This article was downloaded by: [Harvard College]

On: 27 July 2013, At: 03:46

Publisher: Routledge
Informa Ltd Registered in England and Wales Registered Number: 1072954 Registered
office: Mortimer House, 37-41 Mortimer Street, London W1T 3JH, UK

Cognition and Instruction

Publication details, including instructions for authors and subscription
information:
http://www.tandfonline.com/loi/hcgi20

Cognitive Load Theory and the Format of

Instruction
Paul Chandler & John Sweller
Published online: 14 Dec 2009.

To cite this article: Paul Chandler & John Sweller (1991) Cognitive Load Theory and the Format of
Instruction, Cognition and Instruction, 8:4, 293-332, DOI: 10.1207/s1532690xci0804_2

To link to this article: http://dx.doi.org/10.1207/s1532690xci0804_2

PLEASE SCROLL DOWN FOR ARTICLE

Taylor & Francis makes every effort to ensure the accuracy of all the information (the
“Content”) contained in the publications on our platform. However, Taylor & Francis, our
agents, and our licensors make no representations or warranties whatsoever as to the
accuracy, completeness, or suitability for any purpose of the Content. Any opinions and
views expressed in this publication are the opinions and views of the authors, and are not
the views of or endorsed by Taylor & Francis. The accuracy of the Content should not be
relied upon and should be independently verified with primary sources of information. Taylor
and Francis shall not be liable for any losses, actions, claims, proceedings, demands, costs,
expenses, damages, and other liabilities whatsoever or howsoever caused arising directly or
indirectly in connection with, in relation to or arising out of the use of the Content.

This article may be used for research, teaching, and private study purposes. Any substantial
or systematic reproduction, redistribution, reselling, loan, sub-licensing, systematic supply,
or distribution in any form to anyone is expressly forbidden. Terms & Conditions of access
and use can be found at http://www.tandfonline.com/page/terms-and-conditions
 COGNITION AND INSTRUCTION, 1991, 8(4), 293-332
Copyright @ 1991, Lawrence Erlbaum Associates, Inc.

Cognitive Load Theory and the

Format of Instruction
Paul Chandler and John Sweller
University of New South Wales

Cognitive load theory suggests that effective instructional material facilitates learn-
Downloaded by [Harvard College] at 03:46 27 July 2013

ing by directing cognitive resources toward activities that are relevant to learning
rather than toward preliminaries to learning. One example of ineffective instruc-
tion occurs if learners unnecessarily are required to mentally integrate disparate
sources of mutually referring information such as separate text and diagrams. Such
split-source information may generate a heavy cognitive load, because material must
be mentally integrated before learning can commence. This article reports findings
from six experiments testing the consequences of split-source and integrated infor-
mation using electrical engineering and biology instructional materials. Experiment
1 was designed to compare conventional instructions with integrated instructions
over a period of several months in an industrial training setting. The materials chosen
were unintelligible without mental integration. Results favored integrated instruc-
tions throughout the 3-month study. Experiment 2 was designed to investigate the
possible differences between conventional and integrated instructions in areas in
which it was not essential for sources of information to be integrated to be under-
stood. The results suggest that integrated instructions were no better than split-source
information in such areas. Experiments 3, 4, and 5 indicate that the introduction
of seemingly useful but nonessential explanatory material (e.g., a commentary on
a diagram) could have deleterious effects even when presented in integrated for-
mat. Experiment 6 found that the need for physical integration was restored if the
material was organized in such a manner that individual units could not be understood
alone. In light of these results and previous findings, suggestions are made for
cognitively guided instructional packages.

Over the last decade, there have been considerable interest and debate in areas
of cognition and education. Nevertheless, until recently, our knowledge of the
cognitive processes involved in understanding instructional material has been some-
what limited. In the last few years, however, cognitive science has progressed
to a point where it is becoming obvious that traditional methods of instructional
-
Requests for reprints should be sent to John Sweller, School of Education, University of New
South Wales, P.O. Box 1, Kensington, New South Wales, Australia 2033.
 294 CHANDLER AND SWELLER

design based on visual elegance, common sense, and convenience are inadequate.
Recently, new instructional procedures guided by cognitive theory have become
available. In this article, we present findings that have implications for the presen-
tation of instructional materials. The experiments reported were generated by cog-
nitive load theory, which is discussed in the next section.

COGNITIVE LOAD THEORY

AND INSTRUCTIONAL PROCESSES

Cognitive load theory (see Sweller, 1988, 1989) is concerned with the manner
in which cognitive resources are focused and used during learning and problem
solving. Many learning and problem-solving procedures encouraged by instruc-
tional formats result in students engaging in cognitive activities far removed from
Downloaded by [Harvard College] at 03:46 27 July 2013

the ostensible goals of the task. The cognitive load generated by these irrelevant
activities can impede skill acquisition.
The lack of concordance between the cognitive demands of some tasks and
the goals of those tasks first became apparent in studies concerned with relations
between learning and problem solving (e.g., see Sweller, Mawer, & Howe, 1982;
Sweller, Mawer, & Ward, 1983). Subjects could solve problems, in some cases
repeatedly solve problems, and remain oblivious to their essential structure. It
was theorized that the search strategies used, although important in attaining
problem goals, were ineffective as learning devices. The extraneous cognitive
load imposed by the problem-solving strategy interfered with learning.
There are many experiments demonstrating that conventional problem solv-
ing can have negative learning consequences. This body of evidence questions
the usefulness of solving large numbers of conventional problems (in the areas
of mathematics and science, see Cooper & Sweller, 1987; Owen & Sweller, 1985;
Sweller & Cooper, 1985; Sweller et al., 1983; Ward & Sweller, 1990).
The use of worked examples is one technique designed to circumvent the in-
terference with learning caused by some forms of problem solving. Sweller and
Cooper (1985) and Cooper and Sweller (1987) found that a heavy use of worked
examples resulted in more rapid learning than the conventional emphasis on solving
a large number of problems. In a longitudinal study, Zhu and Simon (1987) found
worked examples to be highly effective. Chi, Bassok, Lewis, Reimann, and Glaser
(1989) found more able students were better than less able students at generating
detailed explanations of worked examples and had a greater awareness of com-
prehension failures. These results indicate the importance to learning of an abili-
ty to properly process worked examples.
Cognitive load theory has been used to explain why studying worked exam-
ples can facilitate learning compared with problem solving (e.g., see Cooper &
Sweller, 1987; Sweller, 1988). In essence, searching for suitable problem-solving
 COGNITIVE LOAD THEORY 295

operators is cognitively demanding and directs attention away from aspects of

the problem important to learning. In contrast, many worked examples are far
easier to process than the equivalent problems and direct attention more appropri-
ately. Nevertheless, not all worked examples appropriately direct attention and
reduce cognitive load.
Tarmizi and Sweller (1988), using geometry, and Ward and Sweller (1990),
using physics, explained how worked examples in some areas could be ineffec-
tive. Many worked examples consist of multiple sources of mutually referring
information. Diagrams and sets of explanatory statements provide a very com-
mon example. Most geometry worked examples consist of a diagram and a set
of statements describing relations between entities on the diagram. Little mean-
ing may be extracted from the diagram alone, and the statements may be totally
meaningless in isolation. To follow such examples, attention must be focused
on mentally integrating the various sources of information, for alone they are
Downloaded by [Harvard College] at 03:46 27 July 2013

unintelligible. Mental integration requires searching for appropriate referents in

the diagram or set of statements and understanding that a statement such as "an-
gle ABC" is identical to a particular entity in the diagram. This search process
can be expected to have the same consequences as searching for operators to solve
a problem. Attention is misdirected, and cognitive resources are devoted to an
activity that, although an essential precursor to learning, is itself unrelated to learn-
ing. If cognitive effort is devoted to activities other than learning, we might ex-
pect interference with learning.
Tarmizi and Sweller (1988) and Ward and Sweller (1990) obtained results in
accordance with the cognitive load hypothesis as applied to worked examples.
They found that worked examples requiring learners to split their attention be-
tween multiple sources of information were no more effective and possibly less
effective than problem solving. It was hypothesized that, if such worked exam-
ples are reformatted to reduce or eliminate split attention, the reduction in cogni-
tive load due to the reduction in the need to search for relevant referents and
mentally integrate them should result in effective worked examples. For exam-
ple, by physically integrating geometry statements with the diagram, mental
integration is no longer necessary, and the cognitive load involved in mental in-
tegration should be eliminated and learning enhanced. Many experiments con-
firmed this suggestion.
In summary, we can hypothesize that presentation techniques frequently result
in high levels of extraneous cognitive load that influence the degree to which learn-
ing can be facilitated. It follows that information should be presented in ways
that do not impose a heavy extraneous cognitive load. Problem solving frequent-
ly fails to meet this criterion. Alternatives to conventional problem solving, such
as worked examples, are only useful if they also meet this criterion. For this rea-
son, worked examples that require learners to mentally integrate multiple sources
of information are ineffective.
 296 CHANDLER AND SWELLER

