Cognitive Load Theory in E-Learning

Abstract

Cognitive load can be assessed and monitored using a multitude of subjective (self-reports, i.e. Hart & Staveland, 1988; Paas, 1992) and more objective methods (dual tasks, eye-tracking, heart-rate measurements, skin conductance measurements, cf. Brünken, Plass, & Leutner, 2003; Beatty, 1982, Paas, van Merriënboer, & Adam, 1994), either during the learning or afterwards, so that instruction can be optimized based on mental effort data using iterative design (a cyclic process of prototyping, testing, analyzing, and refining a product or process, ultimately improving the quality and functionality of the design). Computer simulations provide an excellent environment to apply CLT principles. However, such e-environments are technically complex and therefore add to extraneous load. Separating the technical knowledge of how to use the computer interface from the actual conceptual knowledge using sequencing should reduce this load to a reasonable extent (cf. Clarke, Ayres & Sweller, 2006). The authors provide guidelines on how to use CLT in the design of e-environments and discuss what future directions can be taken to further optimize the design of such environments.

Full text

Zheng Yan
University at Albany, USA
Encyclopedia of
Cyber Behavior
Volume I

Encyclopedia of cyber behavior / Zheng Yan, editor.
p. cm.
Includes bibliographical references and index.
Summary: “This book offers a complete look into the field of cyber behavior, surveying case studies, research, frameworks,
techniques, technologies, and future developments relating to the way people interact and behave online”-- Provided by
publisher.
ISBN 978-1-4666-0315-8 (hardcover) -- ISBN 978-1-4666-0316-5 (ebook) -- ISBN 978-1-4666-0317-2 (print & perpetual
access) 1. Internet--Psychological aspects. 2. Computer networks--Psychological aspects. I. Yan, Zheng.
BF637.C45.E53 2010
302.23’1--dc23
2011044984
British Cataloguing in Publication Data
A Cataloguing in Publication record for this book is available from the British Library.
All work contributed to this book is new, previously-unpublished material. The views expressed in this book are those of the
authors, but not necessarily of the publisher.
Managing Director: Lindsay Johnston
Senior Editorial Director: Heather A. Probst
Book Production Manager: Sean Woznicki
Development Manager: Joel Gamon
Acquisitions Editor: Erika Gallagher
Typesetter: Jennifer Romanchak, Nicole Sparano
Cover Design: Nick Newcomer, Lisandro Gonzalez
Published in the United States of America by
Information Science Reference (an imprint of IGI Global)
701 E. Chocolate Avenue
Hershey PA 17033
Tel: 717-533-8845
Fax: 717-533-8661
E-mail: cust@igi-global.com
Web site: http://www.igi-global.com
Copyright © 2012 by IGI Global. All rights reserved. No part of this publication may be reproduced, stored or distributed in
any form or by any means, electronic or mechanical, including photocopying, without written permission from the publisher.
Product or company names used in this set are for identification purposes only. Inclusion of the names of the products or
companies does not indicate a claim of ownership by IGI Global of the trademark or registered trademark.
Library of Congress Cataloging-in-Publication Data

1178
Copyright © 2012, IGI Global. Copying or distributing in print or electronic forms without written permission of IGI Global is prohibited.
Chapter 97
Christa M. van Mierlo
Open University of the Netherlands, The Netherlands
Halszka Jarodzka
Open University of the Netherlands, The Netherlands
Femke Kirschner
Erasmus University Rotterdam, The Netherlands
Paul A. Kirschner
Open University of the Netherlands, The Netherlands
Cognitive Load Theory
in E-Learning
ABSTRACT
Cognitive load can be assessed and monitored using a multitude of subjective (self-reports, i.e. Hart &
Staveland, 1988; Paas, 1992) and more objective methods (dual tasks, eye-tracking, heart-rate mea-
surements, skin conductance measurements, cf. Brünken, Plass, & Leutner, 2003; Beatty, 1982, Paas,
van Merriënboer, & Adam, 1994), either during the learning or afterwards, so that instruction can be
optimized based on mental effort data using iterative design (a cyclic process of prototyping, testing,
analyzing, and rening a product or process, ultimately improving the quality and functionality of the
design). Computer simulations provide an excellent environment to apply CLT principles. However, such
e-environments are technically complex and therefore add to extraneous load. Separating the technical
knowledge of how to use the computer interface from the actual conceptual knowledge using sequencing
should reduce this load to a reasonable extent (cf. Clarke, Ayres & Sweller, 2006). The authors provide
guidelines on how to use CLT in the design of e-environments and discuss what future directions can be
taken to further optimize the design of such environments.
INTRODUCTION
There exist 3 major areas of research regarding
CLT: how to lower load on working memory, how
to stimulate the consolidation of new information
in mental schemata in long-term memory and how
to measure cognitive load for design purposes.
CLT effectively deals with the limitations that
are induced by working memory by creating in-
structions that lower the intrinsic (content-based),
extraneous (presentation-based) and germane
(information consolidation-based) cognitive load
on working memory (Chandler & Sweller, 1991;
DOI: 10.4018/978-1-4666-0315-8.ch097

1179
Cognitive Load Theory in E-Learning
Sweller, Van Merriënboer & Paas, 1998; Kirsch-
ner, 2002, Van Merriënboer, Kirschner, & Kester,
2003). This can be done by dividing the whole
task in separate learning tasks that are whole-task
experiences but increase in difficulty as experience
is gained (4C/ID; Van Merriënboer, 1997), by us-
ing worked-examples (Paas & van Merriënboer,
1994; Sweller & Cooper, 1985), avoiding splits
of attention (Chandler & Sweller, 1991), avoid-
ing redundancy (Chandler & Sweller, 1991), and
using multimedia instead of only one modality
(Mayer, 1997).
Consolidation of new knowledge in schemata
in long-term memory can be facilitated by e.g.
using scaffolding and fading support (Van Mer-
rienboer, Clark & de Croock, 2002), just-in-time
information presentation (Kester, Kirschner,
van Merriënboer, 2001), and/or using increas-
ingly incomplete examples that learners have to
complete so that the different steps of which the
mental model should consist are made explicit
(Van Merriënboer & Kramer, 1990).
Cognitive load theory (CLT, Sweller, Van Mer-
riënboer, & Paas, 1998) is a theoretical framework
of learning based on human cognitive architecture.
It assumes that learning is constrained by the lim-
ited processing capacity of the learner’s cognitive
architecture. The cognitive capacity a person has
available to actively process and store informa-
tion is limited to between 7±2 familiar elements
(Miller, 1956) and 4±1 elements if they are novel
(Cowan, 2001). As a consequence, instruction that
places a high demand on this capacity, either by
including too much or too complex information,
or by presenting it in ways that do not contribute
to learning or even hamper learning, reduces the
acquisition of new knowledge or skill. The aim of
researchers in the field of CLT has, therefore, been
to develop techniques to manage the cognitive load
(CL) imposed by a learning task in order to facili-
tate learning. Because user interfaces in E-learning
are often technologically complex compared to
the traditional paper-based materials (e.g., use of
multi-media, multitasks or multi-facetted tasks,
and/or an often non-linear information organiza-
tion), the instructional principles derived from CLT
are particularly relevant and useful for designing
effective E-learning environments. Fortunately,
E-learning environments can accommodate the
majority of CLT principles in relatively simple
ways (Jochems, Van Merriënboer, & Koper, 2004).
This chapter first introduces CLT and its main
players, second we will elaborate on how to mea-
sure CLT, third we will discuss the main effects
of CL and how to deal with them in e-learning,
fourth we will discuss the 4C/ID model that offers
a standardized approach on how to design new
e-learning materials while taking the effects of
CL into account, and finally the chapter end with
our perspective on future approaches that can be
taken in CLT and e-learning.
COGNITIVE LOAD THEORY
CLT describes learning of complex cognitive tasks,
in which the number of interactive information
elements that need to be processed simultaneously
before meaningful learning can commence often
overwhelms learners. According to CLT, indi-
vidual learning depends on the limited processing
capacity of the learner’s cognitive architecture
and the CL imposed by a task. According to the
Atkinson and Shiffrin model (1971) the cognitive
architecture consists of an effectively unlimited
long-term memory (LTM), which interacts with a
working memory (WM) that is very limited in both
capacity (Baddeley & Hitch, 1974; Cowan, 2001;
Miller, 1956) and duration (Peterson & Peterson,
1959). For new, yet to be learned information, pro-
cessing capacity is limited to only 4±1 information
elements, and if not rehearsed, the information
is lost within 30 seconds (Cowan, 2001). LTM
consists of cognitive schemata schemas (Van
Lehn, 1996) that store and organize knowledge
by incorporating multiple elements of informa-
tion into a single element (also referred to as a
chunk; Miller, 1956) with a specific function (i.e.,

1180
Cognitive Load Theory in E-Learning
learning). Long-term learning processes result
in schemas that can represent vast quantities of
information (e.g., a whole chess game). Because a
schema can be treated by WM as a single element
or even bypass WM if it has become sufficiently
automated after long and consistent practice, the
limitations of WM disappear for more experienced
task performers when they deal with previously
learned information stored in LTM. Hence, the
main area of interest lies in the events in WM.
According to CLT there are three types of CL
that learners can experience in WM (Sweller et
al., 1998). The load is considered to be ‘intrinsic’
if it is imposed by the number of information
elements in a task and the interactivity between
those elements (i.e., task complexity). The more
elements within a task and the more interac-
tion between them, the more complex the task
and the higher the intrinsic cognitive load will
be. When the load is imposed by the manner in
which the information is presented to learners
and by the learning activities required of them,
it is called either ‘extraneous’ or ‘germane’ CL.
Extraneous load is imposed by information and
activities that do not directly contribute to learning
(e.g., difficulties searching for the information in
discovery-based learning), while germane load
is caused by information and activities that foster
learning processes (e.g., elaborating on informa-
tion). Intrinsic, extraneous, and germane CL are
often considered additive in that, when summed,
the total load cannot exceed the total WM capac-
ity that is available to the student if learning is
to occur (see, Paas, Tuovinen, Tabbers, & Van
Gerven, 2003).
The relations between the three forms of CL
are asymmetric. Intrinsic load provides a ‘base’
load that is irreducible by instructional design. It
can only be lowered by constructing additional
schemata and automating previously acquired
schemata; in other words, by an increase in
expertise or by deconstructing the task so that
less elements interact (see Pollock, Chandler, &
Sweller, 2002).
Any available WM capacity remaining after
resources have been allocated to deal with intrinsic
load is allocated to extraneous or germane load.
Manipulations of the different loads can work
in tandem in that, for example, a reduction in
extraneous load by using a more effective instruc-
tional design can free capacity for an increase in
germane load on the basis of which new cognitive
schemata can be formed, resulting in a reduction
in intrinsic load and freeing of WM capacity for
using the newly learned material (i.e., the newly
acquired schemata) to acquire more advanced
schemata. A new cycle, thus, commences and
over many cycles, very advanced knowledge and
skills may be acquired.
Instructional control of overwhelming CL in
learning has become a major focus of CLT. In
the last ten years CLT has undergone important
developments driven both by theoretical progress
and changes in the field of instructional design.
One significant change is the shifting focus from
printed materials to online learning tasks (Van Mer-
riënboer & Ayres, 2005). E-learning tasks often use
2- and 3D graphics, audio narration, animations,
background colors and/or interactions of these
possibilities. Additionally, the information might
not be linearly organized, because most e-learning
environments allow learners to navigate freely
through the available information, with or without
the aid of hyperlinks and graphical organizers. This
increases the extraneous load, because, consider-
ing the way information is normally organized
in text books, which that is stored in long-term
schemata as a heuristic on how to deal with any
new information in the future, the organization of
the e-environment does not comply with learner’s
expectations about how the information in the
e-learning environment is organized. This results
in learners’ cognitive resources being allocated
to searching for relevant information and, while
doing so, forcing them to inadvertently to process
irrelevant information, thus indirectly influenc-
ing the amount of germane load the learner can
allocate to the task. As a result, the CL imposed

1181
Cognitive Load Theory in E-Learning
by e-learning environments may be too high for
novices, and could seriously hamper learning.
Several studies investigated the effects of
different instructional manipulations to decrease
extraneous CL, increase germane load and/or man-
age intrinsic load (i.e., optimize CL) on learning
outcomes. The manipulations that lead to better
schema construction and higher performance for
novice learners in individual and collaborative
e-learning environments are discussed in section
3.1. Before that we will discuss in the following
section how CL can be captured.
MEASURING COGNITIVE LOAD
According to Paas and Van Merriënboer (1994)
cognitive load is a multidimensional construct
representing the load that is imposed on a learner’s
cognitive system when executing a certain task. It
can be conceptualized in two dimensions: a task-
based dimension (i.e., mental load) which refers
to the load imposed by the task itself including
for instance its content complexity, and a learner-
based dimension (i.e., mental effort) which refers
to the amount of cognitive capacity a learner actu-
ally allocates to accommodate the task demands.
Both dimensions affect learner performance (for
a detailed account of the concept of CL see Paas
& Van Merriënboer, 1994).
Thus, CL needs to be assessed by measuring
mental load, mental effort, and performance. The
most commonly used measurement techniques are
designed to measure mental effort and its relation
to performance. Note that mental effort refers to
“internal processes of information processing that
cannot be observed directly” (Brüncken, Plass,
& Leutner, 2003; p. 55). Hence, when study-
ing e-learning environments with respect to the
amount of CL they impose on a learner, inven-
tive measuring techniques must be used. Over
the years three major categories of mental effort
measurement techniques have been developed:
subjective measures, physiological measures, and
task-and performance-based measures (Wierwille
& Eggemeier, 1993). In the following paragraphs
each category will be briefly described. For in
depth discussions of techniques for measuring
CL see Brünken, Plass, and Leutner (2003), Paas,
Ayres, and Pachman (2008) and Sweller, Ayres,
and Kalyuga (2011).
Subjective Techniques
When using subjective techniques to measure
mental effort, one must assume that people can
reliably monitor and introspect on their cogni-
tive processes and report the amount of invested
cognitive effort. Some research has shown that
people are capable of assigning numerical values
to their perceived mental effort (e.g., Gopher
& Braune, 1984; Paas, 1992), although others
have shown that this numerical estimate is only
of ordinal scale and depends on both individual
characteristics and context (for a review on this see
Annett, 2002). Subjective techniques are usually
implemented by means of multi- or unidimensional
rating scales (e.g., Hart & Staveland, 1988; Paas,
1992) on which participants can indicate their
experienced level of mental effort. While multi-
dimensional measures usually take place after
the learner has executed a sequence of learning
tasks are (NASA-TLX by Hart & Staveland,
1988), unidimensional measurements can also
be sampled during the sequence (i.e., after each
subtask; Paas, 1992), resulting in a process mea-
sure. This process measure is usually converted
into an ‘efficiency-measure’ taking into account
the amount of mental effort invested in learning a
task, the amount of mental effort perceived while
being tested on this task and the learning result (i.e.,
performance) (Paas & Van Merriënboer, 1993;
Van Gog & Paas, 2008). This technique has been
widely used in research on educational psychology
(for a review see Paas, Touvien, Tabbers, & Van
Gerven, 2003). Furthermore, this measure has
lately been shown to not only capture individual
mental effort, but also the amount of mental effort

