Newton’s
Cradle animation (with each drawing to be done on a separate
sheet). We asked for six drawings to allow for depiction of the ini-
tial state of the device plus the five phases described above. Partic-
ipants were told they were free to use whatever style of depiction
they wished and not to worry about their drawing ability.
In the traced/copied drawing condition, participants were pro-
vided with the same drawing materials as above but asked to trace
over ‘join-the-dots’ depictions of six given states extracted from
the animation provided on separate cards and then to copy them.
The cards were presented in a random sequence not to advantage
the participants in this condition with additional reinforcing infor-
mation about the sequence that was not available to the partici-
pants in the other conditions. The materials produced by learners
in the traced/copied condition were reviewed to ensure the fidelity
of treatment, an essential requirement for the validity of the find-
ings regarding this condition. All learners traced over the dot line
drawings and copied them. Appendix A introduces the six cards gi-
ven to the participants in this condition.
The trace/copy and self-generation treatments allowed us to
compare (i) drawing as a mere physical reproduction activity with
(ii) drawing as a mental construction activity. They also allowed
control for the time on task spent by participants in the two draw-
ing conditions. The no-drawing condition was included to provide
a baseline for comparison.
3.4. Measures
3.4.1. Self-generated drawings
Drawings self-generated by the learners were used as measures
of the quality of the graphical activity undertaken in the service of
comprehension.
3.4.2. Immediate comprehension
Participants’ responses to the following task were used to mea-
sure participants’ immediate comprehension of the animation:
‘‘Please write an explanation of what happens in the animation
for a schoolmate who has not viewed it. Try to be as clear and com-
plete as possible in your explanation.’’ They were told not to worry
about mistakes in spelling or grammar. As mentioned above, our
focus was on how well the 7th graders extracted and internalized
precursor information suitable for building further understanding
of the phenomenon, not on high level physics principles. Learners
216 L. Mason et al. / Contemporary Educational Psychology 38 (2013) 211–224
who cannot acquire this fundamental information about dynamic
relations would not subsequently be able develop an understand-
ing of the Newton laws and conservation of energy, momentum
per se, in classical mechanics. Because participants were asked to
produce an explanation of what happens (rather than a descrip-
tion), the focus is upon comprehension of the animation rather
than rote recall of isolated events. Participants were required to
represent mentally the dynamic relations that are at the basis of
the essential mechanism of conservation of energy. Such compre-
hension is a prerequisite for a deep understanding of Newton laws
and important physics principles.
3.4.3. Delayed comprehension
After a 2 month break, participants again produced written
explanations of what had happened in the animation viewed on
the earlier occasion.
We also considered topic prior knowledge and achievement in
science to be two basic variables that could influence comprehen-
sion of animation.
3.4.4. Topic prior knowledge
To assess participants’ prior knowledge of the phenomenon de-
picted in the animation, they were shown a static representation of
the initial state of the animation with ball 1 raised just before being
released (Fig. 2). They were asked to predict as clearly and com-
pletely as possible what would happen after release of the ball.
3.4.5. Achievement in science
Participants’ overall achievement in science was measured
using their grade in science (as at the end of the previous term).
In the Italian school system grades range from 1 to 10 (highest
grade = 10).
3.5. Coding
3.5.1. Topic prior knowledge
Participants predicted what would happen after release of the
ball raised in the static representation of the initial state of the ani-
mation. These predictions were awarded 1 point (range 0–5) for
each event correctly anticipated.
3.5.2. Self-generated drawings
Participants’ self-generated drawings were scored according to
the five phases of the Newton’s Cradle behavior (range 0–5). One
point was awarded if a phase in a production was essentially cor-
Fig. 2. The initial phase of the phenomenon illustrated in the Newton’s Cradle.
rect. If a phase was missing, incorrectly represented or incomplete,
no point was awarded. It should be noted that the depiction of the
least perceptually salient information, motion transmission, was
awarded one point when the transitional contact between the balls
was clearly apparent or when arrows were used to represent con-
tact transmissions. Appendix B gives examples of participants’
depictions of the various animation phases. A total score for the
set of drawings was calculated for each participant. This score re-
flected the overall correctness and completeness of the graphical
representation of the animation. Examples of superior and inferior
drawing sets are presented in Appendix C.
3.5.3. Immediate and delayed comprehension
Comprehension scores were derived from the written produc-
tions by awarding one point if the explanations for each phase
were correct (even if not expressed in entirely appropriate lan-
guage; range 0–5). The following are examples of overall satisfac-
tory explanations that refer to the motion transmission through
the activity of the middle balls: ‘‘There are 5 pendulums next to
each other; when one at the end touches the pendulum that is clos-
est, the force passes through all pendulums and transfers to the
last one that swings’’ (P52); ‘‘In this instrument we see the depar-
ture of the first ball that is suspended in the air, it knocks the sec-
ond which seems to transmit the vibrations until the last, which
comes back and hits the fourth ball transmitting the vibration in
a reverse cycle’’ (P10). No points were awarded when an overall
explanation failed to capture the micro-level changes that are
characteristic of the transmission process. Unsatisfactory overall
explanations include: ‘‘The balls rock faster and then more and
more slowly until they remain immobile’’ (P30); ‘‘There is a pendu-
lum with an end ball that moves, as soon as it touches the other
balls, the ball at the opposite end starts moving leaving the others
still’’ (P3). Scores for immediate and delayed comprehension re-
flected the overall correctness and completeness of the verbal
explanations of the animation. In this regard, it should be pointed
out that the coding did not require specific scientific terms to have
been used. Rather, credit was given for meaning and not specialist
terminology. This approach extended even to the term ‘‘force’’ that
appeared in some explanations. Because this term is not exclusive
to physics (but is also used in everyday language), it was given
credit only when part of an effective explanation.
All self-generated drawings and written productions (at pre,
post, and delayed posttests) were scored by two independent judges
who were familiar with the phenomenon illustrated in the anima-
tion. Their overall agreement, as measured by Pearson’s r, ranged
from .88 to .94. Inter-rater reliability for coding of the specific phases
in the set of drawings, as measured by Cohen’s k, is reported in Ta-
ble 1. Inter-rater reliabilities for coding of the written productions
at pre, post, and delayed posttests were .96, .93, and .92, respec-
