Grimley
Grimley
Grimley
Differences
Dr Michael Grimley
University of Canterbury, School of Education, Private Bag 4800, Christchurch,
New Zealand. Email. [email protected]
Learning from Multimedia Materials: The Relative Impact of Individual
Differences
3
Abstract
This paper explores whether the principles of cognitive-load and multimedia theory
are mediated by cognitive style, gender and prior knowledge. Ninety-one children
aged 10 - 11 yrs (54 boys, 37 girls) were assigned to one of two presentation modes.
In Condition 1 children were presented with diagrams supported by printed textual
material, in Condition 2 the same diagrams were supported by narrated text.
Condition 1 was designed in the conventional manner but Condition 2 was designed
to adhere to cognitive load and multimedia theory. Following presentation children
were asked a number of comprehension questions as an outcome measure requiring
assimilation of information from both pictures and words. The Cognitive Style
Analysis was administered to measure wholist-analytic style and verbal-imagery style.
Results from national achievement tests were used as indicators of prior knowledge.
Results indicate that outcome is differentiated by style, gender and prior knowledge
not just instructional design.
What is the best way to present information? Should I use pictures and
diagrams? Do I get better learning if I listen to text or read it? How should I organise
information in order to ensure efficient learning? We ask many questions about how
best to organise and present study media format to aid efficient learning, even more so
with the popularity and flexibility of computer aided learning and hypermedia
delivery modes (Chen & Macredie, 2002). In order to answer these types of questions
we can turn to contemporary cognitive learning theories such as cognitive load
Theory (Sweller and Chandler, 1994) and the cognitive theory of multimedia learning
(Mayer, 1999, 2001). However, few theories consider individual differences
effectively. Clearly, the key to designing good instructional material is dependant
upon having an accurate and informative model of the learner and how the learner
learns best.
This paper considers how three important individual differences, namely,
gender, prior knowledge and cognitive style, mediate learning for tasks designed
using Mayer’s theory of multimedia learning and cognitive load theory. Currently,
both multimedia theory and cognitive load theory predict the learning outcome of
multimedia materials by considering how the materials are designed and presented.
However, little research has been conducted to ascertain how individual differences
affect this outcome in light of the actual design of the learning task. Clearly, it is
important for instructors to know whether the relevant design theory holds for all
individuals and if not how the design should be modified to suit different learners.
Therefore this study presents to two groups of learners the same multimedia learning
materials. These are designed to be either of conventional design or of design that
complies with multimedia theory and cognitive load theory. Learning outcomes are
then examined to establish whether they are predicted purely by the material’s design
or whether individual characteristics such as gender, cognitive style and prior
knowledge play a role.
extensive and thirdly that use of schemas enable information to be simplified and
stored efficiently within long term memory. On this third point, this storage
mechanism allows a schema the capability of being effectively utilised within the
limited capacity of working memory with these schemas eventually becoming
automatic and working via long term memory in parallel with the conscious processes
of working memory (Sweller & Chandler, 1994).
Working memory (Baddeley, 1974, 1986, 1999) can be described as a
multifaceted mechanism with a number of elements running in parallel. It is argued
that working memory constitutes a number of processors each with limited capacity.
The visuo-spatial scratch pad manipulates visuo-spatial information, the phonological
loop processes verbal material and the central executive controls and co-ordinates the
other two.
presented. This forces the learner to pay attention to extra material not necessary for
the particular element being learned. Consequently, when redundant information
interferes with relevant information in this way it also places a higher cognitive load
on the learner.
Information enters sensory memory via either the eyes or ears allowing words
to enter as sound (spoken) or visual stimuli (written). Pictures however can only enter
as visual stimuli. Information enters sensory memory and if selected is transferred to
working memory. Organization of sound or visual information takes place in working
memory to create a model in a form that makes sense to the learner. This model
(pictorial or verbal) is further integrated using prior knowledge. Clearly, words and
pictures can be translated into different forms once selected in working memory;
however this happens at a cost. Hence, pictures can be verbalised and textual
information can be represented in pictorial form (shown by the arrows going between
sounds and images in Fig 1). This model builds upon cognitive learning theory by
assuming active construction of knowledge that is modified by prior knowledge and
implies a cognitive system with limited capacity. Processing of information takes
place within the constraints of sensory and working memory and takes advantage of
dual processing theory as conceptualised by Baddeley (1999) and Clark and Paivio
(1991) and each channel is mutually exclusive with limited capacity.