1182
Cognitive Load Theory in E-Learning
invested by a group of people (Kirschner, Paas,
& Kirschner, 2009a). The NASA-TLX, on the
other hand, captures more variety of the concept
‘working load’, namely mental demand, physical
demand, temporal demand, performance, effort,
and frustration level.
The drawback of subjective measures is that
one has to assume that people are able (i.e., they
have to remember what level of mental effort
they experienced when, which can be a difficult
task in complex situations) and willing to monitor
and report the amount of mental effort that they
invested during task performance. Moreover,
the measure by Paas (1992) captures the overall
amount of load, only when this load (invested in
the learning phase) is compared to the performance
scores in the test phase conclusions about the type
of CL invested can be made. That is, when load
is high and performance is as well, it is assumed
that the load was germane. When the load is high,
but performance low, the load perceived during
learning must have been extraneous. This post-hoc
reasoning has been criticized as it gives no direct
insight into which type of load (i.e., extraneous,
intrinsic, or germane) was demanding the learner’s
cognitive capacities (e.g., Brünken, Plass, &
Leutner, 2003; de Jong, 2009), in contrast, for
instance to the NASA technique. Consequently,
researchers have been trying to investigate the
possibility to measure the three different types of
CL directly (e.g., Scheiter, Gerjets, & Catrambone,
2006). However, these techniques are not widely
adopted yet, because they have not been proven to
be reliable enough. Another criticism of the Paas
(1992) measure is that it is not standardized (i.e.,
one cannot tell from the scale scores, whether
the amount of perceived mental effort is high or
low; de Jong, 2009). Hence, it is very difficult
to compare empirical results using this measure
across different studies. Even when taking into
account that this measure is an indicator of dif-
ferent load types depending on the scenario, the
efficiency measures that are calculated based on
this subjective measure could be standardized,
but in most of the studies this is not done either.
It is important to know the origin of the Paas-
scale to better understand the reasoning and cri-
tique based on it. The Paas scale is modeled on
a similar scale that is much used in the field of
physical exertion: Borg’s (1977, 1982) Rating of
Perceived Exertion. This 20-point subjective scale
ranges from ‘no exertion’ (approximately 60 heart
beats per minute) to ‘maximum exertion’ (more
than 200 beats per minute) has been shown to be
extremely reliable. The problem of an analogous
subjective mental effort scale is that, in contrast
to a physical one, a rater has no corroborating
cues as to her/his feelings of effort. For physical
exertion, respondents have a number of other
benchmarks or reference points with which to
corroborate their response (i.e., heart rate, level of
perspiration, feelings of fatigue or muscle strain,
et cetera). For mental effort or exertion, however,
no such corresponding measurable or experiential
reference points exist. Hence, one may question
the ability of people to estimate their perceived
mental effort in a similar way.
Finally, researchers disagree on the interpreta-
tion of empirical results in terms of the validity,
reliability, and sensitivity of this measure (cf.
Paas, Van Merriënboer, & Adam, 1994 versus
de Jong, 2009).
Physiological Techniques
A more objective way to capture the amount of
mental effort invested by learners is by means
of physiological techniques. These techniques
assume that changes in cognitive states evoke
physiological changes. Examples of such mea-
sures are the heart rate (Paas & Van Merriënboer,
1994), muscle response (Leyman, Mirka, Kaber,
& Sommerich, 2004), or MRI (Tomasio, Chang,
Caparellia, & Ernst, 2007). The advantage of
these techniques is that they visualize the detailed
trend and pattern of load (i.e., peak, average and
accumulated load). On the other hand, these

1183
Cognitive Load Theory in E-Learning
techniques are uncomfortable for the participants
as they require intrusive instruments, such as
wires to an electrocardiogram R-wave toptrigger
(heart rate; Paas & Merriënboer, 1994) or surface
electrodes attached to the skin (muscle response;
Leyman, et al., 2004). This discomfort seems not
to be justified as, for instance, the heart-rate mea-
sure has shown to provide similar and even less
sensitive results as the subjective scale of Paas,
even if more information can be obtained (Paas,
1992; Paas & Van Merriënboer, 1994). Another
physiological technique to capture mental effort
that is less intrusive is eye tracking (Holmqvist,
Nyström, Andersson, Dewhurst, Jarodzka, & Van
der Weijer, 2011; Jarodzka, Janssen, Kirschner,
& Erkens, 2011; Van Gog, Kester, Nievelstein,
Giesbers, & Paas, 2009). Eye-tracking parameters
such as fixation duration (Underwood, Jebbett, &
Roberts, 2004) or pupil dilation (Beatty, 1982; Gra-
nholm, Asarnow, Sarkin, & Dykes, 1996; Hyönä,
Tommola, & Alaja, 1995; Klingner, Tversky, &
Hanrahan, 2011) have been shown to increase
with increasing mental effort. Besides being less
intrusive, eye tracking measures easily allow to
relate certain parts of the task to a specific amount
of mental effort.
Physiological measures allow for a real-time
process measure of mental effort; giving insight
in how mental effort changes over time when the
learner carries out a task. A drawback of physi-
ological measures is that they are both difficult and
time-consuming to analyze because there are no
standardized values for a person’s effort, but rather
each person is his or her own baseline (i.e., the
measures are idiosyncratic; for eye tracking data
see Holmqvist et al., 2011). Furthermore, physi-
ological data are ‘noisy’ and require the researcher
to first filter the data which is an error-prone
process requiring a large amount of training and
knowledge on both the underlying physiological
events and technical properties of the given system
itself. Finally, the values obtained are influenced by
many more factors than mental effort alone (e.g.,
changes in brightness of the presented material
van influence pupil dilation). As a consequence,
there is no direct link between the obtained values
and the amount of mental effort invested (see also
Brünken, Steinbacher, Plass, & Leutner, 2002).
This means that eye tracking preferably needs to
be combined with other mental effort measures
and the experiment needs to be set-up in such a
way that it ensures that the trends seen are really
related to the underlying cognitive processes and
not to irrelevant but confounding factors.
Task- and Performance-
Based Techniques
A third type of mental effort measurement tech-
nique is the task- and performance-based approach.
This can either be the performance on the learned
task itself or performance on a secondary task
performed simultaneously with the primary task
(cf. Paas, Touvien, Tabbers, & Van Gerver, 2003);
sometimes referred to as a dual task. In the latter,
the primary task - which is the actual learning task
(e.g., performing a learning task) - is accompa-
nied by a secondary, non-disruptive and usually
non-cognitive task (e.g., tapping a certain rhythm
with one’s foot). Research has shown that with
increasing mental effort devoted to the primary
task, performance on the secondary task decreases
(Brünken, Plass, & Leutner, 2003; Chandler &
Sweller, 1996; Marcus, Cooper, & Sweller, 1996;
Sweller, 1988). Other examples of a seconday task
are detection of a visual stimulus (Brünken, Plass,
& Leutner, 2003; DeLeeuw & Mayer, 2008) or an
audio stimulus (Brünken, Plass, & Leutner, 2004),
or a simple recall task (Ayres, 2001; Chandler &
Sweller, 1996). In the former case, the stimulus
reaction time will give an indication of mental
effort (i.e., the higher the load, the greater the
reaction time), while in the latter, recall complete-
ness is an indicator for induced mental effort (i.e.,
the higher the load, the less complete the recall).
An example of the use of a secondary task was
implemented by Brünken and colleagues (Brünken
& Leutner, 2001; Brünken, Steinbacher, Plass,

1184
Cognitive Load Theory in E-Learning
Leutner, 2002; Brünken, Steinbacher, Schnotz,
& Leutner, 2001). In a series of experiments, the
effect of the design of a computer-based multi-
media learning environment on mental effort was
investigated. Participants were required to react
to a small change in a stimulus presented above
the actual learning material screen (either only
visual or audio-visual). Their results show that
reaction times to these changes were longer when
the design of the environment did not make use
of multiple modalities; thus when mental effort
increased. Another form of a dual-task may be
thinking aloud; asking participants to verbalize
what they are thinking without changing the
thought process itself (Ericsson & Simon, 1993).
This method allows for indications of invested
mental effort because verbalization of cognitive
processes may impose a CL itself. Consequently,
when participants’ cognitive capacity is used up
for the primary task, they are not able to keep
on thinking aloud, resulting in a silence that is
a clear inidcator for high mental effort (e.g., Yin
& Chen, 2007). Moreover, this method may also
yield interesting insights into the type of mental
effort induced when learners’ make utterances
hinting towards it (Jarodzka et al., 2011).
The major advantage of dual-task measures is
that they provide a direct estimate of mental effort
whilst performing the task, similar to physiological
techniques (Brünken, Plaas, & Leutner, 2003). A
drawback is that they may influence performance
on the primary task itself. This is particularly true
for very complex learning tasks where learners will
need to invest most of their cognitive capacity in
the primary task, thus leaving almost no capacity
for performing the secondary task (Van Gerven,
Paas, Van Merriënboer, & Schmidt, 2006).
Measuring Cognitive Load in
E-Learning Environments
E-learning environments often present vast
amounts of complex information (e.g., a whole
course) in novel ways (i.e., using unfamiliar
interfaces), imposing high amounts of CL (i.e.,
intrinsic, extraneous, and germane) on learners.
By measuring mental effort during the develop-
ment and use of those e-learning environments,
designers can estimate the effects of their designs
on learner’s CL during learning. Based on such
estimates of mental effort, they can then manage
CL by iterating the design and support until the
conditions for learning within the environment
are optimized. That is, CL imposed by processes
that do not contribute to learning (extraneous
load) should be reduced so that CL imposed by
processes that facilitate learning (germane load)
can be increased up to the maximum of total
cognitive capacity (ideally).
One way to assess mental effort in e-learning
environments is through self-report questionnaires
that yield a subjective measure of the mental effort
experienced when studying the learning material.
In e-learning, however, learning phases may be
extensive. This means that the learners would
need to reflect on the average mental effort that
they experienced over time, which might be more
difficult to determine than estimates at a particular
moment in time (for more information on timing
issues in CL measurement see De Jong 2009).
Asking for a self-report at different moments
during the task (e.g., via a pop-up window) would
partly solve this problem (as shown by Gopher &
Braune, 1984 and Paas, 1992). However, repeated
reporting of effort might interfere with learning
itself or might be unreliable if the learners do not
report it seriously because it draws away too many
resources from the primary task.
More objective CL measurement in e-learning,
albeit indirect and unspecific, can be attained by
analyzing behavioral patterns or physiological
conditions and functions and relating them to the
learning processes. Time-on-task, for example,
can been seen as an indicator for different load
levels; often the different amounts of time learners
spend learning in different variants of multimedia
instruction are reasoned to be the result of differ-
ent amounts of load (Brünken & Leutner, 2001;

1185
Cognitive Load Theory in E-Learning
Brünken et al., 2001). In hypermedia learning lit-
erature, navigation behavior, navigation errors, and
orientation problems such as lost-in-hyperspace
are seen as indicators of CL (Astleitner & Leutner,
1996). Moreover, tracking of eye movements (i.e.,
fixation durations) and degree of pupil dilation
can provide insight into the mental effort that
is invested by the learner. The main problem in
using such eye tracking measures is that they are
also influenced by other factors such as ambient
lighting (both from the lamps in the room and the
screen itself) and screen contrast.
The method of choice to sample mental ef-
fort directly and in real-time in e-learning is the
dual-task paradigm. A performance measure that
has been successfully used in working memory
research is reaction time to environmental stimuli
(Verweg & Veltman, 1996). As secondary task,
a very simple continuous monitoring task is per-
formed in which the learner is required to react to
a specific signal as soon as possible (ASAP). The
monitoring requires minimal cognitive resources
and therefore does not suppress performance of the
primary task, yet when a reaction is necessary, the
ASAP response is motivating enough to consume
all available resources.
The dual task paradigm has been successfully
used in a range of e-learning studies. In computer-
based learning environments, Reed, Burton and
Kelly (1985) used a dual-task design to study the
effects of writing ability and mode of discourse on
cognitive capacity engagement for different com-
puter-based writing tasks with different levels of
difficulty. The cognitive engagement of the learn-
ers increased from the easiest task that induced a
low level of CL to the moderately difficult task
with a medium level of load, but decreased for the
most difficult task that induced the highest levels
of load, suggesting a relation between motivation
and material complexity. Chandler and Sweller
(1996) studied mental effort for computer-based
learning materials for computer-aided design using
a secondary task generated by a second computer
that was presented on a separate computer screen
simultaneously to the design program on the first
computer. Performance on this secondary task was
related to split-attention and redundancy effects
in the primary task (i.e., extraneous load) as well
as element interactivity (i.e., intrinsic load) of
the material, and decreased as the mental effort
needed for the primary task increased.
Thus, using a dual-task paradigm to estimate
the CL of new e-learning environments is recom-
mended since it is objective, direct and easy to
implement. Other measures are either subjective or
are difficult to implement in E-learning. However
for adaptive environments in which the informa-
tion organization, information presentation and
learner support is adapted to the needs, expertise
and other individual characteristics of the learner,
using one of the behavioral measures such as
measuring eye movements and pupil dilation, and
combining this with running average performance
measures such as the learner’s performance on the
finished items is recommended as this enables
fast estimation of the currently experienced CL
online without introducing the extra memory load
that may be induced by asking learner to perform
a second task.
INSTRUCTIONAL DESIGN
GUIDELINES DERIVED FROM CLT
CLT provides important guidelines on how to
design learning environments. These guidelines
can be divided into guidelines to reduce extrane-
ous load, guidelines for using freed-up cognitive
capacity to foster germane load, and guidelines
for managing intrinsic load. The next paragraphs
will elaborate on the effects of these guidelines,
but for more information see Atkinson, Derry,
Renkl, and Wortham (2000) as well as Sweller,
Van Merriënboer, and Paas (1998).