tively. All disagreements were discussed by the judges in the pres-
ence of the first author until consensus was reached.
3.6. Procedure
The experiment was conducted within the participating
schools. After random assignment to the three treatment condi-
Table 1
Cohen’s k indices and percentages of inter-rater reliability for the coding of the
various phases of the animation as depicted in the drawing set.
Phases Cohen’s k %
End balls alternate swinging 1 100
Motion transmission .75 88
Swinging decrease .90 95.6
Joint swinging .94 97.1
Final stop .96 98.5
 L. Mason et al. / Contemporary Educational Psychology 38 (2013) 211–224
tions, participants took part in two sessions. In the initial session,
their prior knowledge of the topic was first assessed as indicated
above. Next, the animation was shown with participants in all con-
ditions being allowed to watch it freely. As no time limit was set
for studying the animation, participants could observe it as much
time as they needed, which ranged from 2.5 to 5.4 min and was
substantially the same across conditions. It should be noted that
before viewing the animation, participants were told that they
could view it multiple times if needed, but they had to view it care-
fully to comprehend what happens in the dynamic visualization.
After viewing the animation, self-generation condition participants
were asked to generate six drawings to represent what they had
seen, while trace/copy-condition participants were asked to repro-
duce the six provided drawings. These participants were then
asked to write their explanation of the animation. In the no-draw-
ing condition, there was no associated drawing activity prior to
production of the written explanation. The time taken to write
the explanation ranged from 15 to 20 min and was also substan-
tially the same across conditions. In all conditions, participants also
indicated the grade they had in science at the end of the previous
term of the current school year.
The second session for participants in all conditions took place
2 months later. The students again provided written explanations
of the animation they had observed previously. Between immedi-
ate and delayed posttests, the students received no instruction
on material related to the Newton’s Cradle animation.
4. Results
Alpha was set at .05 for all statistical tests, with a Bonferroni
adjustment applied to p values where necessary.
4.1. Research Question 1: Immediate and delayed comprehension of
the animation
To answer the first research question, a repeated measures AN-
COVA was conducted with condition (self-generated drawing,
traced/copied drawing, and no-drawing) as between-subjects vari-
able, time (immediate posttest and delayed posttest) as the with-
in-subjects variable, comprehension scores (immediate and
delayed) as the dependent variables, and topic prior knowledge
(self-generated drawing: M = 0.84, SD = .71; trace/copied drawing:
M = 1.09, SD = .73; control: M = 1.01, SD = .68) and achievement in
science (self-generated drawing: M = 6.97, SD = 1.04; trace/copied
drawing: M = 6.50, SD = 1.13; control: M = 7.22, SD = 1.17) as covar-
iates. We first carried out a preliminary analysis to evaluate the
homogeneity of regression assumption for the two covariates. This
analysis indicated no effect of the interaction time  condi-
tion  and prior knowledge, Wilks’ Lambda = 0.997, F(2, 190) = .30,
p = .738, ns., as well as no effect of the interaction time 
condition  achievement in science, Wilks’ Lambda = 0.975,
F(2, 190) = 2.47, p = .090, n2p ¼ :02. According to these findings, AN-
COVA was an appropriate analysis. We therefore tested the more
parsimonious model which did not include the non-significant
interactions. The repeated measures ANCOVA revealed a small main
effect of time, Wilks’ Lambda = 0.973, F(1, 194) = 5.43, p = .02,
Table 2
Adjusted means and standard errors of immediate and delayed comprehension as a function of condition.
Variables No-drawing (n = 68)
M SE
Immediate comprehension 2.23 .13
Delayed comprehension 1.71 .12
217
n2p ¼ :03, indicating that across all conditions, comprehension scores
decreased from immediate to delayed posttest. There was also a
medium between-subject effect of condition, F(1, 194) = 7.51,
p = .001, n2p ¼ :07, while the interaction time x condition instead
was not significant, F(2, 194) = .008, p = .992, ns. These findings indi-
cate that comprehension scores for participants in the self-gener-
ated drawing condition were superior to those of participants in
the other two conditions, regardless the testing time. Bonferroni-ad-
justed pairwise comparisons indicated that those in the self-gener-
ated drawing condition differed significantly from those in both the
trace/copy condition, t(196) = 3.12, p = .006, d = .46, and no-drawing
condition, t(196) = 3.51, p = .002, d = .49. However, there were
no differences between those in the two latter conditions,
t(196) = .23, p = 1.00, ns. In other words, the score differences be-
tween the conditions were substantially stable in favor of those ob-
tained by participants in the condition of self-generated drawing,
which was overall more effective than the others two conditions.
Both prior knowledge, F(1, 194) = 5.16, p = .05, n2p ¼ :02, and
achievement in science, F(1, 194) = 22.02, p < .001, n2p ¼ :10, were
significantly related to comprehension. Adjusted means and stan-
dard deviations of comprehension scores are reported in Table 2.
4.2. Research Question 2: Overall contribution of self-generated
drawing to animation comprehension
To answer the second research question, we examined the contri-
bution of the quality of self-generated drawings. We carried out cor-
relational analyses that examined the association of all variables
with immediate and delayed comprehension of the animation. Ta-
ble 3 displays means and standard deviations for these variables, to-
gether with zero-order correlations. The total scores for the six
drawings correlated positively with comprehension scores at both
testing times, as well as with each other. The better the overall gra-
phic productions, the better were the explanations of the scientific
phenomenon shown in the animation. In contrast with the previous
analysis, achievement in science and topic prior knowledge were
also correlated with drawing and comprehension scores.
Hierarchical multiple regression analyses were first conducted
to examine whether overall scores for self-generated drawings un-
iquely predicted immediate and delayed comprehension, while to-
pic prior knowledge and achievement in science simultaneously
included in the first step, were controlled. Immediate and delayed
comprehension results are presented separately.
4.2.1. Immediate comprehension
Table 4a summarizes the hierarchical regression analysis for
variables predicting immediate comprehension with total scores
for six drawings entered separately.
Topic prior knowledge and achievement in science were en-
tered on the first step of the hierarchical regression analysis, and
accounted for a significant 15% of the variance in immediate com-
prehension, R2 = .15, F(2, 64) = 5.87, p = .005. This step was the
same for all regression analyses related to immediate comprehen-
sion. In the second step, only the total score for the six drawings
was entered. In combination, the predictors accounted for 44% of
the variance in immediate comprehension, R2 = .44, adjusted
R2 = .41, F(3, 63) = 16.28, p < .001, with an increment of 29%.
Traced/copied drawing (n = 64) Self-generated drawing (n = 67)
M SE M SE
2.25 .14 2.76 .13
1.76 .13 2.27 .12
218 L. Mason et al. / Contemporary Educational Psychology 38 (2013) 211–224