USING COGNITIVE LOAD THEORY

TO DESIGN INITIAL INSTRUCTION

We have argued that a greater than normal use of worked examples is essential
to cognitively guided instruction. However, worked examples are only one part
of instruction. Normally, they are preceded by introductory explanatory materi-
als. The structure of these introductory materials usually is determined by such
variables as visual elegance, convenience, and tradition. The use of multiple
sources of mutually referring information is frequent. For example, it is com-
mon for diagrams and text to be presented separately. With separate diagrams
and text, the information often is unintelligible prior to being mentally integrated.
The same cognitive load principles should apply to initial instruction and to
worked examples. Mental integration is likely to be cognitively taxing whether
required for a worked example or for initial instruction. Ideal formats for initial
Downloaded by [Harvard College] at 03:46 27 July 2013

instruction should reduce extraneous cognitive load. For this reason, we can
hypothesize that all disparate sources of information should be physically integrated
where they can save learners from performing unnecessary mental integrations
that interfere with learning. Sweller, Chandler, Tierney , and Cooper (1990) tested
this hypothesis. Using coordinate geometry and numerical control programming,
they found that, if initial explanatory instructions were presented in conventional
split-source format, learning was substantially slower than if the same materials
were presented using a unified format. Cognitive load theory was used to gener-
ate the relevant experiments.
Under what conditions might the physical integration of disparate sources of
information not be beneficial? We hypothesize that physical integration is impor-
tant only where the disparate sources of information are unintelligible unless in-
tegrated. The requirement to mentally integrate imposes a heavy cognitive load.
For example, the statements associated with a geometry proof are likely to be
meaningless unless integrated with the diagram. Nevertheless, multiple sources
of information frequently are quite intelligible without being integrated. A source
of information may be used as a self-contained unit designed to buttress another
unit of information. For example, the contents of a diagram may be repeated in
textual form. The diagram and the text may be independently intelligible. The
information contained in each may be functionally redundant with mental integra-
tion unnecessary. Under these circumstances, there are no theoretical reasons
for supposing that physical or mental integration is likely to be beneficial.
The current set of experiments was generated by cognitive load theory. These
experiments were designed to test whether the physical integration of split-source
introductory materials is beneficial when the individual components are unintel-
ligible in isolation but irrelevant when the individual components can be under-
stood in isolation. It might be noted that the experiments were not designed
to test the validity of cognitive load theory. They were designed to test whether
the theory could generate findings with direct applications. In this article, our
 COGNITIVE LOAD THEORY 297

only concern is whether cognitive load theory can be used to design effective
i~istruction.
Experiments 1 and 2 were conducted in industrial training settings over a
3-month period, using instructional material relevant to electrical apprentices.
Experiment 1, using the area of installation testing, was designed to test the
hypothesis that integrated instructional materials are superior to conventional,
split-source instructional formats where the material must be integrated for un-
derstanding. The instructional materials for Experiment 2, based on lighting wir-
ing, were different from Experiment 1 in that the sources of information did not
have to be mentally integrated in order to be understood. Experiments 3 and 4
evolved from the findings of Experiment 2. Together, they investigated the pos-
sibility that, in some instructional areas, a single source of information is superi-
or to multiple sources of information, in either an integrated or conventional,
split-attention format. Experiments 5 and 6 attempted to replicate some of the
Downloaded by [Harvard College] at 03:46 27 July 2013

basic findings of the previous experiments using a totally different area-the

process of blood circulation in the human body.

EXPERIMENT 1

This experiment was designed to compare conventional introductory instruction-

al materials with integrated instructions using elementary electrical engineering.
When electrical equipment or new wiring is installed, it must be tested to ensure
that the installation procedure has been carried out correctly. This procedure,
called installation testing, provided materials for the first experiment.
The experiment used detailed instructional booklets and three separate test peri-
ods over a 12-week period. It was conducted in the training center of a major
Sydney company. Two groups were used. The first received introductory instruc-
ti~onalnotes on installation testing in a conventional split-source format. The sec-
ond group was presented with instructional notes in an integrated format. The
integrated materials contained virtually identical information to the conventional
instructions, with the only difference being the format of the presentation. It was
predicted that integrated instructions designed to focus attention appropriately and
reduce cognitive load would be superior to conventional instructions.

Method

Subjects. The subjects were 28 first-year trade apprentices from a Sydney,

Australia, company. All subjects had completed at least Year 10 of high school
and were enrolled in first-year trade courses at various technical colleges. None
of the subjects used in the experiment had any prior formal training in electrical
areas.
 Materials. The instructional materials consisted of two sets of introductory
instructional notes (conventional and modified) on installation testing. Both sets
of notes were designed to introduce apprentices to the equipment used in installa-
tion testing and to demonstrate how this equipment could be used for various
electrical tests. Specifically, there were notes on megger meters (a meter de-
signed to measure insulation and circuit resistance), multimeters (a meter designed
to measure voltage, current, and resistance), conductors in permanent wiring,
electrical continuity of earthing systems, and procedures for testing electrical
appliances.
The conventional instructional notes, in the form of a 10-page booklet, were
very similar to the regular instructions used by the company as part of its electri-
cal training program. The very limited revisions made to the instructions were
only in the interest of clarifying minor ambiguities. These notes were in a con-
ventional, split-source format; text and diagrams were presented separately, and
Downloaded by [Harvard College] at 03:46 27 July 2013

sources of related textual information also were presented separately. An exam-

ple of the instructions from the conventional notes is presented in Figure 1. The
modified instructional notes contained the same information but in an integrated
format. Mutually referring sources of information were integrated, where possi-
ble, into unitary sources of information. An example of the instructions from
the modified notes is contained in Figure 2.
The test materials consisted of a test booklet, as well as the equipment for
practical tests. The test booklet was divided into four sections covering questions
on (a) conductors in permanent wiring, (b) continuity of earthing systems, (c)
testing of electrical appliances, and (d) multimeters. The questions for the first
three of these required subjects to recall the installation testing procedure for each
of the areas. In each of these three areas, the subjects were asked (a) to indicate
to what the megger meter should be set to carry out the test, (b) to indicate on
diagrams how they would conduct the tests, and (c) to indicate the required results
for the tests. The questions relating to the use of the multimeter concerned (a)
the function of a multimeter, (b) safety precautions when using the meter, and
(c) procedures for using a multimeter. In total, there were 28 questions. All ques-
tions were designed to be completely objective and were marked as either cor-
rect or incorrect. One mark was allocated for a correct response, and no mark
was given for an incorrect response.
There were two practical tests. The first involved testing the safety of an elec-
trical appliance. The materials used for this test were an electric kettle with lead
and a megger meter. Four safety tests were required. Each of the four tests in-
volved placing the leads of the megger meter on appropriate parts of the kettle
and comparing it with the required regulation reading. For example, to test the
earth continuity of the appliance, one lead of the megger needed to be placed
on the earth pin while the other lead of the megger meter was on the frame of
the kettle. The required reading by regulation for this test is 2 ohms or less. For
each of the four safety tests, one mark was allocated for successfully performing
 COGNITIVE LOAD THEORY 299

l NSULATlON RESISTANCE TESTS

a) CONDUCTORS IN PERMANENT WIRING

Test : To test Insulation Resistance fmm conductors to earth

How conducted : i ) Disconnect appliances and busways during these tests. Make sure main
switch is "on" and all fusesare "in". Remove main earth fmm neutral bar and
set meter to read insulation. Connect one lead to earth wire at MEN bar and
take first measure by connecting the other lead to the active. Take next measure
by connecting the lead to the neutral.
ii) If resistance is not high enough in either of the two tests in i) then measure
each circuit sepamtely.
Results required : i) At least One Megaohm
ii) Same result as i) above
Downloaded by [Harvard College] at 03:46 27 July 2013

FIGURE 1 Example of the instructional format presented to the conventional group of Ex-
periment 1. A = active, N = neutral, and E = earth. All subjects were aware of these ab-
breviations prior to the experiment.

the test and one mark was given for successfully recalling the required reading.
This gave a total mark out of eight for the first practical test.
The second practical test involved testing the safety of an intermediate light
switch. (An intermediate light switch is used in long corridors and hallways. It
also is used for the lighting of stairs in multistory buildings.) A megger meter
was again required in order to conduct insulation resistance and earth continuity
 300 CHANDLER AND SWELLER

INSULATION RESISTANCE TESTS

a) CONDUCTORS IN PERMANENT WIRING

If resistance is not at least One

Downloaded by [Harvard College] at 03:46 27 July 2013

result for this

test and all other
tesls is at Icast
One Megiaohrn

FIGURE 2 Example of the instructional format presented to the modified group of Experi-
ment 1 .

tests. Each test involved placing the leads of the megger meter on appropriate
parts of the intermediate light switch. The same marking system used in the first
practical test was used here. This scheme gave a total mark out of four for the
second practical test.
Procedure. The experiment was conducted as part of the normal training
program of the company. The company used a "multiskilling" training system
in which apprentices were divided into groups and were rotated through differ-
ent training areas (e.g., industrial wiring, domestic wiring, sheet-metal work,
 COGNITIVE LOAD THEORY 301

and drawing). Installation testing formed a major part of the domestic wiring
course, which ran for 1 week. The installation testing notes were distributed by
the course instructor to apprentices on the second day of the domestic wiring
course. The course instructor randomly allocated subjects to either a modified
or a conventional instructions group with 14 subjects in each group. The ex-
perimenter and the senior instructor who were to carry out tests did not have
access to information on the group to which each subject was allocated. The notes
were used by the apprentices throughout the remainder of the course and formed
a regular part of their training. Other aspects of the training program in installa-
tion testing were practical demonstrations of equipment and written board work,
which were common to both groups. Apprentices had access to the notes at all
times, including when they were engaged in practical work.
There were three test periods, including two follow-up tests. The first test was
conducted at the end of the 1-week course. Apprentices were required to attempt
Downloaded by [Harvard College] at 03:46 27 July 2013

the written test on installation testing, described in the materials section. They
were tested in their respective groups and allowed a maximum of 15 min to com-
plete the test.
The first follow-up test was practical in nature and took place 1 week after
the completion of the domestic wiring course. Apprentices were required to per-
form both practical tests as described in the materials section. All apprentices
were tested individually. Apprentices did not have access to their notes during
these tests. The first practical test involved four individual tests of an electrical
appliance. Apprentices were allowed a maximum of 1 min to perform each test.
After each test, the experimenter inspected the work and scored it. The same
procedure was used for the second practical test, which involved two tests of
an intermediate light switch. After each of these light switch tests, a senior in-
structor inspected and scored them.
A second follow-up test was conducted approximately 12 weeks after the com-
pletion of the domestic wiring course. Apprentices were required to attempt for
a second time the written test on installation testing. The same procedure used
for the first written test was employed. After completing the written test, the

TABLE 1
The Design of Experiments 1 and 2

Test
Period 1 Test Period 2 Test Period 3
Acquisition (Immediate) (I-Week Follow-Up) (12-Week Follow-Up)

Experiment 1 Conventional or Written Written and Written and practical

modified practical
instructions
Experiment 2 Conventional or Written Practical Written
modified
instructions
 apprentices were required to reattempt the first practical test using the electrical
appliance. The test procedure was the same as for the first follow-up. For practi-
cal reasons, testing of a light switch could not be carried out during the second
follow-up. Table 1 provides a diagram of the experimental design.