1186
Cognitive Load Theory in E-Learning
Reducing Extraneous
Cognitive Load
To reduce extraneous CL by instructional design
(1) problems should be formulated in an open
manner (i.e., goal-free), (2) worked examples
should be used as teaching method, (3) learning
material should be presented in an integrated (i.e.,
non-split) format, (4) information given in the
learning material should be non-redundant, and
(5) different modalities should be used to present
the to-be-learned information. Each guideline will
now be addressed in more detail.
The Goal Free Effect
The goal-free effect states that it is less cognitively
demanding to ask learners to find any possible
solution or as many possible solutions to a prob-
lem than to ask them for one specific solution
(Sweller, Mawer, & Ward, 1983). If learners are
provided with a very specific goal, they tend to
use a means-end strategy, which is not beneficial
for learning. When applying such a strategy,
learners must simultaneously consider the prob-
lem state, the goal state, the difference between
the two, and the operators required to transfer
the problem state into the goal state. In addition,
they have to consider all possible sub-goals that
may be needed, depending on the complexity of
the task. For novices, simultaneously carrying
out these steps is likely to overwhelm their WM
capacity and hamper learning,
The Worked-Examples Effect
Example-based learning has a long tradition (At-
kinos, Derry, Renkl, & Wortham, 2000; Sweller,
Van Merriënboer, & Paas, 1998) and has shown
to be more efficient than problem-based learning
or learning-by-doing (Cooper & Sweller, 1987;
Sweller & Cooper, 1985). For instance, Zhu and
Simon (1987) could show that implementing
example-based learning in educational practice
can reduce a mathematical curriculum from three
to two years.
Worked examples consist of a problem state-
ment, the solution, and the worked out steps to
go from the problem to the solution (Renkl, 2002;
Van Merriënboer, 1997). This form of instruc-
tion has been shown to be particularly suited
for skill acquisition in novices. Both are crucial
to consider when applying worked examples in
e-learning. First, to enhance skill acquisition
so-called process-oriented worked examples are
best used (Van Gog, Paas, & Van Merriënboer,
2004). Such examples include reasoning on why
and how certain solution steps were chosen. This
additional information can be presented in an
auditory (Struve, 2008) or written manner (Van
Gog, Paas, & Van Merriënboer, 2006). Along
with cognitive processes, perceptual processes
can also be described in worked examples; so-
called eye movement modeling examples (Van
Gog, Jarodzka, Scheiter, Gerjets, & Paas, 2009).
These examples have been shown to contribute
to learning in perceptual tasks for which the task
performer has to actually inspect an image, video,
or other visualization to be able to execute the
task (Jarodzka, Balslev, Holmqvist, Nyström,
Scheiter, Gerjets, & Eika, 2010; Jarodzka, Scheiter,
Gerjets, Van Gog, & Dorr, 2009). For a detailed
overview of current research and perspectives on
worked examples see Renkl (2011) and Van Gog
and Rummel (2010).
The mere use of worked examples, however,
does not guarantee successful learning. The knowl-
edge level of the learner also plays a crucial role.
Kalyuga and colleagues have repeatedly found
the so-called expertise-reversal effect (Kalyuga,
Ayres, Chandler, & Sweller, 2003; Kalyuga, Chan-
dler, Touvien, & Sweller, 2001). This effect refers
to the fact that learners with a higher level of prior
knowledge not only do not learn from worked
examples, but their learning is hampered by the
presentation of this large amount of information

1187
Cognitive Load Theory in E-Learning
(see also redundancy effect below). Moreover,
design guidelines need to be considered when
implementing worked examples. These guidelines
will be described in the following sections.
The Split-Attention Effect
The split-attention effect occurs when differ-
ent information sources (e.g., pictures and an
explanatory text) need to be integrated into one
schema (Ayres & Sweller, 2005; Chandler &
Sweller, 1991). If both sources are spatially or
temporally distributed, they require constant at-
tention shifts by the learner. The difficulty here is
that information from one source has to be kept
active in working memory in order to interpret
the information from the other source. This so-
called ‘split format’ requires mental integration
processes that use cognitive resources, which will
no longer be available for learning (Sweller, Van
Merriënboer, & Paas, 1998).
An integrated format, on the other hand,
presents related information from both sources
in close spatial and temporal proximity (e.g., the
explanatory text is next to the related pictorial
elements). An integrated format has been shown
to lead to superior learning outcomes compared
to split formats where the learner has to mentally
integrate the information her- or himself (e.g.,
Tarmizi & Sweller, 1988; Ward & Sweller, 1990).
Eye tracking research has provided additional
insight into the processes underlying the split-
attention effect (Van Gog & Scheiter, 2010).
Hegarty and Just (1993), for instance, found that
learners integrate text and pictures at a level of
single elements or groups of connected elements.
Learners first read the text describing one ele-
ment, including re-reading, and then inspect the
corresponding part of the picture. Holsanova,
Holmber, and Holmqvist (2009) illustrated that
an integrated (and serial) presentation format of
text and pictures attracts and sustains the attention
of the readers, which in turn leads to a longer and
deeper processing of the information.
Jarodzka, Janssen, Kirschner, and Erkens
(2011) studied the split-attention effect in com-
puter-based testing. In an actual Art examination,
learners completed an electronic version with
half of the questions presented in the original
spatially split-attention format and the other
half in a spatially integrated format (i.e., within-
subject-design). Students’ visual patterns were
captured with eye tracking to estimate the amount
of visual search required. Eye tracking analyses
showed that in the integrated format, learners at-
tended more to additional information (indicated
by total fixation durations) and processed them
more intensively (indicated by more fixations) in
contrast to information presented in a split format.
By changing the design of such testing environ-
ments, learners’ attention was guided so that
they intensively processed all given information.
Interestingly, the integrated format did not lead to
higher, but to lower testing scores. Attending to all
presented information led to lower performance in
this specific computer-based environment. These
results suggest that part of the given information in
the tests was redundant, leading to a redundancy
effect (see below).
The Redundancy Effect
Presenting multiple sources containing the same
information has been shown to hamper learning
in comparison to presenting the information only
once (Chandler & Sweller, 1991). For instance,
learning material that is presented simultaneously
as written text (i.e., visual presentation) and is also
read aloud (i.e., auditory presentation) leads to
lower learning results than when the material is
presented either only visually or only orally. The
effect has been found in various experimental
settings (e.g., Bobis, Sweller, & Cooper, 1993;
Mayer, Bove, Bryman, Mars, & Tapangco, 1996;
Schooler & Engstler-Schooler, 1990).
This simultaneous presentation of redundant
information requires cognitive resources to inte-
grate both information sources at the same time.

1188
Cognitive Load Theory in E-Learning
These integration processes may overwhelm the
learner’s cognitive capacity, which in turn leaves
no capacity for learning processes. Hence, from a
CL perspective it is better not to present redundant
information.
However, one has to keep in mind that whether
information is redundant for a learner also depends
on the learner’s prior knowledge (cf. the above
described expertise-reversal effect, Kalyuga,
Chandler, & Sweller, 1998; 1999). Hence, when
designing e-learning material, the amount of re-
dundancy should ideally be adapted to the current
knowledge status of the learner.
The Modality Effect
The modality effect describes the finding that
presentation formats that make use of several
modalities (i.e., auditory and visual) result in better
learning outcomes than material that make use of
only one (Mousavi, Low, & Sweller, 1995). Pos-
sible explanations for this effect are provided by
Paivio (1991) and Baddeley (1992). Paivio (1991)
assumed that the capacity needed for information
processing can be effectively used by perceiv-
ing text via an auditory channel, while pictures
are perceived in a visual channel. Moreover, the
information is dually coded, in that the text is
coded in long-term memory as a verbal model
and the picture as a pictorial one. Baddeley (1992)
argued in a similar manner, namely that pictures
are processed in a visual-spatial sketchpad, while
text is processed in a phonological loop. If text is
presented in a written format, it first occupies the
visual-spatial sketchpad. Hence, pictures cannot
be processed there at the same time.
Though there are convincing theoretical argu-
ments and many empirical findings supporting
the modality effect (e.g., Mayer, 1997, 2005;
Mayer, & Moreno, 1998; Mayer, Moreno, Boire,
& Vagge, 1999; Mousavi et al., 1995; Tindall-Ford,
Chandler, & Sweller, 1997), the generalizability
of the modality effect has become increasingly
restricted. For instance, research has shown
that this effect only can be found for short texts,
whereas for longer texts the effect may even be
reversed (Leahy & Sweller, 2011). Furthermore,
the modality effect does not occur for complex
visual material in which the learner has to invest
many cognitive resources to find the element
that the auditory text refers to; the effect can
only be found if the learner is guided with ad-
ditional visual cues (Jeung, Chandler, & Sweller,
1997). Finally, the theoretical explanation of the
modality effect is even questioned. Rummer,
Schweppe, Fürstenberg, Seufert, and Brünken
(2009) offer alternative interpretations of this
effect: First, auditory information may be more
easily retained per se, in a way which is unrelated
to working memory load, but rather as an early
sensory process. Second, eye movements made
during reading may require capacities themselves
for saccade planning process, which may hamper
visuo-spatial retention.
In sum, the following guidelines should be
kept in mind when designing e-learning envi-
ronments that free-up cognitive capacity from
extraneous CL:
1. Information in different presentation formats
that refer to each other should be presented
in close spatial and temporal proximity to
avoid splitting the learner’s attention.
2. The information presented in different
formats should be complementary, not
redundant.
3. The complementary information should be
presented across different modalities so that
different sub-systems of working memory
are loaded.
Increasing Germane Cognitive Load
The previous section discussed how extraneous
load can be reduced. There is, however, no guar-
antee that this freed-up cognitive capacity will
be used for learning. To ensure that this occurs,
germane load needs to be increased. To increase

1189
Cognitive Load Theory in E-Learning
germane load (1) multiple examples should be
presented (2) incomplete examples should be pre-
sented, (3) instructional support should be faded,
and/or (4) the subgoal structure of the examples
should be emphasized. All these means to increase
germane cognitive load aim at fostering self-
explanations. Each guideline will be addressed
in more detail in the following.
Presentation of Multiple Examples
Presenting several examples allows learners to
compare between them. When doing so, learners
have the opportunity to determine the differences
between examples of one category and thus, un-
derstand which aspects of the examples vary and
hence, are irrelevant surface features to come along
a solution. Similarities between examples of one
category, on the other hand, must be relevant fea-
tures for the solution. In this way, learners learn to
abstract across irrelevant example features, which
in turn is a prerequisite for schema induction (for
more on the design of a curriculum with several
examples see Atkinson, Derry, Renkl, & Wortham,
2000 and Van Merriënboer & Kirschner, 2007).
Incomplete Examples
Presenting the learners with incomplete examples
(also known as partially worked-out examples),
that is, with examples that provide the problem
statement and solution, but miss certain sub-steps
of the solution process, coerces learners to engage
in self-explaining and to anticipate the missing
steps, which in turn results in increased learning
outcomes (Stark, 1999). Self-explanations, in
turn, are meant to foster learning. This example
type is also known as completion problems (Van
Merriënboer & Kramer, 1990). Self-explanations,
in turn, are meant to foster learning.
Fading Instructional Support
Fading refers to providing – in a stepwise man-
ner - decreased instructional support in worked
examples over time (Atkinson, Renkl, & Derry,
2003). The idea behind this concept is that
with increasing expertise, the learner is able to
self-explain the missing steps (cf. incomplete
examples) her-/himself. Moreover, an otherwise
possible expertise-reversal effect can be avoided
(Kalyuga, 2007). The idea is to foster as many
self-explanations as possible and to provide as
few instructional explanations as possible.
Emphasizing Sub-Goal Structure
Another possibility to enhance learning from
worked examples is to illustrate the sub-goal
structure of the example (e.g., Catrambone, 1994).
To this end, the example is split up into single
solution steps or meaningful units. The learners
must then self-explain why certain steps belong
together and which goal is achieved by applying
which step.
The last three design interventions are meant
to trigger the learner to self-explain the learning
content and thus, to actively deal with it, result-
ing in increased germane load. It has to be noted
that self-explanations can also be fostered directly
via training (e.g., Renkl, Stark, Gruber & Mandl,
1998). This topic is, however, out of scope of
this chapter.
Managing Intrinsic CL
While the previous sections introduced useful and
concrete ways to redirect the learners’ cognitive
resources from processes that hamper learning
to processes to those which foster learning, these
guidelines are only effective if the task that has
to be performed is not too complex (i.e., if the
intrinsic load is not too high). For highly complex
tasks, even after the reducing of extraneous load,
the element interactivity of the materials may

1190
Cognitive Load Theory in E-Learning
still cause a cognitive overload that will prevent
learners from effectively learning from the task
(Van Merriënboer, Kirschner, & Kester, 2003; Van
Merriënboer & Sweller, 2010). In such cases, it
is necessary to decrease the intrinsic load. It is
important to note that intrinsic load can only be
lowered by an increase in expertise or by decon-
structing the task so that fewer elements interact
(see Pollock, Chandler, & Sweller, 2002). It is
paradoxical that for very high element interactivity
materials, an increase in expertise - i.e., learning -
can only occur after the elements have been stored
in a schematic form in LTM. While constructing
these schemas is not possible because of the lack
in expertise; Pollock, Chandler, & Sweller, 2002),
deconstruction of the task seems like the only op-
tion open. However, task deconstruction means
that less will be learned; the intrinsic load cannot be
reduced without reducing what the learner learns
(i.e., what they come to understand). Because it is
important that learners fully understand the mate-
rial, instructional design principles that manage
intrinsic load must ultimately present the task in
its full complexity. This asks for a strategy that
gradually increases the number of interacting ele-
ments, sequencing learning tasks from simple to
complex and learning environments from low- to
high-□delity.
Recently another option for reducing intrin-
sic load was documented, namely the collective
working memory (WM) effect. This effect states
that when performing a highly complex task, the
intrinsic load that is imposed on the learner can be
lowered by letting learners work together so that
they use each other’s working memory capacity.
Without decreasing the complexity of the task, and
with that the amount of what has to be learned,
learners will be able to process all the relevant
information and become more of an expert (thus
providing an alternative to solving the paradox,
F. Kirschner, Paas, Kirschner, 2010a, 2010b).
The following sections provide a global de-
scription of sequencing and the collective working
memory effect.
Simple–to-Complex Strategy
In a simple–to-complex strategy, highly complex
learning material consisting of a many informa-
tion elements that interact and that thus must be
processed simultaneously, are split up so that they
can be processed serially. Presenting learners
with isolated interacting information elements
(Pollack, Chandler, & Sweller, 2002) minimizes
the risk of overwhelming the learner’s limited
working memory with a high intrinsic CL and
will therefore foster learning.
Four experiments conducted by Pollack and
colleagues (2002) provided strong evidence for
the isolated interacting elements instructional ap-
proach. In a 2-phase learning approach, learners
were first presented with complex materials for
which the element interactivity was artificially
reduced by presenting the material as isolated
information elements that could be serially pro-
cessed. In the second phase, all the information for
understanding was presented. The control group
was simply presented with all the information for
understanding in both phases. The results showed
that for students who do not possess rudimentary
schemas and therefore experience a high cogni-
tive load (i.e., novices), information is better
learnt through the isolated interacting elements
instructional method.
Low-to-High-Fidelity Strategy
Manipulating the fidelity of the learning environ-
ment or material (i.e., the physical resemblance
of the learning environment to the real world
situation) is another way to manage intrinsic load.
Because high-fidelity environments and materi-
als contain more interacting elements than low-
fidelity environments and materials, it is possible
to gradually increase the number of interactive
information elements by gradually increasing the
fidelity (Van Merriënboer & Kirschner, 2007).