Table 3
Descriptive statistics and zero-order correlations for all variables in the self-generated drawing condition (n = 67).

Variables 1 2 3 4 5 6 7 8 9 10
* * *
1. Topic prior knowledge .19 .29 .14 .24 .21 .23 .03 .30 .34**
2. Achievement in science .19 .31* .00 .20 .23 .23 .08 .30* .54
3. Total score for drawings .29* .31* .25 .66** .60** .63** .45* .64** .55**
4. End balls alternate swinging .14 .00 .25* .10 .18 .12 .22 .07 .11
5. Motion transmission .24* .20 .66** .10 .15 .31 .04 .11 47**
6. Swinging decrease .21 .23 .60** .18 .15 .18 .09 .32** 27*
7. Joint swinging .23 .23 .63** .12 .31** .18 .00 .50** 41**
8. Final stop .01 .08 .45** .22 .04 .09 .00 .41** 13
9. Immediate comprehension .30* .30* .64** .07 .53** .32** .50** .14 .70**
10. Delayed comprehension .34** .54** .55** .11 .47** .27* .41** .13 .70**
M .84 6.97 3.36 .99 .49 .70 .49 .76 2.75 2.25
SD .70 1.04 1.15 .12 .32 .46 .30 .43 1.21 1.35
*
p < .05, two-tailed;
**
p < .01, two-tailed.

Table 4 Table 5
Hierarchical multiple regression analyses for variables predicting immediate com- Hierarchical multiple regression analyses for variables predicting delayed compre-
prehension (n = 67) with total scores for drawings entered separately (a) and with hension (n = 67), with total scores for drawings entered separately (a) and with scores
scores for the five phases in the drawings entered separately (b). for the five phases in drawings entered separately (b).

Predictor Drawings Predictors Drawings

DR 2 b DR2 b
(a) (a)
Step1 .15** Step1 .35***
*
Topic prior knowledge .25 Topic prior knowledge .25*
Achievement in science .26* Achievement in science .49***
Step 2 .29*** Step 2 .13***
Topic prior knowledge .11 Topic prior knowledge .15
Achievement in science .10 Achievement in science .39***
Total score for drawings .58*** Total score for drawings .39***
(b) (b)
Step1 .15** Step1 .35***
*
Topic prior knowledge .25 Topic prior knowledge .25*
Achievement in science .26⁄ Achievement in science .49***
Step 2 .32** Step 2 .15**
Topic prior knowledge .09 Topic prior knowledge .13
Achievement in science .09 Achievement in science .40***
End balls alternate swinging .01 End balls alternate swinging .06
Motion transmission .36** Motion transmission .28**
Swinging decrease .15 Swinging decrease .04
Joint swinging .32** Joint swinging .20**
Final stop .10 Final stop .07
* *
p < .05. p < .05.
** **
p < .01. p < .01.
*** ***
p < .001. p < .001.

The only significant predictor in the final regression model was the ence and the total score for drawings. The effect size was large
total score for drawings. The effect size was large (f2 = .78). (f2 = .92).