Results and Discussion

Table 2 indicates the mean scores for the two groups on all written and practical
tests. A 2 (Groups) x 2 (Test Periods) analysis of variance (ANOVA) with repeat-
ed measures on the last factor was calculated to assess any written test differ-
ences between the two groups. (Written tests were conducted during the first and
third test periods; see Table 1.) Results showed a significant main effect for groups
in the predicted direction, F(1, 26) = 8.60, MS, = 553.14, suggesting an over-
all superiority of modified instructions. (The .05 level of significance is used
Downloaded by [Harvard College] at 03:46 27 July 2013

throughout this article unless otherwise stated.) The interaction between groups
and test periods was not significant, F(l, 26) = .63, MS, = 7.14, indicating
a similar pattern of results, favoring the modified instructions group, over both
test periods. Overall, there was an improvement in written test scores from the
first test period to the 12-week follow-up, F(l, 26) = 8.68, MS, = 97.79. This
result is to be expected from increased exposure to installation testing equipment
over the 12-week period. Although the performance of both groups increased
over the test period, the conventional group mean scores remained well below
that of the modified group. In fact, the improved mean score for the convention-
al group was below the mean score for the modified group in the first test period.
This suggests that, even after 12 weeks of access to instructional notes and equip-
ment, the conventional group's written test performance still had not exceeded
that of the modified group's initial performance.
We suggest that the written test results are in accord with the hypothesis that
modified instructions (packaged in an integrated format) imposed a lower cogni-
tive load than conventional instructions. Because modified instructions reduced
the need to reformulate the material in order to be understood, cognitive resources
could be devoted to learning and revising the installation testing material. To un-
derstand the conventional instructions, apprentices presumably needed to make
continual mental integrations throughout the notes to assimilate the material. These
mental integrations also were probably required when the material was revised.
Attention and mental resources may have been continually devoted to a task un-
related to learning. Consequently, relatively fewer resources were available for
acquiring the principles of installation testing.
Practical tests were conducted during the second and third test periods (see
Table 1). Of the 28 apprentices originally tested, only 20 (10 from each group)
were available for the second test period. Data from the first practical test (appli-
ance testing using an electric kettle) were subjected to a 2 (Groups) x 2 (Test
Periods) ANOVA with repeated measures on the last factor. As with the written
 TABLE 2
Written Test and Practical Test Scores
for the Three Testing Periods of Experiment 1

Test Period

Practical

Group Written Test Test 1 Test 2 Written Test Practical Test I

Conventional
M 8.2 1.9 1.2 10.1 2.1
SD 7.3 1.2 0.9 5.6 1.8
Modified
M 13.8 4.5 1.9 17.1 5.4
SD 6.2 1.8 1.5 5.3 1.2
Downloaded by [Harvard College] at 03:46 27 July 2013

test results, there was a significant main effect due to groups, F(1, 18) = 33.72,
MS, = 87.03. The Class x Test Period interaction was not significant, F(1, 18)
= .58, MS, = 1.22, indicating a similar pattern of results over the second and
third test periods. There was no overall difference in practical test performance
between the second and third test periods, F(l, 18) = 1.42, MS, = 3.02.
The second practical test was conducted once only (in the second test period;
see Table 2). There was no significant difference between the two groups on the
second practical test, t(18) = 1.24, although the direction was as predicted.
In general, the practical test results, like the written test results, favored the
modified instructions group. These findings indicate that the hypothesized ad-
vantages of integrated instructions transfer to related practical skills as well as
performance on written tests.
The findings of this experiment have considerable significance. They suggest
that integrated instructional formats are superior to conventional split-source for-
mats. We consider them important for a number of reasons. First, they were ob-
tained in an industrial training setting that is rarely used in experiments based
on cognitive theory. Second, the results were attained using detailed and lengthy
instructional notes. Last, and most important, the differences between the two
groups in both written and practical skills persisted throughout the 3-month study.
Cognitive load theory generated the experiment and the result. We assumed that
integrated instructions would reduce cognitive load and allow attention to be direct-
ed to acquiring knowledge of installation testing principles. The results indicated
that this knowledge continued to affect performance over a relatively long peri-
old and, based on the final tests, may have assisted in the acquisition of further
s l d . Conversely, we hypothesized that the conventional split-source format, which
required numerous mental integrations, misdirected attention and imposed a rela-
tively heavy cognitive load, leaving fewer cognitive resources available for ac-
 quisition of the installation testing principles. Less learning by the conventional
instruction group is in accord with this hypothesis.
It should be pointed out that, because of the realistic environment used in this
study, there were no direct measures of cognitive load. Rather, the experiment
was generated by and the results are in accord with cognitive load theory. Nor-
mally, differences in time required to study learning materials may be used to
indicate differences in the cognitive load imposed by the materials. Such meas-
ures were not possible in this study. It might be noted that differences in study
times favoring integrated materials were obtained by Sweller et al. (1990) using
laboratory studies. In addition, it was possible to record study times in some of
the subsequent experiments reported in the current article.

EXPERIMENT 2
Downloaded by [Harvard College] at 03:46 27 July 2013

Experiment 1 focused on instructional materials that required learners to inte-

grate multiple sources of information mentally, because the instructions were not
intelligible in an unintegrated form. Units of information that are unintelligible
unless they are integrated, either mentally or physically, with other units of infor-
mation, are quite common in mathematics, science, engineering, and other tech-
nical areas. Nevertheless, we need to ask whether integration of disparate sources
of information is always beneficial. In some areas, mental integration is not
essential, because one source of information may be intelligible by itself. For
example, a diagram may be fully understandable without reference to related
textual information; similarly, textual information may be intelligible without refer-
ence to a related diagram. This possibility raises a number of interesting ques-
tions. The findings of Experiment 1 confirmed the superiority of integrated in-
structional formats in areas where mutually referring sources of information needed
to be assimilated in order to be understood. It was assumed that conventional
instructional formats, which require mental integrations to be intelligible, mis-
direct attention and impose a heavy cognitive load. However, when one source
of information can be understood separately, mental integrations are not neces-
sary and may not occur. In these situations, people may learn quickly to attend
only to one source of information and avoid unnecessary mental integrations. Un-
der these circumstances, there is no reason to assume that integrated instruction-
al formats will be of any benefit.
Experiment 2 was designed to compare conventional instructional formats with
modified instructional formats using principles of wiring electrical light circuits.
The conventional split-source instructions were very similar to those regularly
used by the company as part of its electrical training program. The modified in-
structions contained virtually the same information but in an integrated format.
The experiment was conducted at the same time and used the same subjects as
Experiment 1. Unlike the previous experiment, mutually referring sources of in-
 COGNITIVE LOAD THEORY 305

formation did not necessarily need to be integrated in order to be understood.

For example, diagrams could be understood without reference to their accom-
panying text. Understanding could occur without expending mental resources in
the process of integration. As a consequence, physical integration may not reduce
cognitive load.
Figure 3 provides an example of a diagram that is intelligible without refer-
ence to the accompanying text. Consider the first textual statement: "The active
wire goes from the active to the common of Switch 1."This statement adds noth-
ing to information that can be obtained by studying the diagram. In fact, all of
the statements except the sixth are basically redundant. As a consequence, the
integration of Figure 4 may not assist intelligibility.
The text-diagram redundancy of Figures 3 and 4 can be contrasted with the
text associated with the diagram of Figure 1 used in Experiment 1. The first sen-
tence that refers to the diagram reads: "Make sure main switch is 'on' and all
Downloaded by [Harvard College] at 03:46 27 July 2013

fuses are 'in.' " This information is not likely to be gleaned from the diagram.
Similarly, none of the remaining statements provides redundant information that

Internal wiring for intermediate switching

I. The active wire goes fmm the active to the common of switch 1
2. In this type of switching we use an additional switch called the
intermediate switch
3. The wires connecting switch 1 to the intermediate switch and to
switch 2 are called strap wires
4. The switch wire goes from the common of switch 2 to the light
5. The neutral wire goes from the light to the neutral
6. Under no circumstances is the gauge of win used in this type
of circuit to be bmken.

intermediate
switch

FIGURE 3 Example of the instructional format presented to the conventional group of Ex-
periment 2.
 Internal wiring for intermediate switching

In this type of switching

@ Ihe active wire Owe
use an additional
gors from the active switch call* the
lo tile common of intermediate switch
light
, A A

active I '\
\

from the
switch 2 to the light la tl~c
light - neutrnl
-
Downloaded by [Harvard College] at 03:46 27 July 2013

earth-
(green)

@ Under no circumtmces is the gauge of wire used in this

typ of circuit to be broken.

FIGURE 4 Example of the instructional format presented to the modified group of Experi-
ment 2 .

could be obtained readily from the diagram. The advantages of using integrated
material (Figure 2) become manifest.
The aim of Experiment 2 was to investigate possible differences between con-
ventional and integrated instructions in an area where there was no absolute neces-
sity for sources of information to be mentally integrated in order to be intelligi-
ble. Figures 3 and 4 exemplify some of the materials used.

Method

Subjects. The subjects were 28 first-year trade apprentices from a Sydney

company. No subjects had any previous training in wiring lighting circuits.

