You are on page 1of 10

See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/220495654

Instructional animations can be superior to static when learning human


motor skills

Article  in  Computers in Human Behavior · March 2009


DOI: 10.1016/j.chb.2008.12.012 · Source: DBLP

CITATIONS READS

163 1,127

7 authors, including:

Nadine Marcus Graham Cooper


UNSW Sydney Southern Cross University
56 PUBLICATIONS   1,731 CITATIONS    23 PUBLICATIONS   2,478 CITATIONS   

SEE PROFILE SEE PROFILE

Fred Paas
Erasmus University Rotterdam and University of Wollongong
318 PUBLICATIONS   34,329 CITATIONS   

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Ace Your Self-study: Using a Mobile Device App to Support Self-regulated Learning View project

PhD "Cueing in multimedia practicals" View project

All content following this page was uploaded by Nadine Marcus on 31 January 2018.

The user has requested enhancement of the downloaded file.


Computers in Human Behavior 25 (2009) 339–347

Contents lists available at ScienceDirect

Computers in Human Behavior


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

Instructional animations can be superior to statics when learning


human motor skills
Anna Wong a,*, Nadine Marcus a,*, Paul Ayres b, Lee Smith c, Graham A. Cooper c, Fred Paas d,e, John Sweller b
a
School of Computer Science and Engineering, University of New South Wales, Sydney, NSW 2052, Australia
b
School of Education, University of New South Wales, Australia
c
School of Multimedia and Information Technology, Southern Cross University, Australia
d
Educational Technology Expertise Center, Open University of the Netherlands, The Netherlands
e
Institute of Psychology, Erasmus University Rotterdam, The Netherlands

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

Article history: Based on the assumption of a working memory processor devoted to human movement, cognitive
Available online 4 January 2009 load theory is used to explore some conditions under which animated instructions are hypothesised
to be more effective for learning than equivalent static graphics. Using paper-folding tasks dealing
Keywords: with human movement, results from three experiments confirmed our hypothesis, indicating a supe-
Instructional animations riority of animation over static graphics. These results are discussed in terms of a working memory
Human motor skills processor that may be facilitated by our mirror-neuron system and may explain why animated
Mirror-neuron system
instructional animations are superior to static graphics for cognitively based tasks that involve human
Cognitive load theory
movement.
Crown Copyright ! 2008 Published by Elsevier Ltd. All rights reserved.

1. Introduction the pace and order of presentations, and engage more active pro-
cessing as learners are required to make inferences from frame
Instructional animations have a varied history. Most studies to frame. We suggest that the poor performance of instructional
have concluded that animation is at best no more effective and animations compared with static graphic can be explained by cog-
sometimes less effective than the equivalent static graphics (e.g., nitive load theory.
Hegarty, Kriz, & Cate, 2003; Mayer, DeLeeuw, & Ayres, 2007; Cognitive load theory (Clark, Nguyen, & Sweller, 2006; Paas,
Mayer, Hegarty, Mayer, & Campbell, 2005; Schnotz, Böckheler, & Renkl, & Sweller, 2003; Sweller, 1999, 2004; Sweller & Sweller,
Grzondziel, 1999; Tversky, Morrison, & Betrancourt, 2002). How- 2006) is concerned with relations between working memory
ever, when additional supporting strategies are introduced such (WM) and long-term memory (LTM) and the effects of those rela-
as segmenting animations into smaller parts (see Hasler, Kersten, tions on learning and problem solving. The theory applies to a par-
& Sweller, 2007; Mayer & Chandler, 2001; Moreno, 2007) or pro- ticular category of knowledge that Geary (2002, 2005, 2007) refers
viding user-control (see Hasler et al., 2007; Mayer & Chandler, to as biologically secondary knowledge. Biologically secondary
2001; Schwan & Riempp, 2004), animations become more effec- knowledge, such as learning differential calculus or how to write,
tive. Nevertheless, a recent meta-analysis by Höffler and Leutner is knowledge that we have not evolved to learn readily, and is ac-
(2007) found that animations facilitated better learning outcomes quired with conscious, effortful processing within WM. On the
than static graphics if they contained certain features, e.g., realism other hand, biologically primary knowledge is not learned con-
and procedural-motor knowledge. Reasons have been proposed for sciously because we have evolved to acquire that knowledge easily
the frequent ineffectiveness of animations, but no comprehensive and automatically. The examples of learning a first language, rec-
theory has been articulated. For instance, Lowe (1999, 2003) ognising faces, learning general problem solving techniques, or
looked at learning from weather map animations and suggested learning about basic social interactions fall into this category. Such
that perceptually salient features of an animation may attract knowledge can be stored directly in LTM without being consciously
learners’ attention away from more-subtle, but thematically rele- processed by WM (Geary, 2007). Because we have not evolved to
vant features. Mayer et al. (2005) proposed that statics allow better effortlessly acquire the biologically secondary knowledge charac-
management of intrinsic processing by allowing learners to control teristically taught in educational institutions, it must be explicitly
taught and we have invented educational institutions in order to
* Corresponding authors.
teach such knowledge. The manner in which such knowledge is
E-mail addresses: annawong@cse.unsw.edu.au (A. Wong), nadinem@cse.unsw. taught needs to take into account the characteristics of WM, LTM
edu.au (N. Marcus). and the relations between them. Cognitive load theory, based on