4.2.2. Delayed comprehension

Corresponding regression analyses computed to examine the
unique contribution of self-generated drawing to delayed compre-
hension produced similar outcomes. Table 5a summarizes the hier-
archical regression analysis for variables predicting delayed
comprehension with total scores for six drawings entered
separately.
Topic prior knowledge and achievement in science were once
again entered on the first step of the hierarchical regression anal-
ysis, and accounted for a significant 35% of the variance in delayed
comprehension, R2 = .35, F(2, 64) = 17.40, p < .001. The first step
was the same for all regression analyses related to delayed com-
prehension. In the second step, only the total score for the six
drawings was entered. In combination, the four predictors ac-
counted for 48% of the variance, R2 = .48, adjusted R2 = .45,
F(3, 63) = 19.30, p < .001, with an increment of 13%. The significant
predictors in the final regression model were achievement in sci-
4.3. Research Question 3: Specific contribution of self-generated
drawing to animation comprehension
To answer the third research question, we examined specific
contributions of the drawing of individual phases. First, the corre-
lation analyses (Table 3) revealed that the accurate depiction of the
two perceptually salient phases (i.e., decreasing amplitude of end
ball swinging and joint swinging of all balls) and accurate depic-
tion of the high relevance, low salience phase of motion transmis-
sion, correlated positively with comprehension scores at both
testing times. Depictions of motion transmission and joint swing-
ing were also correlated. Furthermore, the former correlated with
topic prior knowledge. More fine-grained regression analyses were
performed successively, with the quality score for each of the sin-
gle aspects represented in the set of six drawings entered in the
second step as predictors of immediate or delayed comprehension.
 L. Mason et al. / Contemporary Educational Psychology 38 (2013) 211–224
4.3.1. Immediate comprehension
The hierarchical regression analysis presented in Table 4b was
conducted with scores for the five phases entered separately in
the first step as predictors of immediate comprehension. In
combination, all predictors accounted for 47% of the variance
in immediate comprehension, R2 = .47, adjusted R2 = .41,
F(7, 59) = 7.07, p < .001, with an increment of 32%. Only represen-
tations of two aspects of the phenomenon were significant pre-
dictors in the final regression model: (i) the joint swinging of all
balls as a unit and (ii) motion transmission. The effect size was
large (f2 = .92).
4.3.2. Delayed comprehension
The fine-grained hierarchical regression analysis presented in
Table 5b was conducted with scores for the five phases entered
separately in the second step as predictors of delayed comprehen-
sion. In combination, all predictors accounted for 50% of the vari-
ance, R2 = .50, adjusted R2 = .45, F(7, 59) = 8.65, p < .001, with an
increment of 15%. The significant predictors in the final regression
model were not only the joint swinging of all balls as a unit and
motion transmission as before, but also achievement in science.
The effect size was large (f2 = 1.02).
5. Discussion
This study extends previous research on learning from anima-
tions by demonstrating the utility of self-generated drawing as a
strategy for improving learning effectiveness. It focused on the po-
tential of self-generated drawing to support comprehension of
complex dynamic depictions (not comprehension of complex texts
as in most previous research). More specifically, it examined
whether the constructive transformational activity invoked by
such drawing could improve comprehension of a Newton’s Cradle
animation as reflected in written explanations of the represented
phenomenon. Because our participants were seventh graders, com-
prehension of the animation was characterized in terms of high
relevance information necessary as the foundation for building
subsequent physics understandings (rather than the learning of
the underlying physics principles per se). This is particularly
dependent on the capacity to extract and internalize high rele-
vance aspects that are low in perceptual salience. In addition, we
sought to examine whether there was a relation between the qual-
ity of participants’ self-generated drawings and their comprehen-
sion performance.
Our first research question asked whether participants in the
self-generated drawing condition would outperform, at immedi-
ate and delayed posttests, those who merely reproduced pro-
vided drawings, or who did no drawing at all. The results
supported our hypothesis showing that comprehension perfor-
mance of those in the self-generated condition was superior
overall to the comprehension of those in the other two condi-
tions (which did not differ) regardless of the testing time. This
finding has parallels with the outcomes of previous research in
which drawing was used to support learning from text (Ains-
worth, 2010; Leopold & Leutner, 2012; Schwamborn et al.,
2010; Van Meter, 2001; Van Meter & Garner, 2005). Having stu-
dents generate their own external representations through a ser-
ies of static drawings supported their comprehension of the
animated visualization.
Our second research question asked whether the overall quality
of the information depicted in the six self-generated drawings
would predict immediate and delayed comprehension. We were
interested in how participants’ depiction of information from the
animation that varied in its salience and relevance was related to
their comprehension, as revealed in the written explanations. Of
219