Materials. There were two sets of instructional notes on lighting: conven-

tional and modified. Both sets of notes were designed to introduce apprentices
to lighting in domestic areas. The conventional instructional notes, in the form
of a 10-page booklet, were virtually identical to the regular instructions used by
the company as part of its electrical apprentice training program. As with Ex-
periment 1, the very minor revisions made to the instructions were only in the
interests of clarity. The main emphasis on the notes was on introducing appren-
tices to various electrical light circuits (e.g., single light, two-way switching, in-
termediate switching, and fluorescent light). The conventional notes for each
 COGNITIVE LOAD THEORY 307

lighting circuit were in a conventional split-source format, with an internal wir-

ing diagram and a written explanation of the circuit presented separately. An ex-
ample of the conventional format for an electrical light circuit is presented in
Figure 3. As indicated before, the diagrams of the electrical circuits could be
understood without reference to the accompanying textual information. In addi-
tion to the notes on electrical circuits, there were notes on the operating princi-
ples of fluorescent lamps. These notes also were presented in a split-source for-
mat, with a diagram of a fluorescent lamp and related textual information presented
separately. The textual information could be understood without reference to the
diagram.
The modified instructional notes contained the same information but in an in-
tegrated format. Mutually referring sources of information were integrated, where
possible, into unitary sources of information. An example of an electrical circuit
in a modified format is displayed in Figure 4.
The test materials consisted of a test booklet, as well as the apparatus for prac-
Downloaded by [Harvard College] at 03:46 27 July 2013

tical tests. The test booklet was divided into five sections that covered questions
on each electrical circuit and operating principles of fluorescent lamps. The ques-
tions on electrical circuits simply required subjects to identify components and
lwires from internal wiring diagrams. This was the method by which the compa-
my usually assessed apprentices7knowledge of electrical circuits. As one exam-
ple, subjects were presented with the internal wiring of a fluorescent light and
asked to label specific components of the diagram, such as the switch and neutral
and active wires.
In total there were 31 questions. As with Experiment 1, questions were marked
as either correct or incorrect. One mark was allocated for a correct response,
iwhereas no mark was given for an incorrect response.
There was one practical test, which required apprentices to completely con-
struct an intermediate light switch. The materials required for this test were an
electrical training board (a board used by apprentices to construct different light
switches), various wires, and electrical tools. This was the practical method by
which the company usually assessed apprentices' knowledge of an electrical cir-
cuit. Assessment of this test was simple and completely objective, because the
intermediate light switch could only be classified as operative or inoperative. A
senior electrical instructor tested each apprentice's light switch separately and
judged whether the circuit was operative.

Procedure. The procedure was similar to that of Experiment 1. The experi-

ment was conducted as part of the normal training program of the company. The
lighting notes were distributed by the course instructor to apprentices on the sec-
ond day of the domestic wiring course. Subjects were randomly allocated to either
m modified or a conventional instructions group. The notes were used by the ap-
prentices throughout the remainder of the week-long course. They had access
tlo the notes at all times, including when they were involved in practical work.
 As with Experiment 1, the experimenter and senior instructor were unaware of
the actual allocation of apprentices to groups.
There were three test periods, including two follow-up tests. The first test was
conducted at the end of the course. Apprentices were required to attempt the written
test on lighting. They were tested in their respective training groups and allowed
a maximum of 15 rnin to complete the test.
The first follow-up test (the second test) was practical in nature and took place
1 week after the completion of the domestic wiring course. Apprentices were
required to perform the practical test of constructing an intermediate light switch.
They were tested in their respective groups. Each apprentice was given an elec-
trical training board and had access to appropriate wires and electrical tools. The
apprentices were given up to 40 min to complete the task and did not have access
to their lighting notes. After an apprentice was satisfied the task was complete,
a senior instructor tested the circuit and classified it as either operative or inoper-
Downloaded by [Harvard College] at 03:46 27 July 2013

ative. An experimenter recorded these results.

A second follow-up test (the third test) was conducted approximately 12 weeks
after the completion of the domestic wiring course. Apprentices were required
to reattempt the written test on lighting. The same procedure used for the first
written test was employed. For administrative reasons, it was not possible to run
a follow-up practical test. The experimental design is displayed in Table 1.

Results and Discussion

Mean scores for the two groups on all written and practical tests are displayed
in Table 3. Written test data from the first and third tests were subjected to a 2
(Groups) x 2 (Test Periods) ANOVA, with repeated measures on the last fac-
tor. Inspection of means indicated that the conventional group performed margi-
nally better than the modified group on both test occasions. Nevertheless, there
was no significant difference between the groups, F(l, 26) = 1.76, MS, =
64.29. In addition, there was no performance difference between the test peri-
ods, F(l, 26) = .02, MS, = .29, and a nonsignificant Group x Test Period
interaction, F(1, 26) = 2.43, MS, = 34.57.
Of the original 28 apprentices, only 20 (10 from each group) were available
for the practical test. Seven of the 10 apprentices from the conventional group
successfully constructed an intermediate light switch. This compared with four
successful constructions from the modified group. A Fisher Exact Test on these
frequencies failed to find any significant difference between the two groups.
The findings of this experiment are clearly different from Experiment 1, which
found strong effects favoring modified, integrated instructions. The current ex-
periment did not provide any evidence for the superiority of modified instruc-
tions. We believe that these results are in accord with the theory used to generate
the experiments. As mentioned earlier, the materials used in this experiment were
very different from those of Experiment 1. The instructional materials of the
 COGNITIVE LOAD THEORY 309

TABLE 3
Written Test Scores and the Number of Subjects Successfully Completing
the Practical Task for the Three Testing Periods of Experiment 2

Test Period

I 2 3

Group Written Test Practical Test Written Test

Conventional
M
SD
Modified
M
SD
Downloaded by [Harvard College] at 03:46 27 July 2013

previous study required learners in the conventional group to integrate disparate

sources of information mentally, because the instructions were unintelligible in
an unintegrated form. Experiment 2 used instructions that did not necessarily re-
quire sources of information to be mentally integrated in order to be intelligible.
As a consequence, physical integration did not yield any benefits.
Experiments 3 and 4 were laboratory type experiments conducted to test the
possibility that in some instructional areas a single, meaningful source of infor-
mation is superior to multiple sources of information in either an integrated or
a conventional, split-attention format. By running the experiments under labora-
tory conditions rather than the realistic conditions of Experiments 1 and 2, it should
be possible to obtain more direct data relevant to the theoretical constructs used
to generate the experiments. For example, we have hypothesized that effects
that show up during test periods are due to differences in cognitive load during
acquisition. Although time spent in acquisition may be used as an indicator of
cognitive load, it could not be measured in the previous experiments because of
the environment in which they were run. It was measured in the subsequent
experiments.

EXPERIMENT 3

I11 Experiment 2, it was hypothesized that apprentices given the conventional in-
structions would quickly realize that the lighting material contained redundan-
cies and that they could avoid unnecessary mental integration by attending to only
one meaningful source of information, usually in the form of an internal wiring
diagram. As a consequence, nothing was gained by integrating text and diagrams
for the modified group. Experiment 3 was designed to test this hypothesis.
Two groups were used. Both groups were given conventional instructions for
a direct on-line (DOL) starter control circuit, a circuit used in industrial electrical
 areas. The instructions consisted of an internal wiring diagram of the circuit, as
well as related textual information describing the circuit. The internal wiring di-
agram could be fully understood without reference to the textual information.
The implicit instruction group was simply asked to study the instructions. The
explicit instruction group was not only asked to study the instructions but was
also instructed to make sure the textual information was read and related to the
diagram. Because the explicit instruction group was clearly instructed to assimi-
late the textual information, it might be expected that this group would direct
attention and cognitive resources to mentally integrating the two sources of in-
formation, rather than to understanding the circuit. If such use of cognitive
resources by the explicit instruction group occurs, it can be predicted that the
implicit instruction group would exhibit superior knowledge of the DOL starter
control circuit because many of them might ignore the textual information and
devote attention and mental resources to the diagram.
Downloaded by [Harvard College] at 03:46 27 July 2013

It might be noted that this experiment does not directly test the hypothesis that
the implicit instruction group attends to the diagram to the exclusion of the text.
Nevertheless, given that the diagram contains all necessary information and given
that we know that diagrammatic information normally is massively easier to
process than the equivalent textual information (see Larkin & Simon, 1987), it
is reasonable to predict that the implicit group will exhibit superior performance
due to these subjects attending primarily to the diagram.

Method

Subjects. The subjects were 20 first-year apprentices from a Sydney com-

pany. The subjects had only limited experience in industrial electrical areas. All
subjects had previous exposure to circuits with internal wiring diagrams and related
textual information. Sixteen of the 20 subjects had no previous exposure to a DOL
starter control circuit. The remaining 4 subjects had very briefly viewed a DOL
starter control circuit within the context of a larger DOL circuit. To ensure a
balanced experiment, 2 of these subjects were randomly allocated to the first group
and 2 subjects were randomly allocated to the second group.

Materials and procedure. Subjects were allocated randomly to either a

group directed explicitly to study both the text and diagram and to relate them-
called the explicit instruction group-or to a group merely presented both text
and diagram without explicit instructions to study both-called the implicit in-
struction group. All subjects were tested individually. The experiment was con-
ducted in two phases. The first phase was the instruction phase. The experimenter
began this phase by informing subjects that they would be asked to study some
instruction material for a DOL starter control circuit that would be followed by
a number of test problems. Subjects in the implicit group were asked simply to
 COGNITIVE LOAD THEORY 31 1

study the instructional material. Subjects from the explicit group were given similar
instructions but also instructed to read the textual information and relate it to the
diagram. The instructional materials for both groups are given in Figure 5. The
subjects were asked to read the instructional material at their own pace and to
indicate when they had finished reading. Time for completion was noted.
A test phase followed. Three problems were presented, one at a time. Instruc-
tional materials were not available to apprentices during testing. Subjects were per-
mitted unlimited time on each problem. The first test problem required them to con-
struct and label a DOL circuit. The second problem consisted of a slightly different
diagram from the DOL presented to subjects in the instructions. The diagram dif-
fered from the original in the following ways: (a) The active was located below the
rest of the circuit; (b) the fuse was located directly below the thermal overload
contact (TOC); and (c) the wire between the fuse and the TOC was shorter. Sub-
jects were asked to label the components of this diagram. The third problem con-
Downloaded by [Harvard College] at 03:46 27 July 2013

sisted of three questions requiring written responses. Subjects were asked to in-
dicate where specific components of the circuit were located. The first question
asked for the starting and finishing positions of the neutral wire (see Figure 5).
Wirinn of a Direct On Line (D.0.L) Staiter Control Circuit

I) The active wire goes fmm the active to the fwe.