0747-5632/$ - see front matter Crown Copyright ! 2008 Published by Elsevier Ltd. All rights reserved.
doi:10.1016/j.chb.2008.12.012
340 A. Wong et al. / Computers in Human Behavior 25 (2009) 339–347

the relations between working and long-term memory, primarily LTM. Instructional procedures not designed to alter LTM and which
applies to biologically secondary, not biologically primary ignore the limitations of WM when dealing with new information
knowledge. are likely to be ineffective. Cognitive load theory provides instruc-
tional design principles based on this cognitive architecture and
identifies three categories of cognitive load (see Ayres, 2006a,b;
2. Cognitive load theory and its biological base Paas et al., 2003; Sweller et al., 1998):
Intrinsic cognitive load is the ‘‘natural” load imposed by the
Cognitive load theory treats human cognition as a natural infor- information that must be acquired. It consists of interacting ele-
mation processing system that duplicates other natural informa- ments of information that must be processed simultaneously in
tion processing systems such as evolution by natural selection WM to understand and learn new material. It cannot be altered
(Sweller, 2003, 2004). As adapted to human cognition, such sys- without affecting understanding.
tems can be specified by five principles: Extraneous cognitive load is the load imposed by instructional
1. Long-term memory and the information store principle. Most procedures. It is under the control of the instructional designer.
human cognitive action is driven by the contents of a huge LTM Most work on cognitive load theory has been concerned with pro-
that acts as an information store. For example, activities such as cedures for reducing extraneous cognitive load and a large number
determining previously learned, domain-specific problem solving of instructional design principles have been generated (see Sweller,
moves are heavily dependent on the contents of LTM (e.g., Chase 2004; van Merriënboer & Ayres, 2005 for reviews).
& Simon, 1973; De Groot, 1965). Germane cognitive load refers to the WM resources required to
2. Schema theory and the borrowing and reorganising principle. acquire information. Any reduction in extraneous cognitive load
How does information enter LTM, given its massive capacity? permits an increase in germane cognitive load thus freeing WM
According to the borrowing and reorganising principle, we borrow capacity for learning and understanding (Paas & van Merriënboer,
information from the long-term stores of other people by imitating 1993).
what they do, listening to what they say or reading what they Cognitive load theory has been used to generate many instruc-
write. Techniques for facilitating knowledge acquisition through tional procedures (Sweller, 2003, 2004). Efficient instructional de-
the borrowing principle are central to cognitive load theory (for signs should therefore manipulate instructional materials and
summaries, see Paas et al., 2003; Sweller, 1999, 2005; Sweller, procedures to reduce extraneous or unnecessary cognitive load
van Merriënboer, & Paas, 1998; van Merriënboer & Sweller, via the borrowing principle and enhance schema construction
2005). Knowledge elements are assumed to be stored in LTM in and automation.
schematic form. A schema is a cognitive construct that allows peo-
ple to categorize multiple elements of information into a frame-
work of knowledge according to the manner in which it will be 3. Human movement, mechanical object movement, and a
used (Sweller, 1999). Schema acquisition based on the borrowing human movement working memory processor
principle attempts to copy other people’s schemas but rarely re-
sults in an exact copy of another person’s schema resulting in inev- Animation has a critical characteristic that is frequently over-
itable changes, some of which are random. looked when used for instructional purposes – it is usually transi-
3. Problem solving and the randomness as genesis principle. The tory (see Ayres & Paas, 2007a,b). The transitory nature of animation
borrowing principle does not create new knowledge except insofar has significant working memory implications. Unless rehearsed,
as borrowing is inexact. New knowledge is created through prob- material will remain in WM for no more than a few seconds. When
lem solving and the randomness as genesis principle. We need to presenting complex, novel instructional material in animated form,
solve problems when unable to borrow knowledge from others learners are likely to find it difficult to both process current infor-
or from LTM. Failing the availability of relevant knowledge, ran- mation and remember previous information. If both current and
domly generating problem solving moves and determining their previous information interact, and so must be considered simulta-
effectiveness provides us with new knowledge. neously in order to be understood, comprehension may be difficult,
4. Working memory and the narrow limits of change principle. because information presented during earlier phases of an anima-
Working memory processes novel information of the sort dealt tion may be lost. The transitory nature of animation linked with
with by the randomness as genesis principle. Because of the ran- the limited duration of WM may be a major reason for failure in
dom nature of novel elements, only a few elements of novel infor- instructional animation as it creates an extraneous cognitive load
mation can be processed in working memory at a time (Miller, due to the need to attempt to hold information in WM for extended
1956; Peterson & Peterson, 1959) resulting in a limited capacity periods. In contrast, static graphical displays can be revisited a
and limited duration working memory. Attempting to organize number of times when required, in a manner difficult or impossible
and establish meaningful relations for more than a few elements when using animation. Consequently, static graphics may reduce
in working memory will result in combinatorial explosions. this source of extraneous cognitive load and possibly eliminate
5. Relations between long-term and working memory and the envi- the problem of transitory information.
ronmental organizing and linking principle. This principle allows us While working memory limitations may limit the effectiveness
to bring unlimited amounts of previously organized schematic of many animations, one possibility is that humans may have
information from long-term to working memory to allow us to evolved a specific component of working memory that may allow
function in our environment (Ericsson & Kintsch, 1995). Thus, us to cope with some animations better than others. As humans
working memory load is reduced through schema acquisition. learn some movements by imitating others via the borrowing
Schemas, which consist of multiple elements, can be transferred and reorganising principle, we may have developed a separate
to and processed by working memory as a single element and so working memory processor for this purpose, just as we have work-
reduce cognitive load. Schema automation, which further de- ing memory processors for auditory and visual information
creases the amount of processing required in working memory (Baddeley, 1992). As a consequence, we may be able to learn from
(e.g., Schneider & Shiffrin, 1977; Shiffrin & Schneider, 1977), can animations more readily when dealing with human movement
further reduce cognitive load. than with mechanical object movement. The following section
These characteristics of human cognition suggest that the pri- outlines some evidence for a human movement, WM processor
mary function of instruction is the acquisition of schemas in and uses cognitive load theory to predict that animations can be
A. Wong et al. / Computers in Human Behavior 25 (2009) 339–347 341