particular interest was how well the extraction of specific aspects
from the animation would predict immediate and delayed compre-
hension. Results from regression analyses in which other variables
that may contribute, such as topic prior knowledge and achieve-
ment in science were controlled, supported our hypothesis.
Concerning immediate comprehension, the overall quality
with which the various aspects of the animation were depicted
in the self-generated drawings predicted the quality of partici-
pants’ written explanations. Richer and more accurate drawings
were associated with better comprehension of the animation.
Prior knowledge of the topic and achievement in science were
also positively related to posttest comprehension in the first step
only, that is, before entering the drawing score into the regres-
sion analysis.
Findings from regression analyses also supported our hypothe-
sis with respect to delayed comprehension. The overall quality of
the drawings predicted delayed comprehension. In this case,
achievement in science also contributed to the criterion variable
in the second step of the analysis, while in the first step, both topic
prior knowledge and achievement in science were significant
predictors.
Our third research question asked which individual aspects
of the self-generated depictions would predict immediate and
delayed comprehension. Finer-grained regression analysis find-
ings again supported our hypothesis. The depiction of both
information with high perceptual salience (joint swinging) and
information that was much less salient but high in conceptual
relevance (motion transmission) predicted immediate compre-
hension. This outcome suggests that participants who generated
their own drawings were able not only to process conspicuous
aspects of the animation but also to process relatively incon-
spicuous aspects in order to extract key thematically relevant
information.
Finer-grained analysis regarding the specific predictors for de-
layed comprehension produced findings similar to those for imme-
diate comprehension, confirming our hypothesis. Depiction of both
the joint swinging and motion transmission predicted students’
performance. This outcome further supports the view that partici-
pants who performed well in terms of comprehension were able to
extract and internalize visually subtle but conceptually relevant
aspects which allowed them to construct a relatively stable mental
model of the phenomenon. Achievement in science also contrib-
uted to the quality of verbal explanations produced 2 months later.
This finding suggests that a higher general competence in science
can aid construction of a robust mental representation from the
animation.
Taking into account the Generative Theory of Drawing Con-
struction (GTDC, Van Meter & Garner, 2005; Van Meter et al.,
2006), a plausible interpretation of the overall results of this
study is that self-generation of drawings aids immediate and
longer term comprehension because it provokes careful observa-
tion and extensive mental manipulation of the presented ani-
mated content. By this account, the recursive cycling back and
forth between the animation and the evolving drawing sup-
ported the processes of selecting the key elements of the given
external representation, organizing them, and constructing an
internal representation.
Chi (2009) attributes the benefits of constructive learning activ-
ity, such as self-generated drawing, to its requirement that learners
transform the original learning material into a different output.
This can involve a variety of processes ranging from more intensive
scrutiny of the starting material to its elaboration beyond what is
explicitly provided.
In the present study, the self-generated drawing task encour-
aged both greater scrutiny of the information available in the ani-
mation and the making of inferences about how the dynamic
220 L. Mason et al. / Contemporary Educational Psychology 38 (2013) 211–224
changes it presented were taking place. In undertaking this task,
participants were required to search out and make explicit key
conceptually relevant information such as the perceptually incon-
spicuous aspect of motion transmission. In contrast, such scrutiny
and elaboration were not necessary for those in the trace/copy
condition where drawing was a mere physical activity. As a result,
their comprehension was no better than that of participants who
did no drawing at all. By this interpretation, drawing-to-learn is
somewhat analogous to writing-to-learn in its support for transfor-
mative manipulations that lead to deeper processing of the learn-
ing material (Boscolo & Mason, 2001). However, as pointed out by
Ainsworth et al. (2011), unlike written summaries or verbal self-
explanations, ‘‘visual representations have distinct attributes that
match the visual–spatial demand of much of science learning’’ (p.
1097).
In sum, the present investigation extends earlier research
into the effects of self-generated drawing on learning from
science texts to encompass its effects on learning from anima-
tion. Given the proliferation of animated explanations in today’s
digital learning environments, the finding that this activity has
positive effects has important implications for educational
practice.
5.1. Limitations and directions for future research
This research reported here is a first study in a novel area of
investigation. As such, it has some limitations that should be ad-
dressed in future research. First, we measured comprehension of
the animation using only one task that asked an explanation of
what happens in the dynamic visualization. In further studies it
would be beneficial to examine transfer as additional evidence