Mayer (2001, 1999) outlines seven guiding principles for presenting information in a
multimedia format.
Multimedia Principle –there is better learning when learners receive words
and corresponding pictures rather than words alone. This is clearly predicted by the
7
model. The benefit of using both channels is that the capacity is increased and the
learner receives the benefit of two models/representations.
Spatial Contiguity Principle–there is better learning when words and
corresponding pictures are near rather than far from each other. The benefit of having
words and pictures near to one another is that both representations can be held in
working memory simultaneously and cognitive resources are reserved because the
learner does not have to search for relevant material.
Coherence Principle–there is better learning when extraneous words and
pictures are excluded rather than included. Extraneous words and pictures tend to
compete with more relevant information for limited resources.
Modality Principle–there is better learning when words are presented as
narration rather than in visual form. This is predicted by the model because both
channels are used, one for the text and one for the picture rather than just the visual
channel which would tend to be overloaded.
Redundancy Principle–there is better learning from animation and narration
than from animation, narration, and on-screen text. On-screen text would tend to
compete with animation for resources.
Temporal Contiguity Principle–there is better learning when corresponding
words and pictures are presented simultaneously rather than successively. If words
and pictures are processed in parallel they will be integrated into a more coherent
model.
Individual Difference Principle–there are stronger effects for low rather than
high prior knowledge learners and high rather than low spatial learners. This principle
suggests that it is particularly important to implement good multimedia design for low
knowledge and high spatial ability learners.
Mayer's model is limited in the extent to which it takes individual differences
into consideration and this aspect requires further research to include other individual
difference factors. Mayer (2001) states: "Some worthwhile venues for future research
include the role of visual and verbal learning-style preferences." (p181) Hence, the
aim of this research is to further delineate individual differences and their impact on
multimedia design.
8
Cognitive style
Cognitive style is seen as “an individual's preferred and habitual approach to
organising and representing information” (Riding & Rayner, 1998, p.11). The
background to cognitive style has been extensively reviewed by Riding and Cheema
(1991) and Riding and Rayner (1998). They conclude that the various style labels
could be accommodated within two fundamental style dimensions-the wholist-
analytic and the verbal-imagery-which may be described as:
Cognitive style has been found to affect a wide range of behaviours, and this
evidence for its construct validity has been considered by Riding and Rayner (1998;
chapters 5-8). Further; style has been found to be independent of measured IQ (Riding
& Pearson 1994) and also of common personality measures (Riding & Wigley 1997).
one. Analytics on the other hand will have difficulty drawing the parts together to see
the whole picture and may concentrate on only one or two parts of the information at
any one time.
and the pictorial. Taken overall, generally Imagers prefer and learn best from pictorial
presentations, while verbalisers prefer and learn best from verbal presentations.
Riding and Read (1996) with 12-year-old children found that the tendency by imagers
to use pictures, and verbalisers writing, increased with ability. There was evidence
that lower ability children were more challenged by the expected format of the subject
than were those of higher ability.
Sherry Chen and colleagues have carried out some work to explore learning
from hypermedia with particular reference to the effect of cognitive style (Chen,
2002; Chen & Macredie, 2002; Ford & Chen, 2000; Ford & Chen, 2001; Ghinea &
Chen, 2003). There is good evidence to suggest that cognitive style is an important
variable to consider within learning models that deal with media mode and media
structure. Chen’s work concentrates mainly on media structure considering aspects
such as navigation, nonlinear learning structures and learner control. Chuang (1999)
however considered whether individual differences such as cognitive style (Field
dependence/independence), gender and prior knowledge (maths) mediate learning
outcome for different multimedia formats which included animation plus text,
animation plus voice and animation, text plus voice. Results indicated that the field
independent group and males scored higher for animation, text plus voice condition
and those high in prior knowledge performed better on the animation plus text
condition and the animation plus voice condition. No results are reported for
interaction effects however the study does show some evidence of important
individual difference effects related to mode of presentation.
groups, the wholist-imagers and analytic-verbalisers, the tendency is the other way
around with the male wholist-imagers better for P-T, and females better for P-S.