effective when acquiring new knowledge dependent upon human origami model, has been to present a series of static diagrams cap-
movement (motor skills). In doing so, we stress that we intend to turing discrete steps within the procedure in a step-by-step man-
only use one aspect of neuroscience to further understand and en- ner. The learner is generally shown the beginning state of the
hance cognitive load theory. fold and then in the next step, what the model looks like after
Under many theories of WM, humans have separate WM pro- the fold has been made. Ancillary support in the form of arrows
cessors for visual and auditory information (e.g., Baddeley, 1992). is usually given to indicate the direction of the fold to be made,
Effective working memory capacity can be increased by using both along with brief textual descriptions of the actions required. In this
visual and auditory working memory, rather than by either mem- situation, the learner is required to deduce how the change be-
ory stream alone (Penney, 1989). We suggest there may also be a tween the beginning and end step has occurred. In this experiment,
working memory processor for dealing with human movement two versions of a static graphic instructional format were used.
and the perception and imitation of human movement. That pro- One version (the single-static group) was in the traditional format
cessor may relate to the recently described mirror-neuron system described above and consisted of a series of single-static diagrams
(Iacoboni et al., 1999, also see Van Gog, Paas, Marcus, Ayres, & each representing a key fold in the process. In the second version
Sweller, 2007, for a related discussion of the mirror-neuron system (the double-static group) additional information was provided,
and learning by observing). by showing a series of paired static diagrams, each pair illustrating
Iacoboni et al. (1999) provided evidence of similarities in brain the before-and-after shots of each fold. The double-static condition
activities that occur when observing someone carrying out certain was the static graphics analogue of the animation. However, unlike
actions and when carrying out the actions oneself. In neuropsycho- the animated version, where both start and finish of the fold were
logical terms, observing actions made by others activates the cor- shown with one image and an animation demonstrating their rela-
tical circuits responsible for the planning and execution of those tion, the double-static version only depicted the start and end
same actions (Meltzoff & Prinz, 2002). This observation–execu- states of each fold and omitted the animation. The third group
tion-matching system (mirror-neuron system) is thought to play was a narrated animation depicting every action in the paper-fold-
an important role in the understanding of actions made by others. ing process. All three groups received narrated explanations simul-
Recently, Tettamanti et al. (2005) used fMRI scans to show that taneously as the visual information was provided.
when a person listens to another person explain an action they It was hypothesised that students receiving narrated anima-
were about to do (such as put together parts or walk across the tions would learn the paper-folding task more successfully than
room), the neurons in the listener’s brain responsible for those ex- students receiving narrated, static-image based instructions.
act muscle movements were activated. It is believed that these
mirror neurons are responsible for our ability to engage in imita-
tive learning (see Blandin, Lhuisset, & Proteau, 1999) and to empa- 5. Methods
thize. With the discovery of the mirror-neuron system, we can
begin to understand how observing human movements can en- 5.1. Participants and design
hance our mastery of complex motor behaviours. We suggest the
basic capabilities of our mirror-neuron system may be used to help The participants were 32 sixth grade students of mixed levels of
foster more complex motor skill development. If this is the case, ability from a regional New South Wales Public School. The chil-
then animations may tap into these basic capabilities in a way that dren were drawn from four composite 5th/6th grade classes and
static graphics cannot. were in the age range 11–12 years. Participants were randomly as-
Based on these neuropsychological results, we hypothesize that signed to one of three instructional conditions: animated (n = 11),
observational learning from animation showing human manipula- single static (n = 11) and double static (n = 10).
tion may trigger the use of a ‘‘movement” processor in a manner
that does not occur when faced with static graphics or with 5.2. Materials
mechanical object movement. While the above neuropsychological
results by no means prove the existence of a human movement After pilot testing on potential participants in the desired age
processor, they are compatible with the existence of such a range (11–12 years) the Viking Helmet construction (consisting
hypothesised processor. Assuming such a processor exists, observ- of 13 steps) was chosen. The materials for this experiment were
ing human movement may involve biologically primary knowledge developed using a combination of Macromedia Director 8 and Mac-
processed in a very resource-efficient manner and so assist in the romedia Flash 5. The animation material used was manipulated in
acquisition of secondary knowledge. Consequently, instructional Flash 5 and then embedded within a Director 8 projector file,
animations may be potentially ideal for teaching someone to which was used to track the various time attributes related to
manipulate an object. Accordingly, our central hypothesis is that the task. Three different versions of the program were developed,
animations will be superior to equivalent static diagrams when namely, a narrated animation version, a narrated single-static ver-
acquiring knowledge that involves human movement. We tested sion and a narrated double-static version. All three presentations
this hypothesis in a series of three experiments involving paper- used the traditional diagrams for the Viking Helmet model found
folding tasks. in most origami books and all showed each fold to be made.
We sought to make all three instructional programs identical in
as many respects as possible, other than the differences in pre-
4. Experiment 1 sented images. For the animated version, a narration can be heard
explaining the fold as it is made. Fig. 1 shows an example of the
Experiment 1 investigated the benefits of narrated animation- double-static version. Unlike the animated version, where both
based instructions compared to equivalent narrated static gra- start and finish of the fold can be shown in real time, the double-
phic-based instructions in learning a motor skill. An origami pa- static version represents the motion by having two images on
per-folding task was chosen because it requires hand screen together, depicting the start and end states of each fold. A
manipulations and therefore involves human movement and may narration was provided, identical to that used for the animated
thus benefit from the observational learning components of ani- version, to explain how to perform each fold. The single-static ver-
mated instructional materials. A traditional static approach for sion (see Fig. 2) consists of a series of single-static diagrams accom-
demonstrating a complex procedural task, such as how to fold an panied by the narration. As neither of the static versions can
342 A. Wong et al. / Computers in Human Behavior 25 (2009) 339–347

Fig. 1. Screen shot from the double-static version in Experiment 1.