for comprehension. The acquisition of a high quality mental model
from an animation should not only allow learners to explain the
presented subject matter, it should also equip them to apply their
understandings to novel situations. Transfer tests, such as those
used in many multimedia studies (see Mayer, 2005) would provide
valuable insights into the potential of self-generated drawing to
support learning from animations.
A second potential limitation of the study is that we did not
test participants’ graphicacy skills. The effectiveness of self-gen-
erated drawing may depend to some extent on the capacity of
learners to take proper advantage of its potential, particularly
with respect to the quality of the drawings they produce. Their
drawings need to include not only high relevance information
that is visually conspicuous but also relevant information that
is low in perceptual salience. Some students may be less able
to generate such drawings because they lack the graphicacy
skills to portray dynamic content in a static depiction (de Vries
& Lowe, 2010). Future research should consider how learners’
graphicacy skills may influence the potential of self-generated
drawing.
A third limitation concerns the issue of participants’ level of
engagement with the animated material. Research on using self-
generated drawing for learning from text has shown that students
who receive support are more successful than those without sup-
port (Van Meter et al., 2006). Recent studies on writing-to-learn
also indicate that guidance leads to better results. Students’ under-
standing and retention of information is improved if the writing
activity is well supported, for example, by metacognitive prompts
that encourage deeper, more reflective learner processing of stim-
ulus materials (Hübner, Nuckles, & Renkl, 2010; Nückles, Hübner,
& Renkl, 2009). Could this also be the case for drawing-to-learn
from animation? One key aspect of support not addressed in our
study is the role that could be played by instructions about what
type of self-generated drawing is required. In our study, partici-
pants were asked only to produce a certain number of drawings,
but were given no specific instructions about the nature of the
drawings to be produced. Guided and unguided drawing condi-
tions could be compared in future research to gain more informa-
tion on the constructive nature of this activity in the service of
student learning.
One form of support that may be worth investigating is whether
it would be possible to compensate for a lack of graphicacy skills
by providing students with explicit instruction in how to depict
subtle dynamic processes in their drawings by using arrows, dot-
ted lines, etc. Further, providing user control could also help less
able learners to comprehend a demanding animation by allowing
them to interrogate the presented material at their own pace and
with more flexibility (e.g., Boucheix & Schneider, 2009). However,
there can be large individual differences in how well user control
options are used by learners (Lowe, 2008; Tabbers & de Koeijer,
2010).
A fourth limitation of the study is related to the generalizabil-
ity of the findings. Because a very particular type of subject mat-
ter was used in this research, further research is needed to
determine whether our findings for the Newton’s Cradle are
applicable to animations depicting other types of content (e.g.,
biological rather than mechanical; Lowe & Boucheix, 2012). It
would also be important to investigate whether benefits from
self-generated drawing are found for other student populations,
such as those who are older and have more prior knowledge
of the presented content.
Finally, the interplay between learner and media characteris-
tics cannot be overlooked in studies that aim to extend current
understanding of learning from visual media. In this regard,
not only perceptual and cognitive factors as used in this study,
but also various motivational differences should be considered
to advance research on learning from animation (Scheiter &
Gerjets, 2010).
5.2. Conclusion
This study both broadens the repertoire of possible interven-
tions that educators could consider for improving learning from
animation and contributes to the literature on drawing-to-learn
in science (Ainsworth et al., 2011). Self-generated drawing offers
a promising alternative to existing approaches such as user control
and cueing that have produced mixed results. Its effectiveness ap-
pears to stem from the opportunities that transforming content
from one representation to another offer for acquiring key aspects
from provided information and linking these into a coherent
whole. In particular, the task of generating drawings from an ani-
mation provokes deeper processing of the depicted content that in-
volves both more intense inspection of the depicted events and
beneficial mental manipulation of the information extracted as a
result. With the increasing reliance that technology-based learning
materials place on animated explanations, the findings from this
study open up a promising new line of research on ways to help
animations fulfill their undoubted but too often unrealized educa-
tional potential.
Acknowledgments
The study is part of an on-going research project on learning
difficulties in the science domain (STPD08HANE_001) funded by
a grant to the first author from the University of Padova, Italy, un-
der the founding program for ‘‘Strategic Projects’’.
We wish to thank Gaia Mantovani for partial data collection.
We are also very grateful to all the students involved in the study,
their parents and teachers, and the school principals.
 L. Mason et al. / Contemporary Educational Psychology 38 (2013) 211–224 221