1191
Cognitive Load Theory in E-Learning
Collective Working Memory Effect
Recently, group or collaborative learning has be-
come recognized as an alternative way to reduce
intrinsic load and thus to overcome individual
WM limitations (F. Kirschner et al., 2009a, 2009b,
2010a, 2010b), in the sense that groups of col-
laborative learners can be considered information
processing systems (Hinsz, Tindale, & Vollrath,
1997), consisting of multiple limited WMs which
can create a collective working space. Within these
systems, valuable task-relevant information and
knowledge held by each group member is con-
sciously and actively shared (i.e., retrieving and
explicating information), discussed (i.e., encoding
and elaborating information) and remembered
(i.e., personalizing and storing information) (Hinsz
et al., 1997; Tindale & Sheffy, 2002). Instead of
isolating interacting information elements, the
information elements within the task and the
associated CL caused by the intrinsic nature of
the task can be divided across a larger reservoir
of cognitive capacity (F. Kirschner et al., 2009a;
Ohtsubo, 2005).
In terms of CLT, group-based learning envi-
ronments can have two conflicting consequences
for individual group members. On the one hand,
collaborating individuals can invest less cogni-
tive effort compared to learners working alone,
because the task’s interactive information elements
with its associated intrinsic CL can be divided
across a larger reservoir of cognitive capacity
(F. Kirschner et al, 2009a; Ohtsubo, 2005); the
distribution advantage. On the other hand, collabo-
rating individuals need to invest cognitive effort
for communicating information with each other
and coordinating their actions, which individuals
working alone do not have to invest. These so
called transactional activities (Ciborra & Olson,
1988; F. Kirschner et al, 2009b; Yamane, 1996)
can be beneficial for or deleterious to learning.
CLT argues that, while cognitive investment in
beneficial transactional activities such as negoti-
ating common ground (i.e., germane) should be
stimulated, an investment in deleterious activities
such as discussing ways to share information (i.e.,
extraneous) should be minimized. The trade-off
between the advantage of dividing information
processing among group members and the dis-
advantage of doing this in terms of having to
cognitively invest in the associated transactional
activities can be an indicator for the efficiency of
group learning. This so called collective working
memory effect was demonstrated in two studies
by F. Kirschner et al. (2010a,b), one on the effects
of low-complexity (i.e., low intrinsic load) and
high-complexity (i.e., high intrinsic load) tasks
on individual and group learning efficiency, and
one on the effects of worked example study (i.e.,
low extraneous load) and conventional problem
solving (i.e., high extraneous load) tasks on
individual and group learning efficiency. These
studies showed that group learning was superior
to individual learning for high-load tasks, but
inferior for low-load tasks.
Collaborative learning environments are
becoming increasingly computer-mediated or
computer based. These environments can be either
synchronous or asynchronous and/or distributed,
non-distributed or blended. When using collabora-
tive e-learning environments as a way to present
learners which high complexity (i.e., high element
interactivity) tasks, there are two main design
principles that have to be taken into account. First,
the environment has to stimulate group members
to invest continuous mutual effort to learn by solv-
ing the problems together, in contrast to dividing
the labor among group members and learning
individually (Paulus, 2005; Roschelle & Teasly,
1995). Second, the environment should provide
rules, regulations and support that will minimize
deleterious transaction activities.

1192
Cognitive Load Theory in E-Learning
FOUR COMPONENT
INSTRUCTIONAL DESIGN
The Four Component Instructional Design model
(4C/ID) offers a standardized way to incorporate
the previously mentioned CLT design guidelines
either in traditional learning environment or in an e-
learning environment. This model was developed
by Van Merriënboer (1997), further developed by
Van Merriënboer, Clark, & De Crook (2002), and
finally described in depth by Van Merriënboer
and Kirschner (2007).
Complex skills that involve learning to interact
with numerous variables monitored in an elec-
tronic environment – such as operating a train, an
airplane, a ship, a brewery or an energy plant - are
characterized by a multitude of interacting tasks
that need to be performed quickly, accurately and
effortlessly. Progress on the tasks needs to be moni-
tored continuously so that difficulties can be dealt
with before they can influence other processes.
Experts can execute a complex skill quickly and
easily because they have constructed elaborate
cognitive schemata of its subtasks in long-term
memory, and these subtasks are highly automated
so that they are treated as one low-capacity ele-
ment when used in short term memory.
Schemata for such complex skills consist of
two components, namely a) mental models that
allow for reasoning in the domain because they
reflect the way in which the domain is organized,
and b) cognitive strategies that guide problem
solving in the domain because they reflect the way
problems might be effectively approached. Mental
models may contain both general and abstract
knowledge, and concrete cases that exemplify
this knowledge. Strong models allow for both
abstract and concrete (case-based) reasoning. Like
mental models, cognitive strategies contain both
general, abstract knowledge and concrete cases
that exemplify this knowledge. Mental models
can be conceptual models, structural models,
or causal models. Conceptual models (What is
this?) focus on how things are interrelated and
allow for classification or description of objects,
events or activities. Structural models (How is
this organized?) describe how plans for reaching
particular goals are related to each other. Plans
can be distinguished into scripts (What happens
when?) that focus on how events are related in
time and help to understand and predict behavior,
and building blocks or templates (How is this
built?) that focus on how objects are related in
space and help to understand or design artifacts.
Causal models (How does this work?) focus on how
principles affect each other and help to interpret
processes, give explanations for events, and make
predictions. Mental models may also combine
these three different perspectives and thus allow
for qualitative reasoning in a particular domain.
Parts of a complex skill are recurrent, that is,
consist of rules that can be used in more than one
application of the skill or in more than one of the
tasks that embody the skill. Other aspects only
apply to one particular instance or one particular
subtask and thus are non-recurrent. Recurrent
aspects of the skill can be automated by repeated
use, which results in a situation in which their
execution no longer needs conscious effort and
thereby a reduction in the capacity it requires in
short term working memory. Automation of recur-
rent aspects of a skill allows experts to execute
complex skills quickly and effortlessly.
Acquiring such complex skills inherently
encompasses heavy pressure on short term work-
ing memory because novices lack the elaborate
schemata in long-term working memory that are
needed to complete the task within a reasonable
amount of time. Also, because they are character-
ized by interacting elements, complex skills cannot
be broken down into separate simple skills that
can be practiced in isolation. Instruction methods
for complex skills, thus, require learning environ-
ments that:

1193
Cognitive Load Theory in E-Learning
1. Minimize extraneous CL,
2. Stimulate quick formation of schemata in
long term working memory that address the
relation among elements, and
3. Stimulate rule automation as quickly as
possible.
The 4C/ID model takes these rules into ac-
count when designing new learning materials and/
or revising existing ones and has been shown to
result in improved skill acquisition. For instance,
in the computer domain, several studies including
both classroom studies (Van Merrienboer, 1990a,
1990b) and computer-based training studies
(Schuurman, 1999; van Merrienboer & de Croock,
1992; van Merrienboer, Schuurman, de Croock &
Paas, 2002) demonstrated that materials designed
in accordance with 4C/ID principles led to higher
transfer performance than control strategies based
on conventional design (i.e. based on the good
judgment of the developer). This superiority
became more evident on far transfer problems
for which learners had to, for example, design
and construct new computer programs requir-
ing solutions not reached before. This has been
found in many domains, such as statistics (Paas,
1992, 1993), computer numerically controlled
programming (Paas & van Merriënboer, 1994)
and fault management in process industry (de
Croock, 1999; de Croock, van Merriënboer &
Paas, 1998; Jelsma, 1989).
The 4C/ID model designs instruction for
complex learning in four steps (Van Merrienboer,
Clark, & de Croock, 2002), namely:
1. Designing and sequencing learning tasks
2. Designing and developing supportive
information
3. Designing and developing just-in-time
information
4. Designing and developing (if needed) part-
task practice
Step 1: Designing and Sequencing Learning
Tasks
Learning tasks need to be whole-skills to promote
schema construction for also the non-recurrent
aspects of the skill and, to a certain degree, rule
automation by compilation for the recurrent as-
pects (Van Merriënboer & Kirschner, 2007). By
including all aspects of the skill, albeit in very
simple setting for the first few tasks, the interac-
tions between the different elements of the skill
are given proper attention in the construction of
learners’ mental schemata. For the non-recurrent
aspects, learning tasks should promote schema
construction and elaboration by inductive pro-
cessing, that is, through mindful abstraction away
from the concrete experience that the learning task
provides (i.e., generalization and discrimination).
Task classes define the simple-to-complex
categories of learning tasks and should steer the
process of selection and development of suitable
learning tasks. Learning tasks within the same task
class are equivalent, that is, they can be performed
on the same body of knowledge. Complexity is
determined by the number of constituent skills
involved, the number of interactions between
these skills, and the amount of knowledge neces-
sary to perform them. A more complex task class
requires more knowledge or more elaboration of
knowledge for effective performance. Once task
classes are defined, learning tasks can be selected
and developed for each class. Learners need to
start with relatively simple learning tasks to avoid
cognitive overload, which impairs learning and
performance (Sweller, Van Merrienboer & Paas,
1998) and slowly progress towards more complex
tasks. For each task class, a sufficient number of
learning tasks should be provided to ensure that
the learner reaches mastery. Also, tasks within
the same class should show a high variability to
promote the development of rich schemata that
allow for schema-based transfer from the training
program to the real world. They can vary on the
saliency of the defining characteristics, the context

1194
Cognitive Load Theory in E-Learning
in which the task is performed, its familiarity, or
any other dimensions that also varies when ap-
plying the skill in the real world.
Learning tasks within the same task class have
equal difficulty, but are not given equal support.
Much support is given for tasks early in the task
class, and no support is given for the final one.
This process of diminishing support as learners
acquire more expertise is called scaffolding and
is repeated for each subsequent task class, yield-
ing a saw-tooth pattern of support throughout the
whole training program. A general framework of
human problem solving is used to distinguish sup-
port structures. Four elements (based on a general
framework of problem solving, Newell & Simon,
1972) describe learners’ work on a learning task:
A. The given state that the learner is confronted
with
B. The criteria for an acceptable goal state
C. A solution: a sequence of operators that
enables smooth transition from the given
state to the goal state
D. A problem-solving process: the tentative
application of mental operations in order to
reach the solution.
Product-orientated support relates to the first
three elements: given state, goal state and solution
(cf. Van Gog, Paas, & van Merriënboer, 2007).
Highest product-oriented support is provided by
a case-study or worked-out example in which the
learner is confronted with a given state, a desired
goal state, and a solution, one or more intermedi-
ate solutions, or both. Typically, learners need to
answer questions that provoke deep processing
and the induction of mental models from the
given materials.
Process-orientated support also takes the
problem-solving process into account (Van Gog,
Paas, & van Merriënboer, 2004, 2006). Process-
orientated support is provided by a modeling
example, in which an expert performs the task
and explains how and why (s)he is executing
each step. This way, the learners get a clear idea
of the systematic approaches and rules of thumb
professionals use. Another way to give process-
orientated support is by providing performance
constraints and performance support structures.
Both are based on a cognitive task analysis of
strategic knowledge, which yields a description
of cognitive strategies as systematic approaches
to problem solving (SAPs) that experts use to
solve problems in the domain of interest. A SAP
distinguishes the successive phases in a problem-
solving process and the rules of thumb or heuristics
that may be helpful to successively complete each
of the phases. Performance constraints typically
require the learner to complete one phase before
moving to another. Performance support structures
are less directive than constraints and typically
take the form of problem-solving support (i.e.,
process worksheets) that give the main phases
and useful rules of thumb for each of the phases
or computer-based learning tools that show how
an expert would do it. Recently, process-oriented
support for perceptual tasks has been introduced
in the form of eye movement modeling examples
(Van Gog, Jarodzka, et al., 2009). These are mod-
eling examples that provide visual information on
where the expert is focusing her/his attention along
with verbal explanations of why and how (s)he is
executing the task. This is done by displaying the
eye movements of the expert recorded while (s)
he was executing a task. This form of support is
particularly helpful for tasks with high perceptual
complexity such as classifying locomotion pat-
terns of reef fish (Jarodzka et al., 2009) or diag-
nosing epileptic seizures of infants (Jarodzka et
al., 2010). As eye movement modeling examples
are provided in form of educational videos, they
can easily be incorporated in e-learning environ-
ments, such as instructional videos about surgery.