Fig. 2. Screen shot from the single-static version in Experiment 1. Note: The single-static version is similar to the animated version except that in the single-static
presentation, the end state of the fold is not shown and arrows are used to convey the fold, rather than real-time movement.

directly show changes through motion, arrows were used to show the students in the animated condition, pressing this button re-
the direction in which the folds were to proceed. played the animation and narration for the step they were cur-
In each version buttons were present at the bottom of the rently on, while for each of the static conditions the button
screen showing the folds that could be accessed in any particular replayed the narration for that step. The only other materials pro-
stage. Students were able to move through the steps using these vided were the paper squares used for performing the required
buttons. A replay button was also available to all students; for folds during testing.
A. Wong et al. / Computers in Human Behavior 25 (2009) 339–347 343

5.3. Procedure within the animation group were able to complete the required
origami model than students in both the single-static group
Participants were informed that they were going to view a short (p < 0.05) and the double-static group (p < 0.025). No differences
computer tutorial showing how to fold an origami model of a Vik- were observed between the single-static and double-static groups.
ing Helmet and then shown a completed Viking Helmet. Each par- The results of Experiment 1 support our hypothesis that students
ticipant was run individually in one session and was presented who were given the animated version of the paper-folding task
with either animated, double-static or single-static origami performed better than students given either of the equivalent sta-
instructions. In order to ensure that all participants would feel tic graphics versions, and also took less time during acquisition.
comfortable using the program and the procedure of events, a
small demonstration version was made in replica of each instruc-
7. Experiment 2
tional condition demonstrating a simple paper fold. The interface
was explained and participants were informed that they would
Two underlying issues which Tversky et al. (2002) suggest con-
be able to click through each of the steps. It was made explicit that
found the effectiveness of animation over static graphics are: (1)
they should pay careful attention to how each fold was done so
lack of informational equivalence between animated and static
that they would be able to remember them. In the acquisition
graphics conditions, and (2) the inclusion of learner control. In a
phase, participants were told that there would be three stages con-
survey of existing studies, Tversky et al. (2002) found that static
sisting of five folds, three folds, and five folds, respectively, and that
graphics were often at a disadvantage compared to animation be-
after they had worked through the instructions for a particular
cause the static graphics condition often presented less informa-
stage, they would be handed a piece of paper with which they
tion compared to the equivalent animation. It is also known from
could attempt the folds they had just seen. They were also told that
the literature that learner control and interaction can improve
if they could not complete all the folds in a stage, then a model al-
learning from animations (see Hasler et al., 2007; Schnotz et al.,
ready folded to the correct place would be provided to them so that
1999). It is plausible therefore that the superior performance of
they could continue at the next stage. They were then informed
the animation group in the first experiment was due to learner
that after they had completed all three stages there would be a
control and/or less information provided in the static conditions.
revision stage in which they would be able to review all 13 steps
As a result of these considerations a number of conditions were
again in sequence. In the test phase, after the revision phase, par-
changed in Experiment 2. Both narration and learner control were
ticipants were asked to fold the complete model without reference
eliminated. The static graphics were re-designed to convey move-
to any instructional materials. The maximum time allocated for
ment more effectively by increasing the number of frames used per
each stage in the acquisition phase was 125 s for Stage 1, 75 s for
step. To avoid overwhelming the participants, individual frames
Stage 2, and 125 s for Stage 3. For the revision and test phases, par-
composing the steps were included or excluded based on pilot test-
ticipants were allocated 270 and 300 s to work through all three
ing results, which indicated whether they helped convey the
stages, respectively.
movement of the fold effectively when added to the existing group
of frames. To improve equivalency between conditions in terms of
6. Results and discussion the amount of information given, acquisition times were also stan-
dardized. To further generalise the results of Experiment 1 we
The total time taken to complete the three tasks during acquisi- chose a much younger group of students to work with (age 9–
tion was recorded, as was the time spent during the revision period 10). The instructional content chosen was learning to construct
(see Table 1). The number of completed tasks was also recorded an origami basket, which again required hand manipulations (hu-
during the test phase for each student (see Table 1). A one-way AN- man movement) and was novel for the participants used. Lastly,
OVA indicated that there was a significant difference between the the double static group of Experiment 1 was not used, as there
three groups for time in acquisition, F(2, 29) = 3.53, MSE = 9557.24, was no significant difference found between it and the single-static
p < 0.05, g2p = 0.20. Subsequent pair wise comparisons using New- group. Again, we tested the hypothesis that the animation group
man–Keuls post-hoc tests (a = 0.05) indicated that the animated would outperform the static group.
group spent less time in acquisition than the single-static group.
No other group differences were found. An ANOVA was also con-
ducted on time in the revision phase and indicated a significant dif- 8. Methods
ference between groups, F(2, 29) = 7.43, MSE = 2815.37, p < 0.05,
g2p = 0.34. A Newman–Keuls comparison (a = 0.05) revealed that 8.1. Participants and design
the animation group spent less time in revision than both the sin-
gle-static and double-static groups, which did not differ from each The participants were 26 children (16 female, 10 male) in year 3
other. For scores assessing how many students successfully com- and 4 with an age range of 8–9 years from a co-educational Sydney
pleted the paper-folding task during the test phase, a chi-squared primary school. Most of the students had a small amount of previ-
test revealed a between group difference, v2(2, N = 32) = 7.28, ous experience making origami figures, but none had previously
p < 0.05. Fisher Exact Probability tests indicated that more students learnt how to fold the origami basket used in the current experi-
ment. Participants were randomly assigned to one of two instruc-
tional conditions: animation (n = 15) or static graphics (n = 11).
Table 1 Initially, we began with 19, 8–10 year old children in each group
Mean times and standard deviations taken per group and number of participants that but random allocation resulted in unequal distributions of older
successfully completed the paper-folding task in the test phase across each condition
children (10 year olds) between groups. Removing all students
in Experiment 1.
who were 10 years or older (four from the animation group and
Group Time taken (s) Number of participants that eight from the static graphics group) equalized ages across groups.
completed tasks
Acquisition Revision Test
8.2. Materials
Animated 409 (141) 188 (81) 8
Single static 514 (42) 270 (0) 3
Double static 495 (83) 257 (41) 2 The animation was developed using the multimedia-authoring
program Macromedia FlashTM. It consisted of a series of 13 steps
344 A. Wong et al. / Computers in Human Behavior 25 (2009) 339–347