Basically with males, complementary groups are best on a separation of the channels
of pictures and words received aurally, while females are best on the single channel of
picture and words. With the unitary groups the males are best on a single channel,
while the females are superior on separate channels. Therefore evidence indicates an
interaction of mode and style with gender, and while the precise nature of this is not
yet clear it is likely to be of practical importance. Basically with males,
complementary groups are best on a separation of the channels of pictures and words,
while the females are best on the single channel of pictures and words. With the
unitary groups, the males are best on a single channel, while the females are superior
on separate channels. This hints at a fundamental gender difference in information
processing which also involves style. Further, gender is an important consideration
when investigating any individual difference factors related to learning as is
evidenced by numerous reports of gender differences in learning worldwide.
Prior Knowledge
Prior knowledge is an important element in cognitive learning theory, with
incoming information being integrated with the aid of prior knowledge, often referred
to as long-term memory (Mayer, 2002). This element of knowledge construction is
seen in Mayer’s multimedia model (Mayer, 2001, 1999) and is also reflected in his
seven principles with the suggestion that badly designed multimedia presentations
will impact more on low prior knowledge (low knowledge) individuals than on high
prior knowledge (high knowledge) individuals. This is supported by cognitive load
theory with its emphasis on schema and automaticity (Sweller & Chandler, 1994). As
individuals develop greater knowledge structures their range of schemas become
greater and their use of automatic processing increases. Consequently, design criteria
become less important because individuals rely more on schemata and automaticity.
Consequently, design effects can be observed for low knowledge individuals while
there are no effects for high knowledge individuals (Kalyuga, Chandler & Sweller,
1998, 2000). Finally, Riding and Read (1996) found that there was a tendency for
verbalisers to use writing and imagers to use pictures/diagrams more so as knowledge
level increased. Clearly, prior knowledge is an important factor to consider when
12
describing learning outcome, but its exact nature is as yet unexplored in combination
with other individual difference factors.
Method
Participants
Two urban state primary schools took part in this study. The participants were
91 children aged 10-11yrs (54 boys, 37 girls) comprising 61 children from school one
and 30 children from school two. All children between the ages 10 and 11 yrs from
school one were included in the study, with no exceptions, and all children from one
particular 10-11 yr class from school two were included, with no exceptions, thus,
giving a good spread of personalities and abilities.
Design
In Condition one the child reads the textual material (Appendix A) whilst
referring to the pictorial information (Appendix B). In Condition two children were
given a narrated version of the textual material whilst referring to the same pictorial
information. Prior to their exposure to the textual or narrated information they were
allowed to study the pictures for one minute in order to familiarize themselves with
the pictures, which were common to both groupings, in order to avoid any confusion
that may have been caused by the pictures themselves. According to cognitive load
Theory (Sweller & Chandler, 1994) Condition one imposes more cognitive load due
to split attention effects than Condition two thus predicting poorer learning in
Condition one. In Condition one participants were required to switch between reading
and looking at the pictures in order to assimilate the information. However, for
Condition two (narrated) children listen to information at the same time as looking at
the pictures. Similarly, Mayer (2001, 1999) would predict via multimedia theory that
Condition two would result in better learning than Condition one due to its use of dual
channels and contiguity of information (Modality Principle and Temporal Contiguity
Principle). Consequently, Condition one (written) is designed to be a poor multimedia
model for instruction and Condition two (narrated) a good multimedia model for
instruction.
Materials
The learning materials for this experiment were taken from a text that attempts
to describe how things in the everyday world work (Macaulay D, 1988). A set of
diagrams illustrating how mammoths could be used as sensors or detectors were taken
from the book and all the writing explaining the pictures was removed including the
14
titles. The only written information remaining on the picture/diagram was referential
text labelling the separate picture groups e.g. Fig 1, Fig 2 etc (see Appendix B).
A sheet of written information about sensors and detectors was constructed
(see Appendix A) by combining the original information from the text describing the
illustrations and a passage containing technical information about real sensors and
detectors. The passage was written in such a way as to be readily understood by 10 -
11 yr. olds. This passage was then recorded on to an audiotape at 107wpm (normal
speech rate being approximately 150wpm.). The audio-taped passage was 3minutes
17 seconds in length and contained 352 words. Six questions were devised to test
knowledge and comprehension that could be assimilated from the text and pictures
together. In attempting to answer these questions the learner needed to split their
attention (Sweller & Chandler 1994) between the two sources of information (textual
and pictorial) in order to assimilate the information. The questions and answers that
were used are shown below:
1. How was the mammoth used to detect burglars?
Answer: By putting its trunk under the mat for the burglar to tread on.