1195
Cognitive Load Theory in E-Learning
Step 2: Designing and Developing Supportive
Information
Supportive information needs to reflect both
the mental models and the cognitive strategies
needed to perform a skill. Because the same body
of knowledge underlies all learning tasks within
the same task class and because it is not known
beforehand which knowledge is precisely needed
to successfully perform a particular learning task,
supportive information is coupled to task classes
and not to individual learning tasks.
Supportive Information for
Developing Mental Models
Supportive information should provide a bridge
to the learners’ prior knowledge and should there-
fore primarily aim at elaboration (i.e., broaden-
ing the schemata by establishing non-arbitrary
relationships between new elements and familiar
knowledge; Van Merriënboer & Kirschner, 2007).
Expository instructional methods explicitly pres-
ent the non-arbitrary relationships to the learners.
Inquiry methods, on the other hand, ask the learner
to ‘discover’ the relationships. The latter methods
are time-consuming, but directly build on prior
knowledge and may, therefore, be appropriate
for interconnecting new information to already
existing cognitive schemata.
A particularly important relationship between
skill elements is the experiential one which relates
general, abstract knowledge to concrete cases.
4C/ID distinguishes between the presentation of
general information (i.e., a didactic specification
of conceptual, structural and causal models) and
concrete cases or case studies that illustrate this
information. If a conceptual perspective is taken,
case studies may describe concrete objects, events
or situations. For models with a structural per-
spective, case studies may be designed to reach
particular goals. And for models with a causal
perspective, case studies may illustrate real-life
processes (e.g., in computer-based simulations). In
such simulations, learners can change the settings
of particular variables and study the effects of those
changes on other variables. The goal here is not
primarily to practice the complex target skill but
to help learners construct mental models of how
the world is organized through active exploration.
The 4C/ID-model further distinguishes be-
tween inductive and deductive strategies for
presenting supportive information. In an inductive
strategy, one or more case studies is first pre-
sented as part of the supportive information. Then,
general, more abstract information is dealt with.
Finally, learning tasks are given. In an inductive
inquiry strategy, one or more case studies is first
presented and then learners are asked to identify
the relationships between pieces of information
illustrated in the case. This is time-consuming and
should be used only if there is enough instruc-
tional time available; learners have no experience
with the skill, and a deep level of understanding
is required. In an inductive expository strategy,
case studies are presented along with the explicit
explanation of relationships between the pieces
of information illustrated in the cases. The 4C/ID
model suggests using this as a default approach
because it is time-effective, and starting with
concrete, recognizable case studies works well
for learners with little prior knowledge. The third
alternative is a deductive strategy where learners
work from general, abstract information directly
towards learning tasks that fulfill the role of case
studies. One starts by explicitly presenting rela-
tionships between pieces of information (i.e., the
theory) and then illustrates this with one or more
learning tasks with maximum product-oriented
support. Learners without prior knowledge may
have severe difficulties understanding and stor-
ing the general information. This method should
only be used when instructional time is limited,
learners have already some experience with the
skill, and a deep level of understanding is not
strictly necessary.

1196
Cognitive Load Theory in E-Learning
Supportive Information for
Cognitive Strategies
Cognitive strategies may be analyzed as SAPs
describing the successive phases in a problem-
solving process and the heuristics that may be
helpful to successfully complete each of the phases
(Van Merriënboer & Kirschner, 2007). Instruc-
tional methods for presenting cognitive strategies
closely resemble methods for presenting mental
models, and in particular structural and causal
models. For instance, one might ask a learner to
explain why one phase should precede the other,
predict the effects of rearranging phases, explain
how the use of particular rules of thumb brought
a particular state of affairs about, or predict the
effects of the use of particular heuristics. One
way to illustrate how the application of SAPs
can help to reach a solution is by using modeling
examples. Such a modeling example might be an
expert performing the task whilst simultaneously
explaining why particular decisions and actions
are taken, preferably interspersed with questions
that require the learner to think critically about
the shown problem solving process (i.e., epistemic
questions). Because of the highly abstract character
of cognitive strategies, the 4C/ID model prescribes
only an inductive expository strategy for their
presentation. That is, one should start with the
presentation of one or more modeling examples
and then explicitly present the problem-solving
phases and rules of thumb that are illustrated by
those examples.
Feedback on Non-Recurrent
Aspects of the Skill
Cognitive feedback, that is feedback on the qual-
ity of performance on the non-recurrent aspects
of the skill, should promote schema construction
(Van Merriënboer & Kirschner, 2007). It should
stimulate learners to reflect on the quality of
their personal problem-solving processes and the
solutions found, so that more effective mental
models and cognitive strategies can be developed
(as in cognitive apprenticeship models; Collins,
Brown & Newman, 1989; Kluger & DiNisi,
1998). Because non-recurrent performance is
never ‘correct’ or ‘incorrect’, but only more or
less effective, cognitive feedback is only provided
after learners have finished one or more learning
tasks, or even after they have finished a whole
task class. Examples of cognitive feedback are
debriefing sessions, peer or expert critiques, and
group discussions. The effectiveness of personal
problem-solving processes might be evaluated
by comparing them to presented SAPs, modeling
examples that presented those SAPs, problem solv-
ing processes shown by other learners, presented
general information, case studies that presented
this general information, or solutions found for
previous problems (either by the learner self or
by others).
Step 3: Designing and Developing Just-in-
Time Information (JIT)
Rules that enable learners to correctly carry out
the recurrent aspects of a complex skill (i.e., rules)
are formed through practice. Learning such rules
is facilitated when the information necessary for
forming the rules is directly available in WM
precisely when learners need it (Kester, Kirsch-
ner, van Merriënboer & Bäumer, 2001). This
information can describe the rules themselves
(or procedures that combine those rules) and/or
describe the knowledge elements (i.e., facts, con-
cepts, plans or principles – the same knowledge
elements that make up complex schemata) that
are prerequisite to learning and carrying those
rules out. The instructional methods for presenting
JIT information primarily promote compilation
through restricted encoding of situation-specific
knowledge in cognitive rules (Van Merriënboer &
Kirschner, 2007). JIT information can be given, for
example, in the form of directions that an instructor
(e.g., teacher, trainer) typically gives to learners
during practice, acting as ‘an assistant looking

1197
Cognitive Load Theory in E-Learning
over your shoulder’. Because JIT information
is specified at the learner’s entry level, it does
not need to be embedded in existing schemata in
declarative memory and no particular reference
has to be made to related knowledge structures
in long-term memory.
Information Displays
JIT information needs to be organized in small
units called information displays. This is essential
because this prevents processing overload during
practice. Information displays include didactic
rule specification describing correct performance
along with the prerequisite knowledge to correct
rule application (Van Merriënboer & Kirschner,
2007). For instance, the whole complex skill
of flying an airplane can be broken down into
separate learning tasks, such as plotting flight a
pattern, initiating a take-off, checking the altitude,
checking the fuel, radioing flight control, land-
ing, and so on. Correct handling of each of these
tasks could be illustrated in a simulation showing
how the pilot should operate the different buttons
and displays. The learner then practices various
instances of the same task in isolation, trying to
replicate expert pilot operations as accurately as
possible. As the JIT information is identical for
many learning tasks requiring the same recurrent
constituent skills, it is typically provided during
the first learning task for which the skill is relevant
and faded away on subsequent tasks as learners
gain more expertise. If training takes place on
the job, the designer of the instructional method
often has no control over which learning task is
first presented to the learner. Learning aids such
as online-help systems, checklists, and manuals
provide a good alternative in such situations. Al-
though not directly presented when needed, the
information is both easily available and readily
accessible. Demonstrations of the recurrent aspects
of a complex skill ideally coincide with suitable
learning tasks such as modeling examples and
instances of prerequisite knowledge elements
ideally coincide with suitable learning tasks such
as case studies. This is a deductive-expository ap-
proach where the generalities (i.e., the information
displays) are presented simultaneously with the
examples which are part of the same learning task
as the information display that it is connected to.
Feedback on Recurrent Aspects of Skill
Feedback should be provided on the recurrent
aspects of performance that promote compilation
(Van Merriënboer & Kirschner, 2007). If rules that
algorithmically describe effective performance are
not correctly applied, the learner is said to make an
‘error’. Corrective feedback on such errors should
be given directly after misapplication of that rule.
This feedback is essential to learning because it
is necessary for the learner to preserve informa-
tion about the conditions for applying a particular
rule in WM until feedback (e.g., right, wrong) is
obtained. Only then can a rule be compiled that
attaches the correct action to its critical conditions.
Any feedback delay may hamper learning. Well-
designed feedback should inform the learner why
there was an error and hint as to how to change
the rule to reach the goal (e.g., in the form of an
example or a demonstration). The correct action
should not simply be given because this would
not allow for the practice that is critical for com-
pilation. It might also be necessary to indicate to
the learner how to recover from the error made.
Step 4: Designing and Developing Part-Task
Practice
If a very high level of automation of particular
recurrent aspects is required, the learning tasks
might not provide sufficient repetition to provide
the necessary amount of strengthening. Part-task
practice consists of practice items to promote rule
automation for selected recurrent aspects of the
whole complex skill (Van Merriënboer & Kirsch-
ner, 2007). Overreliance on part-task practice is
not helpful to complex learning.

1198
Cognitive Load Theory in E-Learning
It is critical to start part-task practice within
an appropriate cognitive context because this
practice has been found to be effective only after
exposure to the whole complex skill or a simple
version of it (Carlson et al., 1990; Schneider &
Detweiler, 1988, both in van Merrierboer, Clark
& de Croock, 2002). One should, thus, identify
the first task class for which performance of the
recurrent aspect is required and initiate part-task
practice during this task class – preferably after
case studies or other learning tasks with ample
learner support already have been worked on.
This allows learners to identify the activities that
are required to integrate the recurrent aspect in
the learning tasks.
For part-task practice, it is important that the
whole set of practice items is representative for
all situations that can be treated by the rules. This
is necessary to develop a broad set of situation-
specific rules that may subsequently yield optimal
rule-based transfer to new problem situations.
Performance support for part-task practice
takes the form of procedure support. Special
practice items may be relevant if algorithms leave
learners error-prone, or are easily confused. A
well-known strategy for ordering practice items
is the recognize-edit-produce sequence (REP,
Gropper, 1983, in Van Merrierboer, Clark, &
de Croock, 2002). Learners start with items that
require them to recognize which rules to apply,
continue with items for which they have to edit
incorrect applications of the rules, and finish with
conventional items for which they have to apply
the rules to produce the solution. Performance
constraints may, for example, take the form of
training-wheel interfaces (Carroll et al., 1988,
in Van Merrierboer, Clark, de Croock, 2002);
if particular rules leave learners error-prone,
one may make the actions related to those rules
‘unreachable’ early in the training. Such training
wheels might also be used to support learning
recurrent aspects during whole-task practice on
learning tasks.
The principle of presenting JIT information
may be carried even further for part-task practice
by providing relevant information for applying
a particular rule and its prerequisite knowledge
precisely when this one rule has to be applied
(single-step or step-by-step instruction, Landa,
1982 in Van Merrienboer, Clark, & de Croock,
2002).
To automate a recurrent skill, extensive training
may be necessary which changes the underlying
learning process from one of compilation to one of
strengthening. The ultimate goal of automation of
a recurrent skill is not always 100% accuracy, but
might be acceptable accuracy combined with high
speed and the ability to perform the skill together
with other skills, and ultimately, in the context of
the whole task. To attain this, the recurrent skill is
first practice under speed stress. After the desired
execution speed is reached, the skill is practiced
under time-sharing conditions simultaneously
with other effort-demanding skills. Finally, the
whole skill is practiced in the context of the whole
task. Finally, part-task practice is best intertwined
with learning tasks if more than one recurrent skill
needs to be learned, as this provides distributed
practice and enables the learner to relate each skill
to other constituent skills and to the whole complex
skill (Schneider, 1985, in Van Merrienboer, Clark
& de Croock, 2002).
Appropriate Media
Each of the four components of the 4C/ID model
corresponds to a different category of learning
processes, and each of the learning processes is
best supported by particular media. According
to the 4C/ID model, learning tasks are the main
component in the model and thus determine
what primary media is used (Van Merriënboer
& Kirschner, 2007). The instructional systems
developed will, thus, typically involve a real or
simulated task environment in which the whole
skill can be practiced, such as a problem, simula-

1199
Cognitive Load Theory in E-Learning
tion, case, or scenario. Secondary media are related
to supportive information (e.g., books, hypertext
systems, lectures), JIT information (e.g., online
help-systems, job-aids, pop-up menus, balloons),
and part-task practice (e.g., drill-and-practice
computer programs, part-task trainers).
E-learning environments provide an ideal
setting for training complex skills developed on
the basis of the 4C/ID model because they are
flexible and immediate. Computer simulations are
often more easily built and adjusted than real-life
environments (e.g., a nuclear reactor, chemical
plant, Wall Street), its content can be organized
in a network of pages that can be accessed in
non-linear ways (i.e., mimicking information
organization in the brain), specific learning routes
in this network can be enforced depending, for
example, on the current expertise of the learner,
and support and feedback can be presented ad-
jacent to and during learning tasks and part-task
practice at specific times.
CURRENT STATUS OF
CLT IN E-LEARNING
Conventional methods to lower extraneous CL
(e.g., using goal-free tasks, worked examples or
completion problems, avoiding split attention and
redundancy) often fail to lower it to an acceptable
level in e-learning environments because of the
high intrinsic complexity of the learning mate-
rial (Van Merriënboer & Ayres, 2005). There are
several tools that can be used to diminish CL in
e-environments.
Sequencing methods can be used to decrease
intrinsic CL by reducing element interactivity in
the early phases of learning (cf. Clark, Ayres &
Sweller, 2006). For instance, one might first pres-
ent information with only a few relevant element
interactions present, and then gradually increase
the number of required interactions. Another ap-
proach is first presenting a simple version of the
task and then gradually increasing the complexity.
In e-environments, not only is it so that the
content of the learning material has high element
interactivity, but the technology that presents the
information is also often complex. Intrinsic CL
may be increased if both technology knowledge/
skills and specific subject content concepts are
learned concurrently. Sequencing the learning of
the technology skills before learning the content
has been shown to improve learning, especially
for learners with low levels of technology knowl-
edge/skill (Clark, Ayres, & Sweller, 2005; Van
Merriënboer, Kirschner, & Kester, 2003).
Animations (e.g., visuals, simulations) that
can be manipulated enhance learning in learners
who have the necessary prerequisite knowledge
of the topic (Schnotz & Rasch, 2005).
Presenting two integrated non-redundant rep-
resentations (i.e., verbal and visual) requires less
mental effort than using just one (i.e., verbal or
visual) and has been shown to increase learners’
knowledge acquisition (Moreno & Valdez, 2005).
As described in the foregoing, instructional
methods that work well for novice learners may
have no positive or even negative effects for learn-
ers with more expertise (i.e., expertise reversal
effect: Kalyuga, 2007). Through the possibility
of electronic environments, it might not be neces-
sary to develop completely different instructional
designs for learners at different expertise levels.
In e-learning, an adaptive system can allow one
overall design to be effective for different learners
(e.g., displaying either graphics or text or both
depending on the expertise of the learner; Van
Gog, Ericsson, Rikers, & Paas, 2005). Addition-
ally Van Gog et al. (2005) found that designing
deliberate practice that includes feedback on errors
that is adapted to the current level of expertise of
the learner, may enhance germane CL. This effect
is enhanced if learners are motivated to make ef-
fective use of the feedback to correct errors and
change the appropriate schemas.
E-environments cannot only be adapted to the
learner’s expertise, but also to the CL the learner is
experiencing. This is related to expertise: learners