showing the origami basket being folded and lasted a total of 64 s. they had just seen. Participants in the animation condition were
Each fold consisted of between 6 and 15 individual frames from asked to study an animation on the computer monitor, which
start to finish, i.e., enough frames to obtain smooth movement. taught them how to fold a sheet of paper into an origami basket.
Over the entire animation sequence, there were 1032 frames. Total Participants in the static graphics condition were asked to study
time was established based on the results from pilot testing, i.e., a set of pictures on the computer monitor, which taught them
participants had sufficient time to watch the animation and still how to make the same origami basket. For the animation condi-
be able to successfully complete approximately 60% of the folds tion, the pace of the presentation was not under learner control.
when asked afterwards to perform the folding task. Those in the animation condition were shown the animation once
The static graphics were screenshots of individual frames cap- and those in the graphics condition were shown the graphics for
tured from the animation. The graphic design program Adobe Illus- 64 s. Participants were not allowed to practice folding the basket
tratorTM was used to compile the frames together into the final during this time. After watching the presentation, the testing phase
static graphics presentation. The frames chosen for inclusion into was administered. Participants were given an A4 sheet of paper
the static graphics were selected on the basis of which combina- and asked to fold the piece of origami they had just seen. After par-
tion of frames would make the fold clearest to the participant. ticipants had either finished the task or indicated they had folded
More specifically, individual frames were included or excluded as much as they could remember, the experimenter collected the
based on pilot testing results, which indicated whether they helped origami piece.
convey the movement of the fold effectively when added to the
existing group of frames. The presentation had a total of 13 steps
9. Results and discussion
and was shown on a single screen. Each step comprised of one
complete movement, e.g., folding a piece of paper in half and fold-
A t-test on the folding scores revealed a significant group differ-
ing it back out consisted of one step (see Fig. 3). We aimed to elim-
ence, t(24) = 2.58, p < 0.05, g2p = 0.22. Participants who received
inate redundancy but still effectively convey the movement at each
animation performed better (M = 6.60, SD = 3.96) than those who
step. Each step thus contained 3 to 4 frames (depending on the per-
received static graphics (M = 3.10, SD = 2.51). The results of this
ceptual complexity of the fold), showing the fold being made (with
experiment indicate that the animation group learnt more than
the exception of step 13). Step 13 contained 6 frames, as it was
the static group, thus supporting our hypothesis. The pattern of re-
found during pilot testing that it was perceptually the most diffi-
sults replicate those found in Experiment 1, suggesting that not
cult step to understand and that participants required more guid-
only was lack of informational equivalence not an issue, but that
ance. In total, there were 42 frames in the entire static graphics
the superiority of animation over static graphics remained when
sequence. The last frame of the previous fold was shown as the first
learner control and interactivity were removed.
frame of the next fold to avoid split attention effects, particularly at
the end of lines.
Both animation and static graphics conditions were presented 10. Experiment 3
on a computer to maintain equivalence between the two condi-
tions. Participants were presented a single sheet of A4 paper and Experiment 3 investigated the nature of the results obtained
tested on whether they could fold the entire origami basket from in Experiment 2 further, by exploring the hypothesis that anima-
start to finish. The test was scored out of 13 rather than a suc- tion is superior to static graphics when the task requires human
cess/fail score as in Experiment 1, as we wanted a more sensitive movement but that static graphics may be superior to animated
measure of performance. Each fold was counted as one mark. Half instructional aids for tasks that do not involve human move-
a mark was deducted from the mark allocated for the fold if a par- ment. Such a result would lend further support to the existence
ticipant made an incorrect fold, but then successfully corrected it, of a separate working memory processor devoted to human
as it indicated partial schema formation. Participants were given a movement. Again, a paper-folding task was used to distinguish
total score out of 13, according to the number of correct folds whether animation was superior, but in addition cognitive tasks
made. were also included, which did not involve manual manipulation.
For the cognitive tasks learners were required to recognise the
8.3. Procedure previous or next fold of the paper-folding task, given a particular
state of completion. Recognising the next fold was included as a
The students were presented with learning materials and tested control task, because although it could be classified as a non-
individually. The experiment consisted of consecutive learning and movement based task, it still involves recognising forward move-
test phases. Participants were randomly allocated to either the ani- ment and so could also activate the proposed WM human move-
mation or static graphics condition. They were asked if they had ment processor. In contrast, recognising the previous fold may
any previous experience in origami and paper-folding tasks and be less likely to activate our motor memory processor, as the
if so, were asked to recall what kinds of figures they had previously task requires inferences to be made in the opposite direction
folded. Any origami figures that participants had previously learnt to what occurs naturally in time. Consequently, it was predicted
to fold were noted. that there would be a group ! task interaction, where animation
After collecting the preliminary data, the learning phase began. would be superior for the actual folding task, but this advantage
All participants were told they would have 64 s to study the would disappear for purely cognitive tasks (in particular for the
instructions and that afterwards they would be asked to fold what backwards-based task).

Fig. 3. Example of a folding step and the individual frames that form it (static graphics condition) in Experiment 2.
A. Wong et al. / Computers in Human Behavior 25 (2009) 339–347 345