7) From the fuse to the Thermal Overload Contact (T.0.C).
3) From out of the T.0.C to the stop button.
4) The circuit is then Gred from the stop button to the start button from stop button to
one side of holding contact (24).
5) Then from the stan button to the contactor coil & from the start button to the other side of
holding contact (a).
6) The neutml wire goes from the contactor coil to the neutml bar.

Active
wire

Thermal Overload
Contact (T.0.C)
Neutral

~on\tactor
Holding Contact (24) Coil

I
FIGURE 5 Instructional material presented to both groups of Experiment 3.
 3 12 CHANDLER AND SWELLER

The second question asked for the starting and finishing positions of the active
wire. These two questions required two written responses each. The third ques-
tion asked for the component that was located between the fuse and the stop but-
ton. Only one response was required for the third question. Thus, the third problem
consisted of three questions requiring five written responses.

Results and Discussion

Table 4 indicates mean instruction times and test scores for both groups. A t test
indicated that the implicit group, which was asked simply to study the material,
spent significantly less time processing the instructions than the explicit group,
which, in addition to being asked to study the material, was instructed to assimi-
late the text with the diagram, t(18) = 3.28. The difference between the groups
was quite large, with the explicit group spending almost twice as much time
Downloaded by [Harvard College] at 03:46 27 July 2013

processing instructions as the implicit group.

This result is consistent with the cognitive load hypothesis. Assume subjects
from the implicit group quickly realized the nature of the instructions and avoid-
ed unnecessary mental integrations. Instead, they directed their attention and men-
tal resources to the self-explanatory diagram. As a consequence, cognitive load
was reduced. A reduction in cognitive load by the implicit group is consistent
with the relatively rapid rate at which the material was assimilated.
Test scores for each problem were calculated as follows. The first problem
required apprentices to construct and label a DOL circuit similar to the one present-
ed during the instruction period. One mark was allocated for each component
in its correct position. One mark was also allocated for each correct label. In
total, 30 marks were allocated for the first problem. The second problem required
apprenticesto label (but not construct) a diagram that differed from the one present-
ed during instruction. One mark was allocated for each correct label, giving a
total out of 12. The third problem required five written responses. One mark

TABLE 4
Instruction Times (in Seconds) and Test Scores
on the Problems of Experiment 3

Problem

Group Znsrruction Time 1 2 3

Implicit
M
SD
Explicit
M
SD
 was allocated for each correct response, giving a total out of five. The implicit
group scored significantly higher (using a one-tailed test) than the second group
on the first test problem, t(18) = 1.81. Although the direction was as predicted,
there was no significant difference between the groups on the second problem,
t(l8) = 1.05, or the third problem, t(l8) = .90. The lack of significant effects
on the second and third problems may be due partly to the reduced marking scale
available for these problems. The first problem could be marked out of 30, but
Ihe second and third problems could only be marked out of 12 and 5 marks,
respectively.
The findings from this experiment favor the implicit instruction group. Despite
spending substantially less time on the instructional material, this group performed
significantly better than the explicit instruction group on the first test problem.
The direction of results for the other two problems also favored the first group.
The findings of the study are consistent with the view that apprentices from the
Downloaded by [Harvard College] at 03:46 27 July 2013

implicit instruction group, who were simply asked to study the instructions, rapid-
ly identified the nature of the instructional material, abandoned attempts at un-
necessary mental integrations, and instead directed attention and mental resources
solely to the diagram. On the other hand, it is possible that the explicit instruc-
tion group, which was clearly instructed to assimilate the textual material and
the diagram, unnecessarily directed attention and cognitive resources to this task.
If the explicit instruction group behaved in this manner, fewer resources would
have been available for acquiring knowledge of the circuit.

EXPERIMENT 4

In Experiment 3, we showed that instructing apprentices to mentally integrate

redundant textual information with a self-explanatory diagram impeded knowledge
of an electrical circuit relative to neutral instructions to learn the materials. If
redundant textual information is a handicap to learning, its removal may be an
important step in improving instructional formats. Experiment 4 was designed
tlo test this hypothesis using a mixed electrical circuit. These circuits are used
extensively in domestic electrical areas.
There were three groups in the experiment. The first group received a dia-
gram of the internal wiring of a mixed circuit only. A second group received
the instructions in a conventional format, with textual information presented
separately to the internal wiring diagram. This group was specifically instructed
to assimilate the textual material with the diagram. A third group was given the
instructions in a modified, integrated format. We predicted that the diagram-only
group would exhibit superior performance to the conventional and modified
groups, because attention and mental resources could be devoted entirely to study-
ing the one essential source of information-the diagram.
 31 4 CHANDLER AND SWELLER

Method

Subjects. The subjects were 20 first-year apprentices from a major Sydney

company. The subjects had some experience in domestic electrical areas. All sub-
jects had previous exposure to internal wiring diagrams and related textual infor-
mation. No subjects had seen instructions for a mixed circuit.

Materials and procedure. Subjects were randomly allocated to a diagram-

only, conventional, or modified group. The experiment was conducted in two
phases, and all subjects were tested individually. During the instruction phase,
subjects were informed that they would be asked to study some instructional materi-
al for a mixed circuit that would be followed by a number of test problems. Sub-
jects from the conventional group were also instructed to read the textual infor-
mation and relate it to the diagram. The different versions of the materials are
Downloaded by [Harvard College] at 03:46 27 July 2013

given in Figures 6 and 7. The subjects were asked to study the materials at their
own pace and to indicate when they had finished. An experimenter noted the time
for completion of instructions.
During the test phase, three problems were presented individually, and no time
limit was imposed. The first test problem required subjects to construct and label
a mixed circuit. The second problem consisted of a slightly different diagram
from the mixed circuit presented to subjects in the instructions. This diagram
differed from the original in that the general purpose outlet and associated earth
wire were located above the rest of the circuit. Subjects were asked to label the
components of the diagram. The third problem consisted of three questions re-
quiring eight written responses. As with Experiment 3, subjects were asked to
indicate where specific components of the circuit were located. The first ques-
tion of the third problem asked for the starting and finishing position of the switch
wire (see Figure 6 or Figure 7). The second question asked for the starting and
finishing positions of the active wire. These two questions required two written
responses each. The third question asked for the starting positions and finishing
positions of the primary switch wires. Four responses were required for this ques-
tion, because there were two primary switch wires.

Results and Discussion

Mean instruction times and problem test scores for all groups are presented in
Table 5. Planned contrasts using an ANOVA were run. The diagram-only group
required significantly less time to process the instructions than did the other two
groups, t(27) = 3.08, SEd8 = 14.43. The modified instructions group spent sig-
nificantly less time processing instructions than the conventional instructions group,
t(27) = 4.34, SEW = 24.99.
These results are as expected. The diagram-only group, with only one source
of information to attend to, quickly processed the instructional material. The
 COGNITIVE LOAD THEORY 315

Wiring of a mixed circuit.

1) The active wire goes from the active to the common of SW I.

2) The primary switch wire actually consists of two wires . The fint goes from
terminal I of SW I to the common of SW 2. The second goes from terminal 1
of SW I to the active at the general purpose outlet (GPO).
3) Stmp wires are connected between terminals I to I and 2 to 2 of SW 2
and SW 3.
4) The switch wire goes fmm the common of SW 3 to the active at the light.
5 ) The neutral wire has two parts. The first goes from the neutral at the light to
the neutral bar. The second goes from the neutral at the GPO to the neutral
bar.
6) The earth wire has two pam. The first goes from the light to earth. The
second goes from the earth at the GPO to earth.

active strap wires

Downloaded by [Harvard College] at 03:46 27 July 2013

wire

------
general purpose outletCPO)
1

I
neutral

wire

FIGURE 6 Instructional material presented to the diagram-only and the conventional groups
of Experiment 4. The diagram-only group was not presented with the explanatory text.

relatively short instruction time for this group is consistent with the suggestion
that cognitive load had been reduced because the group had to attend to a dia-
gram only and not written instructions. Although subjects in the modified group
had an integrated instructional format and, therefore, did not have to perform
any mental integrations, they still had to process redundant written material. Thus,
their instruction time was longer than that of the diagram-only group. The con-
ventional group, which spent notably more time on the instructions than the other
groups, was not only required to process redundant written information but in
addition had to perform a series of mental integrations. We expected this process
of mental integration to increase cognitive load and, therefore, to lengthen in-
 Wiring of a mixed circuit.
@ 2:tral
@2tCh wire has
wiregoes twoparis.
from the The fust

the neutral
active
to the at the light
I active at

The earth wire has two

from the wires. The first goes fmm
active to terminal 1 of S W 1 to the the light to earth.
common of SW 2
common
Downloaded by [Harvard College] at 03:46 27 July 2013

neutral
purpose outlet (GPO) I I

-
FIGURE 7 Instructional material presented to the modified group of Experiment 4 .