Appendix A

The six cards of the animation provided in the condition of traced/copied drawing.

Appendix B
Examples of participants’ depictions (coded 1 point each) of the Phase 2: Motion is transmitted via activity of the middle balls
various phases of the animation. (P3)
Phase 1: The balls at the end of the row swing (P31).
222 L. Mason et al. / Contemporary Educational Psychology 38 (2013) 211–224

Phase 3: The amplitude of end ball swinging progressive de-

creases (P31).

Phase 4: All balls swing jointly as a unit (P52).

Appendix C

Example of an inferior drawing set (P36).

Phase 5: All motion finally ceases (P60).

 L. Mason et al. / Contemporary Educational Psychology 38 (2013) 211–224 223

Example of a superior drawing set (P8).

References de Vries, E., & Lowe, R. K. (2010, August). Graphicacy: What does the learner bring to a
graphic? Paper presented at the meeting of the Special Interest Group
‘‘Comprehension of Text and Graphics’’ of the European Association for
Ainsworth, S. (2006). DeFT: A conceptual framework for considering learning with
Research on Learning and Instruction, Tübingen, Germany.
multiple representations. Learning and Instruction, 16(3), 183–198. http://
Gunel, M., Hand, B., & McDermott, M. A. (2009). Writing for different
dx.doi.org/10.1016/j.learninstruc.2006.03.001.
audiences: Effects on high-school students’ conceptual understanding of
Ainsworth, S. (2010). Improving learning by drawing. In K. Gomez, L. Lyons, & J.
biology. Learning and Instruction, 19(4), 354–367. http://dx.doi.org/10.1016/
Radinsky (Eds.). Learning in the disciplines: Proceedings of the 9th international
j.learninstruc.2008.07.001.
conference of the learning sciences (Vol. 2, pp. 167–168). Chicago, IL:
Hegarty, M. (2004). Dynamic visualizations and learning: Getting to the difficult
International Society of the Learning Sciences.
questions. Learning and Instruction, 14, 343–351. http://dx.doi.org/10.1016/
Ainsworth, S., Prain, V., & Tytler, R. (2011). Drawing to learn in science. Science,
j.learninstruc.2004.06.007.
333(6046), 1096–1097. http://dx.doi.org/10.1126/science.1204153.
Hegarty, M., Canhan, M. S., & Fabrikant, S. I. (2010). Thinking about the weather:
Ainsworth, S., & Van Labeke, N. (2004). Multiple forms of dynamic representation.
How display salience and knowledge affect performance in a graphic inference
Learning and Instruction, 14(3), 241–255. http://dx.doi.org/10.1016/
task. Journal of Experimental Psychology: Learning, Memory, and Cognition, 36(1),
j.learninstruc.2004.06.002.
37–53. http://dx.doi.org/10.1037/a0017683.
Ayres, P., & Paas, F. (2007). Can the cognitive load approach make instructional
Hegarty, M., Kriz, S., & Cate, C. (2003). The roles of mental animations and external
animations more effective? Applied Cognitive Psychology, 21, 811–820. http://
animations in understanding mechanical systems. Cognition and Instruction, 21,
dx.doi.org/10.1002/acp.1351.
209–249. http://dx.doi.org/10.1207/s1532690xci2104_1.
Bétrancourt, M. (2005). The animation and interactivity principles. In R. E. Mayer
Höffler, T. N., & Leutner, D. (2007). Instructional animation versus static pictures: A
(Ed.), The Cambridge handbook of multimedia learning (pp. 287–296). New York:
meta-analysis. Learning and Instruction, 17(6), 722–738. http://dx.doi.org/
Cambridge University Press.
10.1016/j.learninstruc.2007.09.013.
Betrancourt, M., & Chassot, A. (2008). Making sense of animation: How do children
Hübner, S., Nuckles, M., & Renkl, A. (2010). Writing learning journals: Instructional
explore multimedia instruction? In R. Lowe & W. Schnotz (Eds.), Learning with
support to overcome learning-strategy deficits. Learning and Instruction, 20(1),
animation: Research implications for design (pp. 141–164). New York: Cambridge
18–29. http://dx.doi.org/10.1016/j.learninstruc.2008.12.001.
University Press.
Klein, P. D. (2000). Elementary students’ strategies for writing-to-learn in science.
Bétrancourt, M., & Tversky, B. (2000). Effect of computer animation on users’
Cognition and Instruction, 18, 317–348. http://dx.doi.org/10.1207/
performance: A review. Le Travail Humain: A Bilingual and Multi-Disciplinary
S1532690XCI1803_2.
Journal in Human Factors, 63, 311–329.
Kombartzky, U., Ploetzner, R., Schlag, S., & Metz, B. (2010). Developing and
Bodemer, D., Ploetzner, R., Feuerlein, I., & Spada, H. (2004). The active integration of
evaluating a strategy for learning from animation. Learning and Instruction,
information during learning with dynamic and interactive visualisations.
20(5), 424–433. http://dx.doi.org/10.1016/j.learninstruc.2009.05.002.
Learning and Instruction, 14(3), 325–341. http://dx.doi.org/10.1016/
Leopold, C., & Leutner, D. (2012). Science text comprehension: Drawing, main idea
j.learninstruc.2004.06.006.
selection, and summarization as learning strategies. Learning and Instruction,
Boscolo, P., & Mason, L. (2001). Writing to learn, writing to transfer. In P. Tynjala, L.
22(1), 16–26. http://dx.doi.org/10.1016/j.learninstruc.2011.05.005.
Mason, & K. Lonka (Eds.), Writing as a learning tool: Integrating theory and
Lowe, R. K. (1999). Extracting information from an animation during complex visual
practice (pp. 83–104). Dordrecht, NL: Kluwer Academic Publishers.
learning. European Journal of Psychology of Education, 14, 225–244. http://
Boucheix, J. M., & Lowe, R. K. (2010). An eye tracking comparison of external
dx.doi.org/10.1007/BF03172967.
pointing cues and internal continuous cues in learning with complex
Lowe, R. K. (2003). Animation and learning: Selective processing of information in
animations. Learning and Instruction, 20(2), 123–135. http://dx.doi.org/
dynamic graphics. Learning and Instruction, 13(2), 157–176. http://dx.doi.org/
10.1016/j.learninstruc.2009.02.015.
10.1016/S0959-4752%2802%2900018-X.
Boucheix, J. M., & Schneider, E. (2009). Static and animated presentations in
Lowe, R. K. (2004). Interrogation of a dynamic visualization during