11. Methods not significant: t(22) = "0.33, p = 0.74, g2p = 0.005. Differences for
the next fold task were also not significant: t(22) = 0.79, p = 0.31,
11.1. Participants and design g2p = 0.04, but in contrast they favoured the animation group.
In summary, we found a significant effect favouring the anima-
The participants were 24 children (11 female, 13 male) in year 3 tion over the static graphics condition for a human movement-
and 4 (age range = 8–9 years old) from a co-educational Sydney based task replicating the results from Experiment 2. However, this
primary school. Most of the students had a small amount of previ- advantage did not carry over to non-movement based tasks, either
ous experience making origami figures, but none had previously in recognition of forward or backward steps.
learnt how to fold the origami basket used in the current experi-
ments. As with Experiment 2, participants were randomly assigned
to one of two instructional conditions: animation (n = 12) or static 13. General discussion
graphics (n = 12).
We have demonstrated that instructional animations that foster
11.2. Materials motor skills can indeed be superior to the equivalent static graph-
ics, consistent with the findings of Höffler and Leutner (2007). Our
The origami task and instructional materials were identical to results also lend support to the possible existence of a separate
Experiment 2. Testing included an additional set of 8 tasks requir- WM human movement processor. In particular, for paper-folding
ing participants to recognise the next or previous fold of the basket tasks requiring motor skills, animated instructional materials were
at varying stages of the folding task. Each task consisted of a target superior to static graphics, a pattern that we consistently demon-
(partially folded model), and four other partially folded models strated across the three experiments. In contrast, for the non-mo-
representing the next fold, the previous fold, two folds forward tor based recognition tasks (Experiment 3), animation was no
and two folds back of the target. Each target and its four accompa- longer superior to static graphics.
nying models were presented on a single sheet of cardboard. Four Both the paper folding and recognition tasks were from the
tasks required recognition of the previous fold to the target and same domain, however, the difference between the tasks focused
four questions required the next fold to be identified. One mark on whether they involved motor skills. Animations led to improved
was given for each correctly recognised fold, with a maximum pos- performance under conditions where people were required to
sible score of four marks for the two different types of tasks. reproduce human movement and thus invoke a hypothesised
‘movement’ based processor. These benefits did not carry over
11.3. Procedure for non-movement based tasks. It is interesting to note that these
effects persisted despite changes in many different variables, such
The procedure for Experiment 3 was identical to Experiment 2, as the folding task, number of static graphics, interactivity, learner
except for the addition of the cognitively based recognition task, control, inclusion of narration and different participant cohorts. In
which was administered after the folding task. Before the recogni- particular, the persistence of the effect indicates that the advantage
tion task was administered, the concepts of the next fold and the of animations over statics graphics for human movement-based
last fold were demonstrated to participants, using a part of the tasks may not necessarily be an artifact of introducing interactivity
folding task that was not tested in one of the questions. The partic- and learner control, but of our inherent ability to learn human-
ipants were also told that they had 10 s to respond. Questions for movement based tasks more efficiently through observation of
the recognition task were presented in random order. Participants the activity being performed. Results similar to ours were obtained
were shown the question and asked which was the last fold or the by Ayres, Marcus, Chan, and Qian (2009) using metal puzzles and
next fold, depending on the question. They were then given 10 s to constructing knots, both tasks that involve human movement.
respond, by pointing to the partially folded model that they Video based animations led to superior learning when compared
thought was the correct answer. Any responses taking longer than to a series of still frames for both the metal puzzle construction
10 s were coded as incorrect. task and the knot-tying task.
We acknowledge the current lack of neurophysiological evi-
12. Results and discussion dence demonstrating a relation between the mirror-neuron system
and WM. Nevertheless, the results are in accord with this putative
Table 2 shows the mean scores and standard deviations for the relation.
folding and fold recognition tasks. A 2 (animation vs static graph- Our results support the existing literature demonstrating that
ics) ! (3) ANOVA with repeated measures (folding task vs forward people can learn motor skills purely through observation, such as
recognition vs backward recognition) showed no significant main found by Blandin et al. (1999), who go further to suggest that when
effects but a significant interaction: F(2, 21) = 4.06, MSE = 4.19, we learn through observation, we also engage in cognitive pro-
p < 0.05, g2p = 0.279. Simple effects tests indicated that for the mo- cesses similar to those occurring during physical practice. Watch-
tor-based folding task performance in the animation condition was ing an animation appears to be tapping into these cognitive
superior to static graphics: t(22) = 2.36, p < 0.05, g2p = 0.20. While capabilities. The lack of benefits of animation over static graphics
differences between means on the last-fold task favoured the static for non-movement based tasks also suggests that the benefits of
graphics group over the animation group, those differences were observational learning do not necessarily transfer to non-motor
tasks that are more cognitive in nature. The results support those
of Iacoboni et al. (1999), who found that people asked to observe,
then execute a hand movement showed greater levels of cortical
Table 2
activation in fMRI scans when given an animated hand as a cue,
Group means (standard deviations) of folding, last and next fold tasks in Experiment
3. than when they were given either spatial or symbolic cues. Our re-
sults may also partly explain the inconsistent outcomes such as
Animation Static graphics
those found for animated materials in the mechanical, biological
Folding task 7.17 (4.30) 3.75 (2.57) and computational domains by Tversky et al. (2002). Over and
Last fold recognition 2.58 (1.08) 2.75 (1.35)
above any methodological shortcomings, mechanical, biological
Next fold recognition 3.17 (0.72) 2.83 (0.84)
and computational systems are typically biologically secondary
346 A. Wong et al. / Computers in Human Behavior 25 (2009) 339–347