TABLE 5
Instruction Times (in Seconds) and Test Scores
on the Problems of Experiment 4

Problem

Group Instruction Time 1 2 3

Diagram only
M 60.4
SD 27.9
Modified
M 92.4
SD 18.1
Conventional
M 136.9
SD 44.9
 COGNITIVE LOAD THEORY 31 7

structional time considerably over a group that had the material physically in-
tegrated (the modified group) and even more so over a group that had the redun-
dant textual information totally eliminated (the diagram-only group). This result
was obtained.
Test scores for each problem were allocated using the same method as Ex-
periment 3. Forty-six marks were allocated for the first problem, 22 for the sec-
ond problem, and 8 for the third problem. The diagram-only group scored sig-
inificantly higher than the other two groups on the first test problem, t(27) =
3.65, SEd8 = 5.42, and on the second test problem, t(27) = 3.86, SEd8 =
2.90. There was no significant difference between the conventional and modi-
ffied groups on either the first test problem, t(27) = 1.09, SEd8 = 3.13, or on
the second test problem, t(27) = 0.72, SEd8 = 1.68. On the third test problem,
although the direction of results favored the diagram-only group, there was no
significant difference between the diagram-only group and the other two groups,
Downloaded by [Harvard College] at 03:46 27 July 2013

t(27) = 1.35, SEd8 = 1.63. There was also no significant difference between
the conventional and modified groups on the third problem, t(27) = 0.64, SEd8
= 0.94. The lack of significant differences on the third problem may have been
due to asymptotic effects. As can be seen from the means, most subjects obtained
very low scores on this problem.
The results of this experiment favor the diagram-only group. Although this
group spent significantly less time processing the instructions, consistent with
a reduced cognitive load, it performed far better than both the conventional and
modified groups on two of the test problems. The findings have direct implica-
tions for instructional formats. In areas where sources of information need not
be integrated to be understood, a redundant source of information may need to
be removed. The experiment showed that, by simply deleting an unnecessary
source of information, instruction time was reduced, and learning was enhanced.

EXPERIMENT 5

Experiment 4 showed that redundant textual information in either a conventional

or an integrated format impeded learning. This result was obtained using electri-
cal wiring materials. Experiment 5 was designed to test the generality of this find-
ing by using a very different learning area. Biology materials in the form of in-
structions explaining the flow of blood around the heart, lungs, and body were
used.
As with Experiment 4, there were three groups in this experiment. The first
group simply received a single self-explanatory diagram of the heart, lungs, and
body. The second group was given instructions in a conventional format, with
textual information presented separately for a diagram of the heart, lungs, and
body. This group was directly instructed to assimilate the textual information with
the related diagram. The third group received the instructions in a modified,
 31 8 CHANDLER AND SWELLER

integrated format. Based on the findings of Experiment 4, we predicted that the

diagram-only instructions would be superior to both the conventional and the in-
tegrated instructional formats.

Method

Subjects. The subjects were 30 Year 9 students from the top two of three
science classes of a Sydney high school. These students had no previous exposure
to instructions dealing with general circulation of blood around the heart, lungs,
and body.

Materials and procedure. The procedure was similar to Experiment 4. Sub-

jects were allocated to a diagram-only, conventional, or modified group. The ex-
Downloaded by [Harvard College] at 03:46 27 July 2013

periment was conducted in two phases. During the instruction phase, the sub-
jects received the instructions and were given unlimited time to process the in-
formation. Subjects in the conventional group were also instructed to read the
written information and to relate it directly to the diagram.
Figures 8 and 9 display the different versions of the instructional material.
The diagram-only group received a diagrammatic representation of the heart,
lungs, and body. Various bodily components were labeled with arrows indicating
the flow of blood. The diagram was a self-explanatory source of information.
The conventional group received textual information, given below the diagram,
explaining the flow of blood around the heart, lungs, and body. The modified
instructions contained textual information identical to the conventional instruc-
tions but in an integrated format. Textual information explaining the flow of blood
was placed at appropriate parts of the diagram.
A test phase followed. Subjects were presented with six test problems, con-
sisting of both diagrammatic and textually orientated questions. All problems were
presented individually, and once a problem was completed, it was covered from
view. No time limit was placed on any of the test problems. The first problem
required subjects to recall six parts of the heart. For the second problem, sub-
jects were given an unlabeled diagram of the heart, lungs, and body and asked
to label it with six parts of the heart. The third problem contained six parts. Sub-
jects were again given an unlabeled diagram. They were asked to place six num-
bers on the diagram to indicate the flow of blood.For example, the first part of
the third problem asks subjects to "put a 1 at the place where blood in the right
atrium flows to." To complete this question, a subject simply had to write the
number I on the unlabeled diagram at the place to which they believed blood
in the right atrium flows. The remaining five parts of the question were as follows:

Put a 2 at the place where the aorta receives its blood.

Put a 3 at the place where the blood in the lungs flows to.
 .Ixai hrolamldxa atp V!M pa~uasa~d
IOU saM dn0.G K ~ I I o - u I EayL
~ ~ ?'S? !luaur!ladxa
~ jo
sdnold @uogua~uo3 a v pm l(~uo-m8r?lpa q oi paiuasaid @palauI@ u o p ~ m ~ s 8 u ~=&J~CJIJ
Downloaded by [Harvard College] at 03:46 27 July 2013
 D i indicating flow of blood through the heart, lungs and body
Downloaded by [Harvard College] at 03:46 27 July 2013

FIGURE 9 Instructional material presented to the modified group of Experiment 5.

 COGNITIVE LOAD THEORY 321

Put a 4 at the place where blood in the right ventricle flows to.
Put a 5 at the place where blood in the body flows to.
Put a 6 at the place where blood in the left ventricle flows to.

The fourth problem required subjects to complete two blood flow chains. A
chain simply places the components in order of blood flow (e.g., left atrium -+

left ventricle -+ aorta -+ body). For the first chain, subjects were told that the
second component of the chain was the left atrium. The subjects then were asked
to complete the chain by listing the components for the first, third, fourth, and
fifth positions of the chain. For the second chain, subjects were informed that
the second component of the chain was the right atrium. As with the first chain,
the subjects were asked to complete the chain by listing the components for the
first, third, fourth, and fifth positions.
For the fifth problem, subjects were given an unlabeled diagram of the heart,
Downloaded by [Harvard College] at 03:46 27 July 2013

lungs, and body. Unlike the diagrams for Problems 2 and 3, this diagram had
lines located on it. These lines were located at the same positions as the arrows
irk the initial instructions diagram. To complete this test problem, subjects sim-
ply had to place an arrowhead on each line to indicate the appropriate flow of
blood. The sixth problem required five written responses to questions relating
to the flow of blood. They were as follows:

Blood in the left atrium flows where?

Blood in the right ventricle flows where?
Blood in the left ventricle flows where?
Blood in the pulmonary artery flows where?
Blood in the aorta flows where?

Results and Discussion

Table 6 displays means and standard deviations for instruction times and problem
test scores for all groups. Planned contrasts using an ANOVA were run. Results
indicate that the diagram-only group required significantly less time processing
instructional material than the other two groups, t(27) = 6.81, SEd@= 18.54.
The modified instructions groups spent significantly less time processing their
material than the conventional instructions group, t(27) = 4.96, SEd8 = 10.70.
These results replicate those of Experiment 4.
Test problem scores were calculated as follows. The first problem simply asked
subjects to name six parts of the heart. One mark was given for each correct
response, giving a total out of six. The second problem required subjects to label
a diagram with six parts of the heart. One mark was allocated for each heart com-
ponent in its correct position, giving a mark out of six. Problem 3 required
 TABLE 6
Instruction Times (in Seconds) and Test Scores
on the Problems of Experiment 5

Problem

Group Instruction Time I 2 3 4 5 6

Diagram only
M 69.1 5.3 4.9 3.7 3.5 14.9 1.8
SD 12.0 1.2 1.5 2.1 2.8 1.4 1.5
Modified
M 105.7 4.5 2.8 1.7 1.4 7.8 0.9
SD 9.6 1.2 2.4 1.6 2.1 4.5 1.1
Conventional
M 158.8 3.5 1.7 0.8 1.1 7.6 0.9
SD 38.5 1.2 1.3 0.8 1.5 4.1 0.9
Downloaded by [Harvard College] at 03:46 27 July 2013

subjects to place six numbers at various parts of an unlabeled diagram. One mark
was allocated for each number in its correct position, again giving a score out
of six. Problem 4 asked subjects to complete two blood flow chains. Each chain
required four responses. For both chains, one mark was allocated for each
component in its correct position on a chain, giving a total out of eight. The
fifth problem asked subjects to place arrowheads on 16 lines to indicate the
appropriate flow of blood. One mark was given for each arrowhead in its cor-
rect position, giving a score out of 16. Problem 6 required five written re-
sponses. One mark was allocated for each correct response, giving a mark out
of five.
Results indicate that the diagram-only group scored significantly higher than
the other two groups on the first five test problems. The critical t values were:
t(27) = 2.86, S E d ~= 0.91 for Problem 1; t(27) = 3.78, SEd@= 1.40 for
Problem 2; t(27) = 3.99, SEd@= 1.23 for Problem 3; t(27) = 2.63, SEd8 =
1.71 for Problem 4; and t(27) = 5.16, SEd8 = 2.79 for Problem 5. Although
the direction was as predicted for Problem 6 , the difference was not significant,
t(27) = 1.92, SEM = 0.94. This result may still have represented a real effect,
.05 -cp -c .l. All three groups recorded low marks on Problem 6. Despite this,
the mean test score for the diagram-only group was still double that of both the
modified and the conventional groups.
Although the direction of results favored the modified group over the conven-
tional group for the first five test problems, none of the differences was signifi-
cant. Both groups recorded equal means for the sixth test problem. The critical
t values were: t(27) = 1.91, SEd@= 0.52 for Problem 1; t(27) = 1.36, SEd#
= 0.81 for Problem 2; t(27) = 1.27, SEd@= 0.71 for Problem 3; t(27) =
0.30, SEdlf = 0.99 for Problem 4 ; t(27) = 0.12, SEd8 = 1.61 for Problem 5 ;
and t(27) = 0.00, SEd8 = 0.54 for Problem 6. Although there was no signifi-
cant difference between the modified and the conventional groups on the first
 COGNITIVE LOAD THEORY 323