2. How was the mammoth used as a smoke detector?
Answer: By hooking its trunk to the ceiling so that it can smell any smoke.
3. How was the mammoth used as a metal detector?
Answer: The mammoth sits on the cases and anything that is not flattened is metal.
4. How was the mammoth used as a portable breathalyser?
Answer: A person is asked to breath into its trunk.
5. What does the mammoth have to do to make the ski lift rise?
Answer: Drink water.
6. The water is squeezed out of the mammoth when it reaches the ground, what should
then happen to the lift?
Answer: It should return to the ground.
The overall mean for the wholist-analytic dimension was 1.20 (SD.0.49) and
the mean for the verbal-imagery dimension was 1.08 (SD. 0.17). A correlation
between the ratios showed a non-significant correlation coefficient of r = 0.12, thus
indicating independent dimensions. A one-way analysis of variance of the two
dimensions with gender as the independent variable gave no significance for either
dimension, indicating that both dimensions were independent of gender. The two
cognitive style dimensions were then divided into two divisions using a median split
technique to give a balanced split between research conditions on the verbal-imagery
data to enable equal numbers in each research condition. The resulting split as can be
seen below is very close to the divisions suggested by Riding (2000) based on the
secondary schools standardisation study (suggested as wholists 0-1.02; analytics 1.03
– 4+; verbalisers 0 – 1.07; imager 1.08 – 4+).
The resulting divisions were: wholist-analytic dimension: wholist, 0.61 - 1.00
(37); analytic. 1.01 - 2.92 (54): verbal-imagery dimension: verbaliser, 0.77 - 1.06
(46); imager, 1.07 - 1.71 (45).
Procedure
Children were allocated randomly to two groups where gender and cognitive
style were counter balanced (Condition one and Condition two).
16
Condition one children were told that they were going to see some pictures and they
were to study them for one minute prior to being given some more information about
the pictures. They were then told that they were going to be given some more
information about the pictures and that they should study the pictures whilst reading
the information by referring to the appropriate figure (labelled on the pictures and
referred to in text) as they read through it. They were also told that they should only
read through the information once and indicate the moment they had finished reading
through it. Finally they were informed that they would be required to answer some
questions about the information following their reading.
Condition one children were given similar instructions except that they would
listen to some further information about the pictures (rather than read it). Similarly, in
Condition two children were given one minute to study the pictures prior to hearing
the recording of the text.
Although no time limit was given to Condition one (written group), times for
reading were very similar to the times on task for Condition two (audio group), with
children taking between 3minutes and 3 minutes 30seconds (audio recording time = 3
minutes 17 seconds).
Following the presentation subjects were given a piece of paper with spaces
for them to answer the recall questions. The experimenter, who was male, then asked
the questions orally allowing all children enough time to write down their answers.
Children were tested in groups of three or four.
17
Results
Analysis
An ANOVA (Analysis of Variance) was performed on the data and an alpha
level of .05 was used for all statistical tests. Independent variables were gender (2),
verbal-imagery style (2), wholist-analytic style (2), presentation condition (2)
(information written or listened to) and prior knowledge level with the dependant
variable being the number of questions answered correctly in the recall test.
In order to present the size of effect associated with the results, differences are also
expressed as percentage increases or decreases between groups or across conditions.
These percentage changes replace conventional effect size measures (such as Cohen’s
d) to allow the reader to put the changes into context.
Design Effects
Results that showed effects between the two different designs (i.e. across
Condition one –written and Condition two - narrated) are now considered.
If the overall difference in recall is examined between conditions one and Condition
two little difference is observed (percentage difference = +1%). However, if the
relative performance of different groups of individuals according to their gender, prior
knowledge and cognitive style are examined a number of differences are seen.
Table 3 below shows a three-way interaction of wholist-analytic style by
gender by presentation condition (F(1, 28) = 9.20, p < .01). Recall improved from
Condition one to Condition two for male wholists and female analytics. Male
analytics and female wholists scored lower in recall between Condition one and
Condition two.