domains of knowledge. They do not involve human movement. The satisfying the congruence and apprehension principles are neces-
biologically secondary nature of the subject matter means that sary but not sufficient to make animations more effective than
processing such information is effortful and we need to make use the equivalent static graphics.
of WM to do so. It is possible that in such cases, WM becomes con- Another possible alternative explanation is that the static
strained, due to the transitory nature of the animation, resulting in graphics may facilitate retention of stages, whereas animations
less effective learning, relative to static graphics-based materials. may facilitate human movement-based learning (Zacks & Tversky,
We believe it is telling that Mayer et al. (2005) in a series of 2001). Due to the spatial nature of static graphics, participants gi-
experiments obtained results equally as consistent as ours but in ven graphics may have learned the paper-folding task in stages,
the reverse direction. Static graphics consistently proved superior reinforcing the learning of the procedure as steps, but conse-
to animations. We suggest that the contrasting results are due to quently leading to less fluid performance. Conversely, the fluid nat-
Mayer et al. (2005) using mechanical object movement that did ure of the animation may render the steps irrelevant when
not require human manipulation and that the transitory nature learning, leading to better performance than the static graphics.
of animations imposed an excessive working memory load that If this were the case, we would expect the static graphics condition
interfered with learning. We suggest that load may be ameliorated to perform as well or better than the animation condition in the
where learners are able to use the hypothesised human movement next fold recognition task in Experiment 3. The next fold task,
processor, a processor that is likely to be irrelevant when dealing which asked participants to recognise the next fold in the paper-
with mechanical object movement. folding procedure, corresponds to fine-grain, bottom-up small
Although the paper-folding task involved human hand move- event boundaries (Zacks & Tversky, 2001). We instead found the
ments, we did not show hands explicitly. If, as we assume, there animation condition still outperformed the static graphics condi-
is a human movement processor, it should not be necessary to tion. As a result, we also doubt this explanation is viable.
explicitly show hands because the relevant hand movements In terms of implications for instructional design, the results
should be implicitly assumed by learners as part of biologically pri- from the current set of experiments indicate that learning some
mary knowledge (see also Van Gog et al., 2007, for a discussion of forms of biologically secondary knowledge (specifically, human
the necessity of the presence of a human actor). Analogously, we movement-based tasks) may be aided by the use of animated
learn tongue movements when learning a first language without instructional materials. The benefits of the animation for the learn-
either explicit tuition or even vision of other peoples’ tongues. ing of motor tasks is most likely directly related to our hard wired
We intentionally did not show hands performing the folds, as abilities to learn by imitating. Geary (2002, 2005, 2007) suggests
we were concerned that the inclusion of hands may obstruct the that tasks can consist of both biologically primary and secondary
participant’s view of the movement. Michas and Berry (2000) per- elements, writing being such an example. Whilst hand movement
formed an experiment comparing learning outcomes from video is biologically primary, the specific motor movements that produce
stills and continuous video showing a first aid person applying letters are biologically secondary. Thus, teaching handwriting
bandages. They showed hand movements and obtained a similar using animations is likely to be more beneficial than using a series
pattern of results to the current set of experiments, where partic- of static graphics (see Marcus, Lu, Kam, & Ayres, 2006). A possible
ipants given the video performed better than in the video stills future direction of research may be to develop other instructional
condition for a real bandaging procedure and for questions relating design techniques that bridge biologically secondary learning
to the steps in the bandaging procedure. material by tapping into the appropriate biologically primary
It could be argued that the results obtained are an artifact of knowledge. Conversely, presenting materials or tasks that are not
procedural and conceptual knowledge, rather than differences be- inherently based around human movement in animated formats
tween biologically primary and secondary knowledge. We doubt may result in less effective learning than if the same material or
this possible explanation. While tasks involving human movement task was to be presented as static graphics.
are clearly procedural, the non-movement tasks of Experiment 3, We reiterate that the human movement WM processor is
while not requiring physical movement, do require the manipula- hypothesised. Our results are compatible with such a construct
tion of a cognitive procedure and in that sense, are procedural. Fur- but further research is needed to firmly establish the existence of
thermore, as indicated above, the mechanical, biological and a separate WM movement processor. In particular, human move-
computational tasks that demonstrated an advantage of static ment based instructional materials need to be directly compared
graphics over animation in other experiments clearly were proce- to mechanical movement based instructions, using a larger num-
dural in nature making it unlikely that our results are due to a pro- ber of participants. More data also needs to be collected using dif-
cedural-conceptual distinction. ferent motor skills.
As is the case for all experiments concerned with cognitive pro-
cesses and instructional design, we acknowledge that there may be
References
alternative explanations for the current set of results. Tversky et al.
(2002) suggested that animations might be effective if the external Ayres, P. (2006a). Using subjective measures to detect variations of intrinsic
and internal representations of both structure and content are cognitive load within problems. Learning and Instruction, 16, 389–400.
commensurate with each other (congruence principle). That is, a Ayres, P. (2006b). Impact of reducing intrinsic cognitive load on learning in a
mathematical domain. Applied Cognitive Psychology, 20, 287–298.
natural correspondence needs to exist between the conceptual Ayres, P., Marcus, N., Chan, C., & Qian, N. (2009). Learning hand manipulative tasks:
information, the change over time and the basis of the animation. When instructional animations are superior to equivalent static represen-
So long as the animated instructional material can be naturally tations. Computers in Human Behavior, 25, 348–353.
Ayres, P., & Paas, F. (2007a). Making instructional animations more effective: A
construed as an animation, then the congruence principle should cognitive load approach. Applied Cognitive Psychology, 21, 695–700.
be satisfied. It is possible that our results can be explained by a cor- Ayres, P., & Paas, F. (2007b). Can the cognitive load approach make instructional
respondence between the internal representation (the cognitive animations more effective? Applied Cognitive Psychology, 21, 811–820.
Baddeley, A. (1992). Working memory. Science, 255, 556–559.
process of folding the piece of paper) and the external representa-
Blandin, Y., Lhuisset, L., & Proteau, L. (1999). Cognitive processes underlying
tion (the physical act of folding the piece of paper). However, we observational learning of motor skills. Quarterly Journal of Experimental
believe the congruence principle is primarily descriptive and fails Psychology, 52A, 957–979.
Chase, W. G., & Simon, H. A. (1973). Perception in chess. Cognitive Psychology, 4,
to sufficiently explain the emerging evidence that animation is
55–81.
more effective than static graphics under specific conditions (e.g., Clark, R., Nguyen, F., & Sweller, J. (2006). Efficiency in learning: Evidence-based
Höffler & Leutner, 2007). Further, Tversky et al. (2002) note that guidelines to manage cognitive load. San Fransisco: Pfeiffer.
A. Wong et al. / Computers in Human Behavior 25 (2009) 339–347 347