1200
Cognitive Load Theory in E-Learning
experience lower CL with specific materials as
they gain experience. Kayuga and Sweller (2005)
used a rapid measure of semantic knowledge
and a subjective CL measure to calculate learner
cognitive efficiency at different moments to adapt
subsequent instruction – including difficulty level -
to each learner’s individual needs. Their adaptable
environment resulted in more efficient learning.
This, thus, may be a promising approach for fu-
ture development of new instructional e-methods,
especially if one can rapidly sample CL during
learning in a more objective way.
FUTURE DIRECTIONS
The use of CLT in the instructional design is
relatively well established; both by applying mul-
timedia theory and using 4C/ID to design learning
materials for e-learning. There are however still
a few relatively unexplored venues.
One such venue is the role of teacher support.
E-learning environments differ from traditional
face-to-face environments with respect to learner
assistance. In face-to-face instruction, the learner
can ask for assistance. With e-learning environ-
ments there is often no teacher directly available.
The learner must email the instructor and wait for
a response, reducing motivation due to the time
delay. Different methods to address this issue will
be discussed below.
Another venue is the development of an objec-
tive cognitive load measure that can be sampled
continuously online during E-learning. Such a
measure would allow for direct estimates of men-
tal effort at each phase of learning, and provide a
basis for adaptive content, adaptive organization
of the environment and adaptive support.
A third issue is that although Cowan (2001)
and Miller (1956) have supplied rules-of thumb,
not all learners have equal cognitive capacity
and hence learners could still have difficulty or
experience material as too easy even if it has been
designed with CL theory in mind. In what ways
can such individual differences in capacity be
sampled and how can e-learning environments
accommodate them?
Finally, even if the design of the learning ma-
terial takes reducing extraneous CL into account,
there is no guarantee that the freed-up cognitive
capacity will be devoted to germane CL. A learner
needs to be motivated to allocate all available
resources to learning if (s)he wants to improve
her/his learning outcome.
The following sections discuss our vision on
promising new approaches to deal with these
issues.
Teacher Support
Holmberg (1989) suggested learning material
designers develop and integrate guided didactic
conversations or internal speech that a learner
would normally have with an instructor into the
system. Two approaches for this have been sug-
gested: 1) the use of animated agents, 2) including
textual annotations based on semantic analysis.
These will be discussed below.
Moreno, Mayer, Spires & Lester (2001) found
that including an intelligent agent in an e-learning
environment in form of an animated character that
guided the learning process, in a similar way to
receiving help of teacher in form of narration,
stimulated deeper learning (as evident from
higher transfer scores) when compared with no
support or providing only visual or auditory text
monologues on the concepts discussed. This effect
was primarily due to the auditory narration that
simulated a teacher-student conversation.
Wallen, Plass, and Brünken (2005) studied
whether using text annotations based on sche-
matic analysis of the subject domain supported
novices’ schema development and resulted in
increased learning. Three strategies using selec-
tion, organization, and integration were tested
individually and in pairs to determine their effect
on learning. Adding definitions of terms with
contextual information (selection level), brief

1201
Cognitive Load Theory in E-Learning
explanations of an idea in the specific context
(organizational level), or by showing links of ideas
in a paragraph (integration level) enhanced recall
of terms and ideas. It appears that integrating one
of these strategies does not increase extraneous
CL significantly, but rather, it enhances germane
CL. Providing definitions and explanations as
an adjunct to the text enhanced the recall of idea
units. Using both selection and organization level
annotation resulted in lower performance than us-
ing only one of these annotations, suggesting that
use of annotations should be restricted to avoid
the redundancy effect.
Some issues about providing support that
simulates teacher-student interaction remain. For
instance, do the effects of including such support
depend on the expertise of the learner? Can such
support be detrimental if the experienced student
considers the animation or support to be redundant
redundant?
New Objective Measures
of Mental Efficiency
A well-known link exists between the eye and
the hand. There is an automatic coupling between
them to ensure that the hand follows where the eye
has been and there is clear cut evidence that the
planning of eye and hand movements are based
on the same information source (cf. Sailer, Eggert,
Ditterich, & Straube, 2000). Delay and error in
this coupling appears to be related to the load the
user is experiencing performing a task; the higher
the CL the larger the delay and the more errors.
This is backed up by the finding that increased
fatigue in a model that simulates user cognition
and motor action when interacting with a human
computer interface results in an increase in eye-
hand coordination noise (Duric et al., 2002).
We are currently developing a novel measure
that objectively samples mental load online
during e-learning using a combination of eye
tracking and mouse logging. This measure takes
advantage of the above discussed eye-hand cou-
pling for its calculations (Van Mierlo, Jarodzka
& Kirschner, 2011). This measure will also be
used to see whether there are specific features in
mouse movements that indicate increased load.
For instance, hovering above, clicking on and
highlighting certain bits of text might be related
to increased effort digesting it.
Adaption to Individual Differences
in Cognitive Capacity
Continuous real-time online mental effort mea-
surement as described in the previous section
opens up a world of possibilities to adjust the
status of an environment to the learner’s current
mental load and motivation.
For instance, if eye-hand errors increase or if
scan paths become less efficient (i.e., an increase
in fixations on irrelevant locations) this would
indicate that a learner is currently experiencing
increased load. Extra support could then be offered
in the form of pop-up messages that indicate that
more support can be provided if the learner wants
it, or in form of pop-up windows that automatically
appear when the mouse moves over complicated
information. The information organization and
layout could also be adjusted based upon such
measurement of CLT. Finally, the structure of
the environment could be intelligently adapted
to the learner’s current mental model – her/his
current expertise level - based on analysis of
eye-hand movements over previously presented
information.
A drawback of using such a highly adaptable
interface is that it if there is no way to trace back
previous steps, the learner might more easily get
lost in hyperspace. This can be avoided by pro-
viding information about navigation history (e.g.,
giving already visited links and pages a different
color with the hue of this color depending on the
time visited; the longer ago, the darker the color).

1202
Cognitive Load Theory in E-Learning
Increasing Motivation
and Commitment through
Collaborative E-Learning
The intrinsic load that learners experience during
e-learning can be reduced by making use of col-
laborative learning. Distribution of large amounts
of learning material across individuals within
a group should effectively reduce experienced
load. In addition to this intrinsic load reduction,
collaborative e-environments may also foster
learning by increasing the learners’ motivation
and commitment. Inspired by research on group
efficacy (Bandura, 1986) F. Kirschner, Paas, and
Kirschner (2011) explored an alternative affective
explanation for results found in previous studies
where learners in groups performed better on com-
plex tasks than individual learners (F. Kirschner,
Paas, Kirschner, 2009ab, 2010ab). Group-efficacy
is an extension of Bandura’s (2007) concept of
self-efficacy, and refers to a person’s belief in
the capacity of the group to perform a specific
task (Bandura, 1986). According to an affective
explanation, individuals working together in a
group have more confidence in their ability to
solve a problem together and that there will,
thus, be a greater motivation for them to carry
out the task in a group than that for individuals
working on their own. By measuring the amount
of mental effort learners expected to invest in
working on a learning task before actually carry-
ing out the task, F. Kirschner et al (2011) showed
that learners who had to collaboratively solve a
high-complexity problem expected to invest less
mental effort than learners who had to solve the
problem alone. From an efficacy viewpoint this
shows that the prospect of collaboration lead to
learners feeling more confident about successful
task completion of high-complexity tasks (i.e.,
tasks that are difficult to solve by a single learner).
This effect can be further enhanced if the vary-
ing preferences of the different individuals in the
group are addressed in the design of the e-course.
For example, letting different groups compete by
rating each of their assignments and comparing
its quality to that of the other groups, would make
males feel satisfied in their need for competition
and females satisfied in their need for collabora-
tion and distributing responsibility (Van Vugt, De
cremer & Janssen, 2007).
Future studies should investigate what role
social group pressure plays on people’s com-
mitment to collaborative learning tasks. Without
collaboration each individual learner will not pass
the criterion that must be reached (i.e., will fail
the course). It can be expected that awareness of
other people’s expectations, and failure or success
to meet those expectations, might result in better
learning outcomes.
REFERENCES
Annett, J. (2002). Subjective rating scales: Sci-
ence or art? Ergonomics, 45(14), 966–987.
doi:10.1080/00140130210166951
Astleitner, H., & Leutner, D. (1996). Applying
standard network analysis to hypermedia systems:
Implications for learning. Journal of Educational
Computing Research, 14, 285–303. doi:10.2190/
W2GB-05NT-VJRN-PGY9
Atkinson, R. C., & Shiffrin, R. M. (1968). Hu-
man memory: A proposed system and its control
processes. In Spence, K. W., & Spence, J. T.
(Eds.), The psychology of learning and motiva-
tion: Advances in research and theory (Vol. 2).
New York, NY: Academic Press.
Atkinson, R. K., Derry, S. J., Renkl, A., &
Wortham, D. (2000). Learning from examples:
Instructional principles from the worked examples
research. Review of Educational Research, 70,
181–214.

1203
Cognitive Load Theory in E-Learning
Atkinson, R. K., Renkl, A., & Merrill, M. M.
(2003). Transitioning from studying examples
to solving problems: Effects of self-explanation
prompts and fading worked-out steps. Jour-
nal of Educational Psychology, 95, 774–783.
doi:10.1037/0022-0663.95.4.774
Ayres, P., & Paas, F. (2009). Interdisciplinary per-
spectives inspiring a new generation of cognitive
load research. Educational Psychology Review,
21, 1–9. doi:10.1007/s10648-008-9090-7
Ayres, P., & Sweller, J. (2005). The split-attention
principle in multimedia learning. In Mayer, R. E.
(Ed.), The Cambridge handbook of multimedia
learning (pp. 134–146). New York, NY: Cam-
bridge University Press.
Baddeley, A. D. (1986). Working memory. Oxford,
England: Oxford University Press.
Baddeley, A. D., & Hitch, G. (1974). Working
memory. In Bower, G. H. (Ed.), The psychology
of learning and motivation: Advances in research
and theory (Vol. 8, pp. 47–89). New York, NY:
Academic Press.
Beatty, J. (1982). Task-evoked pupillary responses,
processing load, and the structure of processing
resources. Psychological Bulletin, 91, 276–292.
doi:10.1037/0033-2909.91.2.276
Beatty, J., & Lucero-Wagoner, B. (2000). The
pupillary system. In Cacioppo, J. T., Tassinary,
L. G., & Berntson, G. G. (Eds.), Handbook of
psychophysiology (2nd ed., pp. 142–162). Cam-
bridge, MA: Cambridge University Press.
Bobis, J., Sweller, J., & Cooper, M. (1993). Cogni-
tive load effects in a primary school geometry task.
Learning and Instruction, 3, 1–21. doi:10.1016/
S0959-4752(09)80002-9
Borg, G. A. (1977). Simple rating methods for
estimation of perceived exertion. In Borg, G.
(Ed.), Physical work and effort. New York, NY:
Pergamon Press.
Borg, G. A. V. (1982). Psychophysical bases
of perceived exertion. Medicine and Sci-
ence in Sports and Exercise, 14, 377–381.
doi:10.1249/00005768-198205000-00012
Braver, T. S., Cohen, J. D., Nystrom, L. E., Jonides,
J., Smith, E. E., & Noll, D. C. (1997). A parametric
study of prefrontal cortex involvement in human
working memory. NeuroImage, 5, 249–262.
doi:10.1006/nimg.1996.0247
Brünken, R., & Leutner, D. (2001). Aufmerksam-
keitsverteilung oder Aufmerksamkeitsfokussier-
ung? Empirische Ergebnisse zur “Split-Attention-
Hypothese” beim Lernen mit Mulitmedia [Split
of attention or focussing of attention? Empirical
results on the split-attention-hypothesis in mul-
timedia learning]. Unterrichtswissenschaft, 29,
357–366.
Brünken, R., Plass, J. L., & Leutner, D. (2003).
Direct measurement of cognitive load in Multi-
media Learning. Educational Psychologist, 38,
53–61. doi:10.1207/S15326985EP3801_7
Brünken, R., Plass, J. L., & Leutner, D. (2004). As-
sessment of cognitive load in multimedia learning
with dual-task methodology: Auditory load and
modality effects. Instructional Science, 32, 115–
132. doi:10.1023/B:TRUC.0000021812.96911.c5
Brünken, R., Steinbacher, S., Plass, J. L., &
Leutner, D. (2002). Assesment of cognitive load
in multimedia learning using dual-task method-
ology. Experimental Psychology, 49, 109–119.
doi:10.1027//1618-3169.49.2.109
Brünken, R., Steinbacher, S., Schnotz, W., &
Leutner, W. (2001). Mentale Modelle und Effekte
der Präsentations und Abrufkodalität beim Lernen
mit Mulitmedia [Mental models and the effect
of presentation and retrieval mode in multimedia
learning]. Zeitschrift fur Padagogische Psycholo-
gie, 15, 15–27. doi:10.1024//1010-0652.15.1.16