test problem, the difference favoring the modified group may still have represented
a real effect, .05 < p < .l.
As with Experiment 4, the results of this study clearly favored the diagram-
only group. Despite spending far less time processing instructional materials, this
group exhibited superior performance over both the modified and the conven-
tional instruction groups. The experiment extended the findings of Experiment
4 in two ways. First, unlike Experiment 4, the diagram-only group was superior
to the other two groups on both diagrammatic and textually based questions. Sec-
ond, the benefits of eliminating unnecessary textual information have been shown
to be effective in a very different learning area, suggesting that this finding may
have considerable generality.
These results are in accordance with the suggestion that subjects in the con-
ventional group were required to direct attention and mental resources to assimilat-
ing redundant text with the diagram. As a result, attention may have been un-
Downloaded by [Harvard College] at 03:46 27 July 2013

necessarily misdirected away from the self-explanatory diagram, and mental

resources used in integrating the text and diagram may have been unavailable
for learning. These constructs were used to hypothesize poor test performance.
In fact, the conventional group means in Experiment 5 fell to below a quarter
of the diagram-only group.
We can hypothesize that, although subjects in the modified group were not
required to split attention between text and diagram, they still had to devote men-
tal resources to processing unnecessary textual information. There are some in-
dications that subjects in the modified group may have performed slightly better
than those in the conventional group, but the performance of the modified group
was still substantially below that of the subjects in the diagram-only group, who
did not have to process any redundant textual information. Consequently, these
differences in speed of learning are in accordance with the suggestion that sub-
jects in the diagram-only group could solely devote attention and mental resources
to processing the self-explanatory diagram.
Using very different instruction areas, data from Experiments 4 and 5 indi-
cate that redundant information can impede learning. The finding that the simple
removal of a redundant source of information can reduce instruction time and
enhance learning has direct implications for instructional formats. In instruction-
al areas where a self-explanatory source of information exists, disparate sources
of information need not be integrated in order to be understood. The single, self-
contained source of information should be identified, and other sources of redun-
dant information should be removed.

EXPERIMENT 6

Experiment 5 showed that subjects who studied a self-explanatory diagram of

the heart, lungs, and body demonstrated superior test performance to subjects
who received redundant textual information in addition to the diagram in either
 324 CHANDLER AND SWELLER

a conventional or an integrated format. From the point of view of cognitive load

theory, this result is consistent with the preceding experiments. Nevertheless,
ignoring our theoretical perspective, Experiment 1 demonstrated a very large ad-
vantage to a group that was presented integrated instructions, whereas the subse-
quent experiments, using different materials, failed to obtain such an effect.
Although the results are in accordance with cognitive load effects, they could
be due to a different, unknown variable associated with the various materials used.
We can reduce the likelihood of such unknown variables contaminating our results
by using materials similar to those of Experiment 5 but modified in such a way
that the diagram and text do not consist of self-contained, redundant materials.
If the diagram cannot be understood, except by reference to the text, results similar
to those of Experiment 1 should be obtained; integrated materials should be su-
perior to disparate sources of information. Using such materials, we should find
that integrated information requires substantially less study time than convention-
ally presented information. A reduction of study time is consistent with reduced
Downloaded by [Harvard College] at 03:46 27 July 2013

cognitive load. A sufficiently large reduction of extraneous cognitive load should

result in superior performance on subsequent test problems.
Experiment 6 used instructional materials similar to those of Experiment 5.
There were two groups in the experiment. One group received its instructions
on the blood flow through the heart, lungs, and body in a conventional split-source
format. Unlike Experiment 5, however, the diagram was not self-explanatory.
The labels for the components of the diagram were replaced with numbers. The
accompanying textual information included each number with its corresponding
component and an explanation of the component's role in the circulation of blood.
Thus, the two sources of information needed to be mentally integrated in order
to be understood. The modified instructions contained the same information in
an integrated format. We predicted, for the reasons detailed before, that the modi-
fied instructions with a unitary source of information would be superior to the
conventional instructions.

Method

Subjects. In this experiment, the school used ungraded Year 9 science class-
es. The subjects used in the experiment were the top 20 students, as judged by
a common science test given by the school. No subjects had previous exposure
to instructions dealing with general circulation of blood around the heart, lungs,
and body.

Materials and procedure. The procedure was the same as in Experiment

5 , except that subjects were allocated to either a conventional or a modified group.
In other respects, the instruction phase was conducted identically to that of Ex-
periment 5. The conventional and modified instructions are shown in Figures 10
 Diagram indicating flow of blood through the heart, lungs and body
Downloaded by [Harvard College] at 03:46 27 July 2013

7. right venmcle - Whrnlhir rmrurc =laxer blood from Lhr right alnwn flows m
-
R. lell venlncle When Ulls smriwe m l a n blood from Uu lcfl ahium flows In.
-
9 pulmonary anew Whcn the right venlncle conuacs Mad is forced into h s s a w
10. B l d enicnng Lhc pulmonary micry supplies the lungs.
I I . aom - When h e lell vcnlncle contraC1S Mood is forced inu) Ulis s a h l r e .
12. Blood entexing the mIfa is pumped back to the M y .

FIGURE 10 Instructional material presented to the conventional group of Experiment 6 .

 Diagram indicating flow of blood thmugh the heart, lungs and body
Downloaded by [Harvard College] at 03:46 27 July 2013

FIGURE 11 Instructional material presented to the modified group of Experiment 6.

 COGNITIVE LOAD THEORY 327

and 11, respectively. The conventional group received textual information listed
below the diagram, which contained numbers representing various bodily com-
ponents. The text consisted of each number and its corresponding bodily compo-
nent, along with a commentary of the role of the bodily component in the flow
of blood around the heart, lungs, and body. The modified instructions contained
virtually identical textual information to the conventional instructions but in an
integrated format. The only difference was that the numbers were not present
on the modified instructions. The test phase was conducted using the same proce-
dure and testing materials as in Experiment 5.

Flesults and Discussion

The variables under analysis were instruction time and test problem performance
on each of the six test items. Means and standard deviations for instruction time
and test scores are displayed in Table 7.
Downloaded by [Harvard College] at 03:46 27 July 2013

A t test indicated that the modified group spent significantly less time process-
ing instructions than the conventional group, t(18) = 5.30. This result is consis-
tent with a reduction in cognitive load for the modified group. We can hypothe-
size that subjects in the modified group simply processed their instructions by
attending to the single integrated source of information. Conversely, the conven-
tional group would have had to direct attention and mental resources to assimilating
the textual information with the diagram, because each source of information was
unintelligible by itself. Consequently, one would expect instruction-processing
time to be considerably longer for this group. This result was achieved.
Test problem scores were calculated using the methods of Experiment 5. Six
marks each were allocated to Problems 1, 2, and 3 . Eight marks were allocated
to Problem 4, 16 marks to Problem 5, and 5 marks to Problem 6 .
Analyses using t tests indicated that the modified group performed significantly
better than the conventional group on the first three test problems. The critical
t values were: t(18) = 3.10, for Problem 1; t(18) = 2.30, for Problem 2; and
t(18) = 2.40, for Problem 3. Although the direction was as predicted for Problem
TABLE 7
Instruction Times (in Seconds) and Test Scores
on the Problems of Experiment 6

Problem
-- -

Group Instruction Time I 2 3 4 5 6

Conventional
M 150.8 4.0 2.7 1.7 3.3 12.8 1.7
,SD 22.1 1.2 2.0 1.3 2.4 4.2 1.6
Modified
M 105.1 5.4 4.7 3.6 4.8 13.6 2.2
SD 16.0 0.7 1.9 2.1 2.5 2.6 1.8
 328 CHANDLER AND SWELLER

4, the difference was not significant, t(18) = 1.38. This difference may, neverthe-
less, represent a real effect on a one-tailed test, .05 < p < .l. The directions
were also as predicted for Problems 5 and 6, although the differences between
the means were quite small. Both of these findings may be due to asymptotic
effects. Subjects had little difficulty with Problem 5, with both groups obtaining
high scores. On the other hand, both groups obtained low scores on Problem 6.
The results of this experiment favored the modified instructions group. Despite
spending less time studying the instructions, this group performed better than
the conventional group on most of the test problems. These results contrast with
those of Experiments 2, 3, 4, and 5, where integrated instructions did not facili-
tate performance on subsequent test problems. They are in accordance with the
results of Experiment 1, despite vast differences in materials between Experi-
ments 1 and 6. Thus, whereas Experiments 5 and 6 used very similar materials
but obtained contrasting results, Experiments 1 and 6 used quite different materials
but obtained similar results. The paradox may be resolved by an analysis of the
Downloaded by [Harvard College] at 03:46 27 July 2013

materials in terms of the cognitive processes required to assimilate the informa-

tion presented in each experiment. We can hypothesize that, in Experiments l
and 6, students had to mentally integrate disparate sources of information when
presented with conventional materials, resulting in a heavy extraneous cognitive
load. In Experiments 2 , 3 , 4 , and 5, mental integration may not have been neces-
sary because of redundant information, and so physical integration had minimal
consequences. The elimination of redundancy proved more effective. The sur-
face structure of the materials (e.g., electrical wiring, human anatomy) proved
irrelevant to these processes.