Discussion
In terms of overall performance the data seems to indicate that the
characteristics of the wholist seem to be better suited for this particular task and
material. This may be due to wholists being better at seeing the bigger picture (Riding
& Cheema, 1991; Riding & Rayner, 1998) (the whole view) and thus better able to
integrate the textual and pictorial information required to answer the comprehension
style recall questions. Analytics would be challenged when assimilating information
from text and pictures, whether of good design or poor design, which was an essential
part of this task for the children. It also appears that female analytics are at a
disadvantage in terms of this particular learning task. This may reflect the difficulty
females have in assimilating information when relying on an analytic style as opposed
to males who perform relatively well with an analytic style. Evidence for these
reversal type effects between different genders but similar styles is evident in other
studies involving mode of presentation (Riding & Grimley, 1999) and further
exploration into these types of effect are warranted. In addition, for this particular task
the wholist-analytic style and verbal-imagery style in combination appear to interact
with gender to show differing information processing characteristics for males and
females. It appears that males and females show different information processing
characteristics in combination with cognitive style.
An examination of the relative performance of different groups of individuals
according to their gender, prior knowledge and cognitive style highlights a number of
differences in outcome across the different design conditions. It appears that cognitive
load and multimedia design effectiveness depend upon the characteristics of the
individual learner. Presentation Condition two gives good outcome results for male
wholists, female analytics, male imagers and female verbalisers but is actually
detrimental to the opposite style-gender groupings. As previously indicated by Riding
and Grimley (1999) a fundamental gender difference in information processing for
particular presentation modes involving style is evident. Riding and Rayner (1998)
20
make comparisons between wholists and imagers and suggest that they are both style
dimensions that have characteristics giving individuals of these style groupings the
benefit of being able to view information from the wholistic/overview perspective.
Additionally, comparisons can also be made between verbalisers and analytics in that
these two styles have an element of both being analytic types of style where
information tends to be viewed in chunks/pieces rather than as a whole. This implies
that the way information is represented by the individual will mediate the processing
strategies, and thus the outcome of the learning task. The task of learners in this study
was very much oriented around the assimilation and recall of distantly positioned
information. Consequently it is possible that this element of the information
processing procedure was a factor in influencing the outcome for different
individuals. Further studies with different task emphasis are needed to explore this
idea further.
Results also show an interaction between prior knowledge and gender with
high knowledge males performing better under Condition one, contrary to what would
be expected given the design characteristics. Previous studies (Mayer, 2001; Kalyuga
et al, 1998, 2000) indicated that low knowledge individuals get greater benefits than
high knowledge individuals from well designed multimedia presentations. These high
knowledge individuals in some cases showed the reverse effect with performance
being better for traditionally designed materials. This finding is also supported in this
study. There is also evidence to suggest that gender is an important factor to consider.
The current study implies that the high-low knowledge rule may only apply to males.
Females appear to require well designed materials for improved learning outcomes
whether prior knowledge is high or low.
Application of good design according to cognitive load and multimedia
theories increased learning outcomes for low knowledge verbalisers by a massive
83%. Therefore this finding supports Mayer’s multimedia model with low knowledge
and low imagery individuals gaining the most from designing the presentation to
comply with this model. Mayer (2001) goes some way to describing how individual
differences might interact with the model proposed, however, it is clear from the
results of this study that further work is needed and elements of cognitive style and
gender need to be included in order to develop a fuller and more comprehensive
model.
21
Conclusion
This study set out to explore whether the multimedia and cognitive load
models could be used generically across all individuals or whether such models
needed to be modified to suit individual needs. This study supports the idea that
cognitive style, gender and prior knowledge mediate learning outcome as predicted by
these two models. In particular there appears to be a mediating effect caused by an
interaction with gender and both the verbal-imagery dimension and the wholist-
analytic dimension. Both of these effects may actually be synonymous with a
particular characteristic of the way an individual organises or represents the material
and is gender specific. Models such as Mayer’s multimedia model and cognitive load
clearly boost our capability to design more effective learning materials, however, any
model designed to improve learning needs to take account of individual differences,
as ultimately this is what learning is about. In particular, given the current climate of
differential educational outcomes between boys and girls it would be desirable to
learn more about gender interactions and their effects on information processing.