1204
Cognitive Load Theory in E-Learning
Catrambone, R. (1994). Improving examples
to improve transfer to novel problems. Mem-
ory & Cognition, 22, 606–615. doi:10.3758/
BF03198399
Chandler, P., Ayres, P., & Sweller, J. (2006). The
impact of sequencing and prior knowledge on
learning mathematics through spreadsheet ap-
plications. Educational Technology Research
and Development, 53, 15–24.
Chandler, P., & Sweller, J. (1991). Cognitive
load theory and the format of instruction. Cog-
nition and Instruction, 8, 293–332. doi:10.1207/
s1532690xci0804_2
Chandler, P., & Sweller, J. (1996). Cognitive load
while learning to use a computer program. Applied
Cognitive Psychology, 10, 151–170. doi:10.1002/
(SICI)1099-0720(199604)10:2<151::AID-
ACP380>3.0.CO;2-U
Cooper, G., & Sweller, J. (1987). Effects of schema
acquisition and rule automation on mathematical
problem-solving transfer. Journal of Educational
Psychology, 79, 347–362. doi:10.1037/0022-
0663.79.4.347
Cowan, N. (2001). The magical number 4 in short-
term memory: A reconsideration of metal storage
capacity. The Behavioral and Brain Sciences,
24, 87–114. doi:10.1017/S0140525X01003922
De Croock, M. B. M. (1999). The transfer para-
dox: Training design for troubleshooting skills.
Unpublished PhD thesis. Enschede, The Nether-
lands: University of Twente.
De Croock, M. B. M., van Merriënboer, J. J. G.,
& Paas, F. G. W. C. (1998). High vs. low contex-
tual interference in simulation-based training of
trouble-shooting skills: Effects on transfer per-
formance and invested mental effort. Computers
in Human Behavior, 14, 249–267. doi:10.1016/
S0747-5632(98)00005-3
De Jong, T. (2009). Cognitive load theory, edu-
cational research, and instructional design: Some
food for thought. Instructional Science, 38,
105–134. doi:10.1007/s11251-009-9110-0
DeLeeuw, K. E., & Mayer, R. E. (2008). A compari-
son of three measures of cognitive load: Evidence
for separable measures of intrinsic, extraneous, and
germane load. Journal of Educational Psychology,
100, 223–234. doi:10.1037/0022-0663.100.1.223
Duric, Z., Gray, W. D., Heishman, R., Li, F., Rosen-
feld, A., & Schoellles, M. J. (2002)... Proceedings
of the IEEE, 90(7), 1272–1289. doi:10.1109/
JPROC.2002.801449
Ericsson, K. A., & Simon, H. A. (1993). Protocol
analysis: Verbal reports as data. Cambridge, MA:
MIT Press.
Ginns, P. (2005). Meta-analysis of the modality
effect. Learning and Instruction, 15, 313–331.
doi:10.1016/j.learninstruc.2005.07.001
Ginns, P. (2006). Integrating information: A meta-
analysis of the spatial contiguity and temporal
contiguity effects. Learning and Instruction, 16,
511–525. doi:10.1016/j.learninstruc.2006.10.001
Gopher, D., & Braune, R. (1984). On the psycho-
physics of workload: Why bother with subjective
measures? Human Factors, 26, 519–532.
Granholm, E., Asarnow, R. F., Sarkin, A. J., &
Dykes, K. L. (1996). Pupillary responses index
cognitive resource limitations. Psychophysiol-
ogy, 33, 457–461. doi:10.1111/j.1469-8986.1996.
tb01071.x
Hart, S. G., & Staveland, L. E. (1988). Develop-
ment of NASA-TLX (Task Load Index): Results
of experimental and theoretical research. In
Hancock, P. A., & Meshkati, N. (Eds.), Human
mental workload (pp. 139–183). Amsterdam, The
Netherlands: North Holland. doi:10.1016/S0166-
4115(08)62386-9

1205
Cognitive Load Theory in E-Learning
Hegarty, M., & Just, M. (1993). Constructing men-
tal models of machines from text and diagrams.
Journal of Memory and Language, 32, 717–742.
doi:10.1006/jmla.1993.1036
Holmberg, B. (1989). Theory and practice of
distance education. New York, NY: Routledge.
Holmqvist, K., Nyström, M., Andersson, R.,
Dewhurst, R., Jarodzka, H., & Van de Weijer,
J. (2011). Eye tracking: A comprehensive guide
to methods and measures. Oxford, UK: Oxford
University Press.
Hyönä, J., Tommola, J., & Alaja, A.-M. (1996). Pu-
pil dilation as a measure of processing load in simul-
taneous interpretation and other language tasks.
The Quarterly Journal of Experimental Psychol-
ogy Section A: Human Experimental Psychology,
48, 598–612. doi:10.1080/14640749508401407
Jarodzka, H., Balslev, T., Holmqvist, K., Nyström,
M., Scheiter, K., Gerjets, P., & Eika, B. (2010).
Learning perceptual aspects of diagnosis in medi-
cine via eye movement modeling examples on
patient video cases. In S. Ohlsson & R. Catrambone
(Eds.), Proceedings of the 32nd Annual Conference
of the Cognitive Science Society (pp. 1703-1708).
Austin, TX: Cognitive Science Society.
Jarodzka, H., Janssen, N., Kirschner, P. A., &
Erkens, G. (2011). Avoiding split attention in
computer-based testing: Is neglecting additional
information facilitative? Manuscript submitted
for publication.
Jarodzka, H., Scheiter, K., Gerjets, P., Van Gog,
T., & Dorr, M. (2009). How to convey perceptual
skills by displaying experts’ gaze data. In N. A.
Taatgen, & H. van Rijn (Eds.), Proceedings of the
31st Annual Conference of the Cognitive Science
Society (pp. 2920-2925). Austin, TX: Cognitive
Science Society.
Jelsma, O. (1989). Instructional control of trans-
fer. Enschede, The Netherlands: Bijlstra & Van
Merriënboer.
Jeung, H., Chandler, P., & Sweller, J. (1997). The
role of visual indicators in dual sensory mode in-
struction. Educational Psychology, 17, 329–433.
doi:10.1080/0144341970170307
Kalyuga, S. (2007). Expertise reversal effect and
its implications for learner-tailored instruction.
Educational Psychology Review, 19, 509–539.
doi:10.1007/s10648-007-9054-3
Kalyuga, S., Ayres, P., Chandler, P., & Sweller,
J. (2003). The expertise reversal effect. Edu-
cational Psychologist, 38, 23–32. doi:10.1207/
S15326985EP3801_4
Kalyuga, S., Chandler, P., & Sweller, J.
(1998). Levels of expertise and instruc-
tional design. Human Factors, 40, 1–17.
doi:10.1518/001872098779480587
Kalyuga, S., Chandler, P., & Sweller, J.
(1999). Managing split-attention and redun-
dancy in multimedia instruction. Applied Cog-
nitive Psychology, 13, 351–371. doi:10.1002/
(SICI)1099-0720(199908)13:4<351::AID-
ACP589>3.0.CO;2-6
Kalyuga, S., Chandler, P., Touvien, J., & Sweller, J.
(2001). When problem solving is superior to study-
ing worked examples. Journal of Educational
Psychology, 93, 579–588. doi:10.1037/0022-
0663.93.3.579
Kalyuga, S., & Sweller, J. (2005). Rapid dynamic
assessment of expertise to improve the efficiency
of adaptive e-learning. Educational Technol-
ogy Research and Development, 53, 83–93.
doi:10.1007/BF02504800
Kester, L., Kirschner, P. A., Van Merriënboer, J. J.
G., & Bäumer, A. (2001). Just-in-time information
presentation and the acquisition of complex cogni-
tive skills. Computers in Human Behavior, 17,
373–391. doi:10.1016/S0747-5632(01)00011-5

1206
Cognitive Load Theory in E-Learning
Kirschner, F., Paas, F., & Kirschner, P. A. (2008).
Individual versus group learning as a function
of task complexity: An exploration into the
measurement of group cognitive load. In Zum-
bach, J., Schwartz, N., Seufert, T., & Kester, L.
(Eds.), Beyond knowledge: The legacy of com-
petence (pp. 21–28). The Netherlands: Springer.
doi:10.1007/978-1-4020-8827-8_4
Kirschner, F., Paas, F., & Kirschner, P. A. (2009a).
A cognitive load approach to collaborative learn-
ing: United brains for complex tasks. Educational
Psychology Review, 21, 31–42. doi:10.1007/
s10648-008-9095-2
Kirschner, F., Paas, F., & Kirschner, P. A. (2009b).
Individual and group-based learning from complex
cognitive tasks: Effects on retention and transfer
efficiency. Computers in Human Behavior, 25,
306–314. doi:10.1016/j.chb.2008.12.008
Kirschner, F., Paas, F., & Kirschner, P. A. (2010a).
Task complexity as a driver for collaborative learn-
ing efficiency: The collective working-memory
effect. Applied Cognitive Psychology, 25(4).
doi:doi:10.1002/acp.1730
Kirschner, F., Paas, F., & Kirschner, P. A. (2010b).
Differential effects of problem-solving demands
on individual and collaborative learning out-
comes. Learning and Instruction, 21, 587–599.
doi:10.1016/j.learninstruc.2011.01.001
Kirschner, F., Paas, F., & Kirschner, P. A. (2011).
Superiority of collaborative learning with complex
tasks: A research note on an alternative affective
explanation. Computers in Human Behavior, 27,
53–57. doi:10.1016/j.chb.2010.05.012
Klingner, J. (2010). Fixation-aligned pupillary
response averaging. Paper presented at the Eye
Tracking Research & Applications, Austin, TX,
USA.
Klingner, J., Kumar, R., & Hanrahan, P. (2008).
Measuring the task-evoked pupillary response
with a remote eye tracker. In K. -J. Räihä & A. T.
Duchowski (Eds.), Proceedings of the 2008 Sym-
posium on Eye Tracking Research & Applications
(pp. 69-72). New York, NY: ACM.
Klingner, J., Tversky, B., & Hanrahan, P. (2011).
Effects of visual and verbal presentation on
cognitive load in vigilance, memory, and arith-
metic tasks. Psychophysiology, 48, 323–332.
doi:10.1111/j.1469-8986.2010.01069.x
Leahy, W., & Sweller, J. (2011). Cognitive load
theory, modality of presentation and the transient
information effect. Applied Cognitive Psychology,
25(6). doi:10.1002/acp.1787
Leyman, E., Mirka, G., Kaber, D., & Sommerich,
C. (2004). Cervicobrachial muscle response to
cognitive load in a dual-task scenario. Ergonom-
ics, 47, 625–645. doi:10.1080/00140130310001
629766
Marcus, N., Cooper, M., & Sweller, J. (1996).
Understanding instructions. Journal of Educa-
tional Psychology, 88, 49–63. doi:10.1037/0022-
0663.88.1.49
Mayer, R., Bove, W., Bryman, A., Mars, R.,
Tapangco, L. R., & Peterson, M. J. (1959). Short-
term retention of individual verbal items. When
less is more: Meaningful learning from visual and
verbal summaries of science textbook lessons.
Journal of Experimental Educational Psychol-
ogy, 58, 193–198.
Mayer, R. E. (2009). Multimedia learning. Cam-
bridge, UK: Cambridge University Press.
Mayer, R. E., Morene, R., Boir, M., Vagge,
S., & Sheffey, S. (2002). Shared information,
cognitive load and group memory. Group
Processes & Intergroup Relations, 5, 5–18.
doi:10.1177/1368430202005001535

1207
Cognitive Load Theory in E-Learning
Mayer, R. E., Moreno, R., Boire, M., & Vagge, S.
(1999b). Maximizing constructivist learning from
multimedia communications by minimizing cog-
nitive load. Journal of Educational Psychology,
91, 638–643. doi:10.1037/0022-0663.91.4.638
Miller, G. (1956). The magical number seven, plus
or minus two: Some limits on our capacity for
processing information. Psychological Review,
63, 81–97. doi:10.1037/h0043158
Moreno, R., Mayer, R. E., Spires, H. A., & Les-
ter, J. C. (2001). The case for social agency in
computer based teaching: Do students learn more
deeply when they interact with animated peda-
gogical agents? Cognition and Instruction, 19(2),
177–213. doi:10.1207/S1532690XCI1902_02
Moreno, R., & Valdez, A. (2005). Research on
cognitive load theory and its design implications
for e-learning. Cognitive load and learning effects
of having learners organize pictures and words
in multimedia environments: The role of student
interactivity and feedback. Educational Tech-
nology Research and Development, 53, 35–44.
doi:10.1007/BF02504796
Moresi, S., Adam, J. J., Rijcken, J., Van Gerven,
P. W. M., Kuipers, H., & Jolles, J. (2007). Pupil
dilation in response preparation. International
Journal of Psychophysiology, 67, 124–130.
doi:10.1016/j.ijpsycho.2007.10.011
Mousavi, S. Y., Low, R., & Sweller, J. (1995).
Reducing cognitive load by mixing auditory and
visual presentation modes. Journal of Educational
Psychology, 87, 319–334. doi:10.1037/0022-
0663.87.2.319
Ohtsubo, Y. (2005). Should information be re-
dundantly distributed among group members?
Effective use of group memory in collaborative
problem solving. Applied Cognitive Psychology,
19, 1219–1233. doi:10.1002/acp.1162
Paas, F., Ayres, P., & Pachman, M. (2008). As-
sessment of cognitive load in multimedia learning.
In Robinson, D. H., & Schraw, G. (Eds.), Recent
innovations in educational technology that facili-
tate student learning (pp. 11–35). Charlotte, NC:
Information Age Publishing Inc.
Paas, F., Tuovinen, J., Van Merriënboer, J. J. G.,
& Darabi, A. (2005). A motivational perspec-
tive on the relation between mental effort and
performance: Optimizing learner involvement in
instruction. Educational Technology Research
and Development, 53, 25–33. doi:10.1007/
BF02504795
Paas, F., Tuovinen, J. E., Tabbers, H., & Van
Gerven, P. W. M. (2003). Cognitive load mea-
surement as a means to advance cognitive load
theory. Educational Psychologist, 38, 63–71.
doi:10.1207/S15326985EP3801_8
Paas, F. G. W. C. (1992). Training strategies
for attaining transfer of problem-solving skill
in statistics: A cognitive load approach. Jour-
nal of Educational Psychology, 84, 429–434.
doi:10.1037/0022-0663.84.4.429
Paas, F. G. W. C. (1993). Instructional control of
cognitive load in the training of complex cognitive
tasks. Unpublished PhD thesis. Enschede, The
Netherlands: University of Twente.
Paas, F. G. W. C., & van Merriënboer, J. J. G.
(1993). The efficiency of instructional conditions:
An approach to combine mental effort and perfor-
mance measures. Human Factors, 35, 737–743.
Paas, F. G. W. C., & van Merriënboer, J. J. G.
(1994). Variability of worked examples and
transfer of geometrical problem solving skills: A
cognitive load approach. Journal of Educational
Psychology, 86, 122–133. doi:10.1037/0022-
0663.86.1.122