GENERAL DISCUSSION

The experiments reported in this article have highlighted the inadequacy of some
conventional methods of presenting instructional materials. We believe that these
findings have some important implications. Before discussing these, however,
we summarize the results.
Experiment 1 used detailed electrical notes in an industrial training setting and
found that integrated instructions were superior to the conventional instructions
previously used by the company. Results favoring integrated instructions in written
test performance and practical skills persisted throughout the 3-month study, clear-
ly demonstrating the long-term advantages of eliminating the need for students
to split their attention between multiple sources of mutually referring information.
The advantages of the integrated instructional package used in Experiment 1
were found in an area where it was essential for two or more sources of informa-
tion to be mentally integrated in order to be intelligible. In contrast, we had no
theoretical reason for supposing that integrated instructions would be effective
in areas where sources of information did not have to be integrated in order to
 COGNITIVE LOAD THEORY 329

be understood. In another long-term study, Experiment 2 showed that integrated

instructions were no more effective than conventional instructions in areas where
it was not necessary to mentally integrate text and diagrams. It was hypothesized
that subjects identified the nature of the instructional material and only devoted
attention and mental resources to the one meaningful source of information.
Unlike Experiments 1 and 2, Experiments 3 and 4 were laboratory experi-
nnents designed to throw further light on the elimination of an effect favoring
integrated instructions in Experiment 2. Experiment 3, using industrial electrical
circuit instructions, found that subjects who were not specifically asked to inte-
grate disparate sources of information required less instruction time and performed
better than subjects who were specifically instructed to mentally integrate related
text and diagrams. We suggested that redundant information could impede learn-
ing and that its removal was another necessary step in improving instructional
materials. Experiment 4 provided strong evidence for this hypothesis. Using
Downloaded by [Harvard College] at 03:46 27 July 2013

domestic electrical circuit instructions, it was found that one self-explanatory

source of information was superior to two redundant sources of information in
either a conventional or an integrated format.
Experiments 5 and 6 duplicated the results of the preceding experiments in
a vastly different area: blood flow in the human body. Experiment 5 obtained
similar results to Experiments 2, 3, and 4, with physically integrated materials
having minimal effects. The elimination of redundant material proved superior.
Seemingly minor alterations to the Experiment 5 materials in Experiment 6 resulted
in a similar pattern of results to that obtained in Experiment 1. Integration proved
superior to disparate sources of information.

THEORETICAL CONSIDERATIONS

The experiments and predictions of this article were generated by cognitive load
theory. The theory was used to hypothesize that some conventionally used in-
structional designs are inadequate. It also was used to design alternative modes
of instruction predicted to be more effective. In general, the data supported our
hypotheses.
It must be emphasized strongly that we were not engaged in a theory valida-
tion exercise. The theory was used solely to attempt to provide results with direct
instructional implications. Our procedure was to use the theory to consider the
cognitive consequences of some instructional designs and to predict, on the basis
ofthose cognitive consequences, the adequacy of the formats used. We then test-
ed the predictions by comparing learning and problem solving after the use of
differing instructional designs. Direct tests of, for example, attentional or cogni-
tive load factors were not carried out. By emphasizing instructional effective-
ness, we have obtained immediate rather than merely potential applications. We
believe direct applications are essential for the health of the cognitive science
enterprise at this time.
 Inevitably, there are both negative and positive consequences to our approach
in this article. Because we have not tested cognitive processes directly in this
set of experiments, there is a possibility that our findings are due to variables
other than those we have postulated. Testing for cognitive processes decreases
the likelihood of alternative, post hoc explanations being available for a set of
results. Although we concede these points, it should be noted that many previous
reports have provided evidence for cognitive load theory using verbal protocols,
differential error scores and error locations, task difficulty as measured by time
to completion, and dual task paradigms (e.g., see Ayres & Sweller, 1990; Owen
& Sweller, 1985; Sweller, 1988; Tarmizi & Sweller, 1988). Notwithstanding this
work, additional detailed tests of cognitive processes need to be carried out.
Nevertheless, given that the theory was constructed solely to generate instruc-
tional applications, we feel it is essential that at some point it does just that. Fur-
thermore, although alternative conceptualizations may be found to explain our
Downloaded by [Harvard College] at 03:46 27 July 2013

results, currently they seem incapable of generating experiments leading directly

to applications. The fact that cognitive load theory is able to do so is itself a form
of incidental validation.
The major positive consequence of our approach is that, if hypotheses are sup-
ported, applications are available immediately. The ultimate aim of any theory
dealing with cognition and instruction must be that it generates new and useful
instructional techniques. Any other aims are merely means toward this end. Spe-
cifically, we do not feel that being able to explain results in a post hoc fashion
is a substitute for generating applications. Cognitive load theory can generate in-
structional applications.

INSTRUCTIONAL APPLICATIONS

The findings of the group of experiments reported in this article indicate that,
in areas where mental integrations are essential in order to make sense of two
or more sources of information, conventional instruction should be replaced by
integrated instructional formats. In areas where mental integrations are not neces-
sary because of redundant information, neither physical nor mental integration
is beneficial. Isolation and elimination of redundant sources of information are
preferable. As shown in Experiments 2, 3 , 4 , and 5, this is a simple process that
requires little more than a quick inspection of instructions. Once the self-
explanatory source of information is located, other unnecessary sources of infor-
mation should be deleted.
The frequently made assumption, sometimes explicit but more frequently
implicit, that redundant technical information is at least neutral and perhaps
beneficial in its effects on learning needs to be called into question. Redundancy
that can lead to mental integration seems to be neutral at best and then only
when learners are aware that it is redundant and, therefore, ignore it. Paradoxi-
 cally, by eliminating redundancy, intelligibility may be increased rather than
decreased.
Although most of us have a pervasive intuition that redundancy can be benefi-
cial, currently there is no information on the relevant conditions. It appears
reasonable to assume that redundant information that does not or cannot lead to
attempts at mental integration can be beneficial. Summaries, for example, are
redundant but are not integrated normally with the original material. They are
probably useful as a reminder of the preceding information. Widely separated
(physically or temporally) redundancy might be beneficial as a mnemonic device.
In addition, we must consider the possibility that redundant technical informa-
tion can be beneficial where information is so poorly structured that almost any
alternative presentation of the same material is useful, even if it does lead to at-
tempts at mental integration.
The results of the current series of experiments using introductory instruc-
tional materials can be combined with studies using worked examples (e.g., Cooper
Downloaded by [Harvard College] at 03:46 27 July 2013

& Sweller, 1987; Sweller & Cooper, 1985; Tarmizi & Sweller, 1988; Ward &
Sweller, 1990) to create guidelines for complete cognitively driven instructional
packages. As mentioned in the introduction, a heavier than normal use of worked
examples has been shown to be very effective (Cooper & Sweller, 1987; Sweller
& Cooper, 1985).
In technical areas, complete instructional packages consist normally of three
parts. The first part consists of introductory explanatory instructions, similar to
those used in the current series of experiments. The second part usually includes
one or two worked examples designed to demonstrate the new material. The third
part normally consists of a large number of problems or exercises. The current
results, along with the previous findings, have clear and direct implications for
each of these components of instruction:

1. Conventional introductory instructions that consist of multiple sources

of information that need to be mentally integrated in order to be under-
stood should be restructured into integrated formats. Sometimes inte-
gration may be impossible, but frequently it can be achieved relatively
easily. If sources of information cannot be restricted to a single entity,
they should be integrated into as small a number of units as possible.
If conventional instructions consist of a source of information that is fully
intelligible by itself, other related information, whether it be textual or
diagrammatic, should be removed to prevent unnecessary attempts at
integration.
2. These same rules apply to the structure of worked examples used to demon-
strate principles taught during initial instruction.
3. Far more emphasis should be placed on worked examples, rather than con-
ventional problems or exercises, by mixing large numbers of integrated
worked examples with conventional problems.
 332 CHANDLER AND SWELLER

On the evidence of the current and previous empirical work, the implementation
of these suggestions should result in substantial benefits.

ACKNOWLEDGMENTS

The work reported in this article was supported by Grant No. AC9031965 from
the Australian Research Council to John Sweller.
The cooperation and support provided by the staff of the New South Wales
Department of Education and by Email Ltd. was greatly appreciated. We extend
our sincere thanks to Email's training staff: Brian Jones, Richard Winter, John
Harley, and Jim Jarrick.

REFERENCES
Downloaded by [Harvard College] at 03:46 27 July 2013

Ayres, P., & Sweller, J. (1990). Locus of difficulty in multi-stage mathematics problems. American
Journal of Psychology, 103, 167- 193.
Chi, M., Bassok, M., Lewis, M., Reimann, P., & Glaser, R. (1989). Self-explanations: How stu-
dents study and use examples in problem solving. Cognitive Science, 13, 145-182.
Cooper, G., & Sweller, J. (1987). The effects of schema acquisition and rule automation on mathe-
matical problem-solving transfer. Journal of Educational Psychology, 79, 347-362.
Larkin, J., & Simon, H. (1987). Why a diagram is (sometimes) worth ten thousand words. Cognitive
Science, 11, 65 -99.
Owen, E., & Sweller, J. (1985). What do students learn while solving mathematics problems. Journal
of Educational Psychology, 77, 272-284.
Sweller, J. (1988). Cognitive load during problem solving: Effects on learning. Cognitive Science,
12, 257-285.
Sweller, J. (1989). Cognitive technology: Some procedures for facilitating learning and problem solving
in mathematics and science. Journal of Educational Psychology, 81, 457-466.
Sweller, J., Chandler, P., Tierney, P., & Cooper, M. (1990). Cognitive load and selective attention
as factors in the structuring of technical material. Journal of Experimental Psychology: General,
119, 176-192.
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.
Sweller, J., Mawer, R., & Howe, W. (1982). Consequences of history-cued and means-ends strate-
gies in problem solving. American Journal of Psychology, 95, 455-483.
Sweller, J . , Mawer, R., & Ward, M. (1983). Development of expertise in mathematical problem
solving. Journal of Experimental Psychology: General, 112, 639-661.
Tarmizi, R., & Sweller, J. (1988). Guidance during mathematical problem solving. Journal of
Educational Psychology, 80, 424-436.
Ward, M., & Sweller, J. (1990). Structuring effective worked examples. Cognition and Instruction,
7, 1-39.
Zhu, X., & Simon, H. (1987). Learning mathematics from examples and by doing. Cognition and
Instruction, 4 , 137-166.