learning dynamic mechanical systems. Learning and Instruction, 19(2), 112–127.
learning. Learning and Instruction, 14(3), 257–274. http://dx.doi.org/10.1016/
http://dx.doi.org/10.1016/j.learninstruc.2008.03.004.
j.learninstruc.2004.06.003.
Chi, M. T. H. (2009). Active–constructive–interactive: A conceptual framework for
Lowe, R. K. (2008). Learning from animation: Where to look, when to look. In R. K.
differentiating learning activities. Topics in Cognitive Science, 1, 73–105. http://
Lowe & W. Schnotz (Eds.), Learning with animation: Research implications for
dx.doi.org/10.1111/j.1756-8765.2008.01005.x.
design (pp. 49–68). New York.
Cox, R. (1999). Representation construction, externalized cognition and individual
Lowe, R. K., & Boucheix, J. M. (2008). Learning from animated diagrams: How are
differences. Learning and Instruction, 9(4), 343–363. http://dx.doi.org/10.1016/
mental models built? In G. Stapleton, H. Howse, & J. Lee (Eds.), Diagrammatic
S0959-4752(98)00051-6.
representation and inference (pp. 266–281). Berlin: Springer.
224 L. Mason et al. / Contemporary Educational Psychology 38 (2013) 211–224
Lowe, R. K., & Boucheix, J. M. (2011). Cueing complex animations: Does direction of
attention foster learning processes? Learning and Instruction, 21(5), 650–663.
http://dx.doi.org/10.1016/j.learninstruc.2011.02.002.
Lowe, R. K., & Boucheix, J. -M. (2012, August). Addressing challenges of biological
animations. Paper presented at the meeting of the Special Interest Group
‘‘Comprehension of Text and Graphics’’ of the European Association for
Research on Learning and Instruction, Grenoble, France.
Mason, L. (2001). Introducing talk and writing for conceptual change: A classroom
study. Learning and Instruction, 11(4–5), 305–329. http://dx.doi.org/10.1016/
S0959-4752(00)00035-9.
Mayer, R. R. (Ed.). (2005). The Cambridge handbook of multimedia learning. New York:
Cambridge University Press.
Narayanan, N. H., & Hegarty, M. (2002). Multimedia design for communication of
dynamic information. International Journal of Human–Computer Studies, 57,
279–315. http://dx.doi.org/10.1006/ijhc.2002.1019.
Nückles, M., Hübner, S., & Renkl, A. (2009). Enhancing self-regulated learning by
writing learning protocols. Learning and Instruction, 19, 259–271. http://
dx.doi.org/10.1016/j.learninstruc.2008.05.002.
Rebetez, C., Bétrancourt, M., Sangin, M., & Dillenbourg, D. (2010). Learning from
animation enabled by collaboration. Instructional Science, 38, 471–485. http://
dx.doi.org/10.1007/s11251-009-9117-6.
Rieber, L. P. (1991). Animation, incidental learning, and continuing motivation.
Journal of Educational Psychology, 83, 318–328. http://dx.doi.org/10.1037/0022-
0663.83.3.318.
Rieber, L. P., Tzeng, S.-C., & Tribble, K. (2004). Discovery learning, representation,
and explanation within a computer-based simulation: Finding the right mix.
Learning and Instruction, 14(3), 307–323. http://dx.doi.org/10.1016/
j.learninstruc.2004.06.008.
Scheiter, K., & Gerjets, P. (2010). Cognitive and socio-motivational aspects in
learning with animations: There is more to it than ‘do they aid learning or not’.
Instructional Science, 38, 435–440. http://dx.doi.org/10.1007/s11251-009-9118-
5.
Schnotz, W., & Kulhavy, R. (1994). Comprehension of graphics. Amsterdam, The
Netherlands: North Holland/Elsevier.
Schnotz, W. (2005). An integrated model of text and picture comprehension. In R. E.
Mayer (Ed.), The Cambridge handbook of multimedia learning (pp. 48–69). New
York: Cambridge University Press.
Schnotz, W., & Lowe, R. K. (2008). A unified view of learning from animated and
static graphics. In R. K. Lowe & W. Schnotz (Eds.), Learning with animation:
Research implications for design (pp. 256–304). New York: Cambridge University
Press.
Schwamborn, A., Mayer, R. E., Thillmann, H., Leopold, C., & Leutner, D. (2010).
Drawing as a generative activity and drawing as a prognostic activity. Journal of
Educational Psychology, 102, 872–879. http://dx.doi.org/10.1037/a0019640.
Sweller, J. (2005). Implications of cognitive load theory for multimedia learning. In
R. E. Mayer (Ed.), The Cambridge handbook of multimedia learning (pp. 19–30).
New York: Cambridge University Press.
Tabbers, H. K., & de Koeijer, B. (2010). Learner control in animated multimedia
instructions. Instructional Science, 38, 441–453. http://dx.doi.org/10.1007/
s11251-009-9119-4.
Tversky, B., Morrison, J. B., & Bétrancourt, M. (2002). Animation: Can it facilitate?
International Journal of Human–Computer Studies, 57, 247–262. http://dx.doi.org/
10.1006/ijhc.101.
Van Meter, P. (2001). Drawing construction as a strategy for learning from text.
Journal of Educational Psychology, 93, 129–140. http://dx.doi.org/10.1037/0022-
0663.93.1.129.
Van Meter, P., Aleksic, M., Schwartz, A., & Garner, J. (2006). Learner-generated
drawing as a strategy for learning from content area text. Contemporary
Educational Psychology, 31(2), 142–166. http://dx.doi.org/10.1016/
j.cedpsych.2005.04.001.
Van Meter, P., & Garner, J. (2005). The promise and practice of learner-generated
drawing: Literature review and synthesis. Educational Psychology Review, 17,
285–325. http://dx.doi.org/10.1007/s10648-005-8136-3.
Wouters, P., Paas, F., & van Merriënboer, J. J. G. (2009). Observational learning from
animated models: Effects of modality and reflection on transfer. Contemporary
Educational Psychology, 34(1), 1–8. http://dx.doi.org/10.1016/
j.cedpsych.2008.03.001.
Zhang, H. Z. (2010). Exploring drawing and critique to enhance learning from
visualizations. In K. Gomez, L. Lyons, & J. Radinsky (Eds.). Learning in the
disciplines: Proceedings of the 9th international conference of the learning sciences
(Vol. 2, pp. 234). Chicago, IL: International Society of the Learning Sciences.
Zhang, H. Z., & Linn, M. (2011). Can generating representations enhance learning
with dynamic visualisations? Journal of Research in Science Teaching, 48(10),
1177–1198. http://dx.doi.org/10.1002/tea.20443.