1208
Cognitive Load Theory in E-Learning
Paas, F. G. W. C., Van Merriënboer, J. J. G.,
& Adam, J. J. (1994). Measurement of cogni-
tive load in instructional research. Perceptual
and Motor Skills, 79, 419–430. doi:10.2466/
pms.1994.79.1.419
Paivio, A. (1991). Dual coding theory: Retrospect
and current status. Canadian Journal of Psychol-
ogy, 45, 255–287. doi:10.1037/h0084295
Pollock, E., Chandler, P., & Sweller, J. (2002).
Assimilating complex information. Learning
and Instruction, 12, 61–86. doi:10.1016/S0959-
4752(01)00016-0
Reed, W. M., Burton, J. K., & Kelly, P. (1985). The
effects of writing ability and mode of discourse
on cognitive capacity engagement. Research in
the Teaching of English, 19, 283–297.
Renkl, A. (2002). Worked-out examples: In-
structional explanations support learning by
self-explanations. Learning and Instruction, 12,
529–556. doi:10.1016/S0959-4752(01)00030-5
Renkl, A. (2011). Towards an instructionally
oriented theory of example-based learning. Manu-
script submitted for publication.
Renkl, A., Stark, R., Gruber, H., & Mandl, H.
(1998). Learning from worked-out examples:
The effects of example variability and elicited
self-explanations. Contemporary Educational
Psychology Review, 2, 123–138.
Schooler, J. W., & Engstler-Schooler, L. (1990).
Verbal overshadowing of visual memories: Some
things are better left unsaid. Cognitive Psychology,
22, 36–71. doi:10.1016/0010-0285(90)90003-M
Schuurman, J. G. (1999). On the proper treatment
of learning and transfer: A study on introduc-
tory computer programming. Unpublished PhD
Thesis. Enschede, The Netherlands: Universiteit
of Twente.
Schuurman, J. G. (1999). On the proper treatment
of learning and transfer: A study on introduc-
tory computer programming. Unpublished PhD
Thesis. Enschede, The Netherlands: Universiteit
of Twente.
Stark, R. (1999). Lernen mit Lösungsbeispielen.
Der Einfluß unvollständiger Lösungsschritte auf
Beispielelaboration, Motivation und Lernerfolg
[Learning by worked-out examples. The impact
of incomplete solution steps on example elabora-
tion, motivation, and learning outcomes.]. Bern,
Switzerland: Huber.
Struve, D. (2008). Process-oriented worked ex-
amples for training older adults to use interactive
systems. Gerontechnology (Valkenswaard), 7,
216. doi:10.4017/gt.2008.07.02.153.00
Sweller, J. (1988). Cognitive load during problem
solving: Effects on learning. Cognitive Science,
12, 257–285. doi:10.1207/s15516709cog1202_4
Sweller, J. (2010). Element interactivity and in-
trinsic, extraneous, and germane cognitive load.
Educational Psychology Review, 2, 123–138.
doi:10.1007/s10648-010-9128-5
Sweller, J., Ayres, P., & Kalyuga, S. (2011). Mea-
suring cognitive load. In Spector, J. M., & LaJoie,
S. (Eds.), Cognitive load theory (pp. 71–85). New
York, NY: Springer. doi:10.1007/978-1-4419-
8126-4_6
Sweller, J., & Cooper, G. A. (1985). The use of
worked examples as a substitute for problem solv-
ing in learning algebra. Cognition and Instruction,
2, 59–89. doi:10.1207/s1532690xci0201_3
Sweller, J., Mawer, R. F., & Ward, M. R. (1983).
Development of expertise in mathematical prob-
lem solving. Journal of Experimental Psychol-
ogy. General, 112, 639–661. doi:10.1037/0096-
3445.112.4.639

1209
Cognitive Load Theory in E-Learning
Sweller, J., Van Merrienboer, J. J. G., & Paas, F.
(1998). Cognitive architecture and instructional
design. Educational Psychology Review, 10,
251–296. doi:10.1023/A:1022193728205
Tarmizi, R. A., & Sweller, J. (1988). Guidance
during mathematical problem solving. Jour-
nal of Educational Psychology, 80, 424–436.
doi:10.1037/0022-0663.80.4.424
Tindale, R., Mayer, R. E., & Moreno, R. (1998).
A split-attention effect in multimedia learning:
Evidence for dual processing systems in working
memory. Journal of Educational Psychology, 90,
312–320. doi:10.1037/0022-0663.90.2.312
Tindall-Ford, S., Chandler, P., & Sweller, J. (1997).
When two sensory modes are better than one.
Journal of Experimental Psychology. Applied, 3,
257–287. doi:10.1037/1076-898X.3.4.257
Tomasio, D., Chang, L., Caparellia, E. C., &
Ernst, T. (2007). Different activation patterns for
working memory load and visual attention load.
Brain Research, 1132, 158–165. doi:10.1016/j.
brainres.2006.11.030
Underwood, G., Jebbett, L., & Roberts, K. (2004).
Inspecting pictures for information to verify a
sentence: Eye movements in general encoding
and in focused search. The Quarterly Journal of
Experimental Psychology Section A, 57, 165–182.
doi:10.1080/02724980343000189
Van Gerven, P. W. M., Paas, F., Van Merriënboer,
J. J. G., & Schmidt, H. G. (2004). Memory load
and the cognitive pupillary response in aging. Psy-
chophysiology, 41, 167–174. doi:10.1111/j.1469-
8986.2003.00148.x
Van Gerven, P. W. M., Paas, F., Van Merriënboer,
J. J. G., & Schmidt, H. G. (2006). Modality and
variability as factors in training elderly. Applied
Cognitive Psychology, 20, 311–320. doi:10.1002/
acp.1247
Van Gog, T., Ericsson, K. A., Rikers, R. M. J.
P., & Paas, F. (2005). Instructional design for
advanced learners: Establishing connections
between the theoretical frameworks of cognitive
load and deliberate practice. Educational Tech-
nology Research and Development, 53, 73–81.
doi:10.1007/BF02504799
Van Gog, T., Jarodzka, H., Scheiter, K., Gerjets,
P., & Paas, F. (2009). Attention guidance during
example study via the model‘s eye movements.
Computers in Human Behavior, 25, 785–791.
doi:10.1016/j.chb.2009.02.007
Van Gog, T., Kester, L., Nievelstein, F., Giesbers,
B., & Paas, F. (2009). Uncovering cognitive
processes: Different techniques that can contrib-
ute to cognitive load research and instruction.
Computers in Human Behavior, 25, 325–331.
doi:10.1016/j.chb.2008.12.021
Van Gog, T., & Paas, F. (2008). Instructional
efficiency: Revisiting the original construct in
educational research. Educational Psychologist,
43, 16–26. doi:10.1080/00461520701756248
Van Gog, T., Paas, F., & Van Merriënboer, J. J.
G. (2004). Process-oriented worked examples:
Improving transfer performance through enhanced
understanding. Instructional Science, 32, 83–98.
doi:10.1023/B:TRUC.0000021810.70784.b0
Van Gog, T., Paas, F., & Van Merriënboer, J. J.
G. (2006). Effects of process-oriented worked
examples on troubleshooting transfer perfor-
mance. Learning and Instruction, 16, 154–164.
doi:10.1016/j.learninstruc.2006.02.003
Van Gog, T., Paas, F., & Van Merrienboer, J. J. G.
(2007). Effects of studying sequences of process-
oriented and product-oriented worked examples
on troubleshouting transfer efficiency. Learning
and Instruction, 18, 211–222. doi:10.1016/j.
learninstruc.2007.03.003

1210
Cognitive Load Theory in E-Learning
Van Gog, T., & Rummel, N. (2010). Example-
based learning: Integrating cognitive and social-
cogntive research perspectives. Educational
Psychology Review, 22, 155–174. doi:10.1007/
s10648-010-9134-7
Van Gog, T., & Scheiter, K. (2010). Eye track-
ing as a tool to study and enhance multimedia
learning. Learning and Instruction, 10, 95–99.
doi:10.1016/j.learninstruc.2009.02.009
Van Lehn, K. (1996). Cognitive Skill Acquisi-
tion. Annual Review of Psychology, 47, 513–539.
doi:10.1146/annurev.psych.47.1.513
Van Merriënboer, J. J. G. (1997). Training complex
cognitive skills: A four-component instructional
design model for technical training. Englewood
Cliffs, NJ: Educational Technology Publications.
Van Merriënboer, J. J. G., & Ayres, P. (2005).
Research on cognitive load theory and its design
implications for e-learning. Educational Tech-
nology Research and Development, 53(3), 5–13.
doi:10.1007/BF02504793
Van Merriënboer, J. J. G., Clark, R. E., & De
Croock, M. B. M. (2002). Blueprints for complex
learning: The 4C/ID-model. Educational Tech-
nology Research and Development, 50, 39–64.
doi:10.1007/BF02504993
Van Merriënboer, J. J. G., & De Croock, M. B.
M. (1992). Strategies for computer-based pro-
gramming instruction: Program completion vs.
program generation. Journal of Educational
Computing Research, 8, 365–394. doi:10.2190/
MJDX-9PP4-KFMT-09PM
Van Merriënboer, J. J. G., & Kirschner, P. A.
(2007). Ten steps to complex learning: A systematic
approach to four-component instructional design.
Mahwah, NJ: Lawrence Erlbaum Associates.
Van Merriënboer, J. J. G., Kirschner, P. A., &
Kester, L. (2003). Taking the load off a learner’s
mind: Instructional design for complex learning.
Educational Psychologist, 38, 5–13. doi:10.1207/
S15326985EP3801_2
Van Merriënboer, J. J. G., & Kramer, H. P. M.
(1990). The “completion strategy” in program-
ming instruction: Theoretical and empirical sup-
port. In S. Dijkstra, B. H. M. Van Hout-Wolters, &
P. C. Van der Sijde (Eds.), Research on instruction
(pp. 45-61). Englewood Cliffs, NJ: Educational
Technology Publications.
Van Merriënboer, J. J. G., Schuurman, J. G., De
Croock, M. B. M., & Paas, F. G. W. C. (2002).
Redirection learners’ attention during training: Ef-
fects on cognitive load, transfer test performance
and training efficiency. Learning and Instruction,
12, 11–37. doi:10.1016/S0959-4752(01)00020-2
Van Mierlo, C. M., Jarodzka, H., & Kirschner, P.
A. (2011, August). Sex differences in e-navigation.
Poster session presented at the 16th edition of a
series of European Conferences on Eye Move-
ments in Marseille, France.
Van Vugt, M., De Cremer, D., & Janssen, D. P.
(2007). Gender differences in cooperation and
competition: The male-warrior hypothesis. Psy-
chological Science, 18(1), 19–23. doi:10.1111/
j.1467-9280.2007.01842.x
Verwey, W. B., & Veltman, H. A. (1996). Detecting
short periods of elevated workload: A comparison
of nine workload assessment techniques. Journal
of Experimental Psychology. Applied, 2, 270–285.
doi:10.1037/1076-898X.2.3.270
Wallen, E., Plass, J. L., & Brünken, R. (2005).
The function of annotations in the comprehension
of scientific texts: Cognitive load effects and the
impact of verbal ability. Educational Technol-
ogy Research and Development, 53, 59–71.
doi:10.1007/BF02504798

1211
Cognitive Load Theory in E-Learning
Ward, M., & Sweller, J. (1990). Structuring effec-
tive worked examples. Cognition and Instruction,
7, 1–39. doi:10.1207/s1532690xci0701_1
Wierwille, W. W., & Eggemeier, F. L. (1993).
Recommendations for mental workload measure-
ment in a test and evaluation environment. Human
Factors, 35, 263–281.
Yamane, D. (1996). Collaboration and its dis-
contents: Steps toward overcoming barriers to
successful group projects. Teaching Sociology,
24, 378–383. doi:10.2307/1318875
Yin, B., & Chen, F. (2007). Towards automatic
cognitive load measurement from speech analy-
sis. Human-Computer Interaction: Interac-
tion, Design, and Usability, 4550, 1011–1020.
doi:10.1007/978-3-540-73105-4_111
Zhu, X., & Simon, H. A. (1987). Learning math-
ematics from examples and by doing. Cogni-
tion and Instruction, 4, 137–166. doi:10.1207/
s1532690xci0403_1
ADDITIONAL READING
Jarodzka, H., Janssen, N., Kirschner, P. A., &
Erkens, G. (2011). Avoiding split attention in
computer-based testing: Is neglecting additional
information facilitative? Manuscript submitted
for publication.
Kirschner, F., Paas, F., & Kirschner, P. A. (2009a).
A cognitive load approach to collaborative learn-
ing: United brains for complex tasks. Educational
Psychology Review, 21, 31–42. doi:10.1007/
s10648-008-9095-2
Van Merriënboer, J. J. G., & Ayres, P. (2005).
Research on cognitive load theory and its design
implications for e-learning. Educational Tech-
nology Research and Development, 53(3), 5–13.
doi:10.1007/BF02504793
Van Merriënboer, J. J. G., Clark, R. E., & De
Croock, M. B. M. (2002). Blueprints for complex
learning: The 4C/ID-model. Educational Tech-
nology Research and Development, 50, 39–64.
doi:10.1007/BF02504993
Van Merriënboer, J. J. G., & Kirschner, P. A.
(2007). Ten steps to complex learning: A systematic
approach to four-component instructional design.
Mahwah, NJ: Lawrence Erlbaum Associates.
Van Merriënboer, J. J. G., Kirschner, P. A., &
Kester, L. (2003). Taking the load off a learner’s
mind: Instructional design for complex learning.
Educational Psychologist, 38, 5–13. doi:10.1207/
S15326985EP3801_2
Van Mierlo, C. M., Jarodzka, H., & Kirschner, P.
A. (2011, August). Sex differences in e-navigation.
Poster session presented at the 16th edition of a
series of European Conferences on Eye Move-
ments in Marseille, France.