Mayer (2001) previously indicated that spatial ability and prior knowledge are
important factors in any model of multimedia design for learning and that they should
be explored further. Research by Riding and Grimley (1999) also indicated that
gender in interaction with cognitive style might be important in mediating any
information-processing model of learning. The study presented here suggests that
further work in this area is indeed warranted. It is evident that all of these factors are
important; however the exact nature of these interactions is still uncertain. Perhaps in
the first instance there is a clear need to attempt to replicate these findings. In
addition, some methodological flaws are evident in that it was difficult to ascribe the
exact time that children spent on the textual version of the task, although timings did
indicate a similar time to the audio version. Therefore there is a need to make sure
that reading time is the same for the narrated version and written version, this may be
possible by presenting the material via computer.
23
References
Riding, R. J., & Douglas, G. (1993). The effect of cognitive style and mode
of presentation on learning performance. British Journal of Educational
Psychology, 63, 297-307.
Riding, R. J., & Dyer, V. A. (1980). The relationship between extraversion
and verbal-imagery learning style in 12-year-old children. Personality and
Individual Differences, 1, 273-279.
Riding, R. J., & Grimley, M. (1999). Cognitive style, gender and learning
from multimedia materials in 11-year-old children. British Journal of
Educational Psychology, 30(1), 43-56.
Riding, R. J., & Pearson, F. (1994). The relationship between cognitive style
and intelligence. Educational Psychology, 14, 413-425.
Riding, R. J., & Rayner, S. (1998). Cognitive style and learning strategies.
London: David Fulton Publishers.
Riding, R. J., & Read, G. (1996). Cognitive style and pupil learning
preferences. Educational Psychology, 16, 81-106.
Riding, R. J., & Watts, M. (1997). The effect of cognitive style on the
preferred format of instructional material. Educational Psychology, 17,
179-183.
Riding, R. J., & Wigley, S. (1997). The relationship between cognitive style
and personality in further education students. Personality and Individual
Differences, 23, 379-389.
Sweller, J. (1988). Cognitive load during problem solving: Effects on learning.
Cognitive Science, 12, 257-285.
Sweller, J., & Chandler, P. (1994). Why some material is difficult to learn.
Cognition and Instruction, 12(3), 185-233.
26
integrating Prior
Knowledge
Selecting images Organising images Long Term
Pictures Eyes Images Memory
Pictorial
Model
Figure 1 Diagram Depicting Mayer’s Multimedia Model. Adapted from Mayer (2001).
27
Appendix A
This passage tells you something about how modern sensors and detectors work, you can see in the
pictures how ancient sensors and detectors worked using the woolly mammoth.
Sensors and detectors are instruments that are used to detect the presence of something and quite often
to measure it. Burglar alarms sense the direct evidence of intruders such as the tell-tale tread of
burglars, smoke alarms detect the presence of smoke particles in the air.
Fig 1 shows the use of the woolly mammoth for detecting unwanted visitors. Mammoths are highly
sensitive creatures. Their physical sensitivity can be used in numerous ways.
Fig 2 shows how the trunk of a sleeping mammoth can be used as a smoke detector. The plants obscure
the creatures bulk and also provide it with the occasional snack.
Other sensors and detectors use penetrating rays or magnetic fields to find objects that can not be seen.
Measuring instruments like radar speed traps are used by the police and are examples of sensors and
detectors that react to something specific and then record its quantity.
It can be seen in Fig 3, 4 and 5 how a highly trained mammoth can be used as a metal detector. One
piece of luggage has been tested and there is no question about the location of bulky items, there’s a
good chance that they are metal. Another type of measuring instrument is the breathalyser used by the
police to detect alcohol on someone’s breath.
In fig 6 and fig 7 the mammoths trunk is employed as a highly sensitive mobile breathalyser.
Sensors and detectors are also very important in automatic machines. Many machines, for example the
autopilot of an aircraft, use feedback. This means that the sensor actually measures the machine’s
performance and then feeds back the results to control the power output. By sensing their own
performance, automatic machines keep within set limits. The mammoth powered weight sensing ski lift
is a simple automatic machine.
Fig 9 shows the specially designed squeezer which forces the water out making the lift return to the
ground.
34
Appendix B
Appendix B Pictorial information for the sensors and detectors learning task