De Groot, A. (1965). Thought and choice in chess. The Hague, Netherlands: Mouton Paas, F., & Van Merriënboer, J. J. G. (1993). The efficiency of instructional conditions:
[Original work published 1946]. An approach to combine mental-effort and performance measures. Human
Ericsson, K. A., & Kintsch, W. (1995). Long-term working memory. Psychological Factors, 35, 737–743.
Review, 102, 211–245. Penney, C. (1989). Modality effects and the structure of short-term verbal memory.
Geary, D. (2002). Principles of evolutionary educational psychology. Learning and Memory and Cognition, 17, 398–422.
Individual Differences, 12, 317–345. Peterson, L., & Peterson, M. (1959). Short-term retention of individual verbal items.
Geary, D. (2005). The origin of mind: Evolution of brain, cognition, and general Journal of Experimental Psychology, 58, 193–198.
intelligence. Washington, DC: American Psychological Association. Schneider, W., & Shiffrin, R. (1977). Controlled and automatic human
Geary, D. (2007). Educating the evolved mind: Conceptual foundations for an information processing: I. Detection, search and attention. Psychological
evolutionary educational psychology. In J. S. Carlson & J. R. Levin (Eds.), Review, 84, 1–66.
Psychological perspectives on contemporary educational issues (pp. 1–99). Schnotz, W., Böckheler, J., & Grzondziel, H. (1999). Individual and cooperative
Greenwich, CT: Information Age Publishing. learning with interactive animated pictures. European Journal of Psychology and
Hasler, B. S., Kersten, B., & Sweller, J. (2007). Learner control, cognitive load and Education, 14, 245–265.
instructional animation. Applied Cognitive Psychology, 21, 713–729. Schwan, S., & Riempp, R. (2004). The cognitive benefits of interactive videos:
Hegarty, M., Kriz, S., & Cate, C. (2003). The role of mental animations and external Learning to tie nautical knots. Learning and Instruction, 14, 293–305.
animations in understanding mechanical systems. Cognition and Instruction, 21, Shiffrin, R., & Schneider, W. (1977). Controlled and automatic human information
325–360. processing: II. Perceptual learning, automatic attending, and a general theory.
Höffler, T. N., & Leutner, D. (2007). Instructional animation versus static pictures: A Psychological Review, 84, 127–190.
meta-analysis. Learning and Instruction, 17, 722–738. Sweller, J. (1999). Instructional design in technical areas. Melbourne: ACER.
Iacoboni, M., Woods, R., Brass, M., Bekkering, H., Mazziotta, J., & Rizzolatti, G. (1999). Sweller, J. (2003). Evolution of human cognitive architecture. In B. Ross (Ed.). The
Cortical mechanisms of human imitation. Science, 286, 2526–2528. psychology of learning and motivation (vol. 43, pp. 215–266). San Diego:
Lowe, R. K. (1999). Extracting information from an animation during complex visual Academic Press.
learning. European Journal of Psychology of Education, 14, 225–244. Sweller, J. (2004). Instructional design consequences of an analogy between
Lowe, R. K. (2003). Animation and learning: Selective processing of information in evolution by natural selection and human cognitive architecture. Instructional
dynamic graphics. Learning and Instruction, 13, 157–176. Science, 32, 9–31.
Marcus, N., Lu, P., Kam, B., & Ayres, P. (2006). Animations with a trace lead to more Sweller, J. (2005). Implications of cognitive load theory for multimedia learning. In
effective learning than equivalent static graphics. Paper presented as part of an R. E. Mayer (Ed.), Cambridge Handbook of Multimedia Learning (pp. 19–30). New
effective learning from animations symposium at the annual meeting of the York: Cambridge University Press.
American education research association, San Francisco. Sweller, J., & Sweller, S. (2006). Natural information processing systems.
Mayer, R. E., & Chandler, P. (2001). When learning is just a click away: Does simple Evolutionary Psychology, 4, 434–458.
user interaction foster deeper understanding of multimedia messages? Journal Sweller, J., van Merriënboer, J. J. G., & Paas, F. (1998). Cognitive architecture and
of Educational Psychology, 93, 390–397. instructional design. Educational Psychology Review, 10, 251–296.
Mayer, R. E., DeLeeuw, K. E., & Ayres, P. (2007). Creating retroactive and proactive Tettamanti, M., Buccino, G., Saccuman, M. C., Gallese, V., Danna, M., Scifo, P., et al.
interference in multimedia learning. Applied Cognitive Psychology, 21, 795–809. (2005). Listening to action-related sentences activates fronto-parietal motor
Mayer, R. E., Hegarty, M., Mayer, S., & Campbell, J. (2005). When static media promote circuits. Journal of Cognitive Neuroscience, 17, 273–281.
active learning: Annotated illustrations versus narrated animations in Tversky, B., Morrison, J. B., & Betrancourt, M. (2002). Animation: Can it facilitate?
multimedia instruction. Journal of Experimental Psychology: Applied, 11, 256–265. International Journal of Human-Computer Studies, 57, 247–262.
Meltzoff, A., & Prinz, W. (2002). The imitative mind: Development, evolution, and brain Van Gog, T., Paas, F., Marcus, N., Ayres, P., & Sweller, J. (in press). The mirror-neuron
bases. Cambridge, UK: Cambridge University Press. system and observational learning: Implications for the effectiveness of
Michas, I. C., & Berry, D. C. (2000). Learning a procedural task: Effectiveness of dynamic visualizations. Educational Psychology Review.
multimedia presentations. Applied Cognitive Psychology, 14, 555–575. van Merriënboer, J. J. G., & Ayres, P. (2005). Research on cognitive load theory and its
Miller, G. A. (1956). The magical number seven, plus or minus two: Some limits on design implications for E-learning, Educational Technology. Research and
our capacity for processing information. Psychological Review, 63, 81–97. Development, 53(3), 5–13.
Moreno, R. (2007). Optimising learning from animations by minimizing cognitive van Merriënboer, J. J. G., & Sweller, J. (2005). Cognitive load theory and complex
load: Cognitive and affective consequences of signaling and segmentation learning: Recent developments and future directions. Educational Psychology
methods. Applied Cognitive Psychology, 21, 765–781. Review, 17, 147–177.
Paas, F., Renkl, A., & Sweller, J. (2003). Cognitive load theory and instructional Zacks, J., & Tversky, B. (2001). Event structure in perception and cognition.
design: Recent developments. Educational Psychologist, 38, 1–4. Psychological Bulletin, 127, 3–21.

View publication stats

You might also like