Suzuki, K. (1987). How is an attitude toward a newly learned skill formed?: An inter-domain interaction study. Unpublished Doctoral Dissertation, Florida State University, U.S.A.
How is an attitude toward a newly learned skill formed?
An inter-domain interaction study.
THE FLORIDA STATE UNIVERSITY
COLLEGE OF EDUCATION
by Katsuaki SUZUKI
Unpublished Doctoral Dissertation,
Florida State University, U.S.A.
A Dissertation submitted to the Department of Educational Reserch
fulfuillment of the requirements for
the degree of Doctor of Philosophy
HOW IS AN ATTITUDE TOWARD PRACTICAL USE OF A NEWLY LEARNED SKILL FORMED?
AN INTERDOMAIN INTERACTION STUDY
(Publication No. )
Katsuaki Suzuki, Ph. D.
The Florida State University, 1987
Major Professor: Waiter W. Wager, Ed. D.
This study explored supportive relationships among
objectives from different domains of learning outcomes,
namely, the learning of a concept classification skill,
information, and attitudes toward the skill. Two versions
of an instructional module on the ARCS motivation model were
randomly assigned to pre-service teacher-education students.
One version contained examples of problem solving, using
teachers as models, whereas the other version used business
training as the context of the module. Together with the
''relevance" of the instruction, self-concept of ability,
attitude toward learning of the ARCS model, actual level of
skill acquisition, and final attitude toward the practical
use of the ARCS model formed a hypothesized path model of
causal inference, which was to be empirically confirmed in
The results indicated that the relevant version of the
material had a positive effect on the pre-instructional
motivation for personal commitment. Such an effect,
however, was not found on the acquisition of the skill and
on attitude formation favoring practical use of the ARCS
model. The hypothesized path model was not found to fit the
data in this study. Among the variables in the path model,
a positive relationship was found between the level of skill
acquisition and favorable feeling toward the learned model.
In contrast, the level of skill acquisition was not related
to the behavioral intentions in a statistically reliable
manner. Some alternatives to the hypothesized model were
identified, although the specification of these models needs
to be supported by future research. Results are discussed
in light of the hyposeses, as well as theoretical frameworks
so that the hypothesized rnodel may be confirmed, or revised,
in future research. Implications for instructional design
practices are also discussed.
It has been a great honor to complete a doctrate degree
under the directorship of the finest committee members:
Professors Waiter Wager (major professor), John K. Mayo,
Richard Tate, Robert A. Reiser, and Distinguished professor
Robert M. Gagne,. The committee was tough, but always in a
constructiveway. I learned a lot from them, wh.ich I shall
try to apply in my future work.
I Wish to thank Dr. Marcy Driscoll and Ms. Gail Delicio
who encouraged their students to participate in this study.
I also thank the students of the Educational Psychology
classes who participated in the formative evaluation of the
materials, the pilot studies, and the experiment. Without
their help, this study could not have been completed.
Dr. John M. Keller and the members of Motivation
Research Group provided me with an opportunity to sharpen my
ideas for this study. Dr. Harold Fletcher and the judges of
the Dean and Silby McClusky Award (RTD of AECT) also
provided me with valuable suggestions and comments. I am in
debted to Ms. Brenda Lichfield for her editorial assistance.
I am deeply grateful to the Rotary Foundation of Rotary
International who made my Study in the U.S. possible. My
gratitude extends to the graduate assistantship with the
Center for Educational Technology. I thank Professor Robert
K. Branson, the Director, and his staff members for the
TABLE OF CONTENTS
LIST OF TABLES.........................................vii
LIST OF FIGURES.......................................viii
I. INTRODUCTION ....................................1
Statement of the Problem
II. REVIEW OF THE LITERATURE ........................4
Domains of Learning Outcomes and Interactions
Gagne's Taxonomy of Learning Outcomes
Integration of Cognitive and Affective
Instructional Strategies: Essential and
Intellectual Skill Learning
Use of Path Analysis for Interdomain
Ordering Theory for Hierarchy Validation
Multiple Regression Techniques
Path Analysis for Causal Modeling
III. CONCEPTUAL FRAMEWORK ............................45
The Experimental Manipulation and its Hypothesized Effects
The Hypothesized Causal Model
IV. METHODOLOGY .....................................57
V. RESULTS .........................................69
The Effects of Relevance
The Causal Model
Limitations of the Study and Suggestions for
Instructional Implications of the Study
↑List Of Tables
1. Essential and Supportive Prerequisites for Five Kinds of Learning Outcomes..........................................................8
2. Possible Ranges, Group Means, Standard, Deviations, and Reliability Coefficients..............................................70
3. Bivariate Correlation Coefficients among Predictor Variables.
4. Summary Of the Effects in the Hypothesized Path Model...............................................................................................77
↑ List of Figures
1. Briggs-Wager Matrix for Interdomain
2. Martin-Briggs Taxonomy for Affective Learning...............................13
3. Unit-level ICM Using Twin Objectives........................................17
4. A General ICM for A Intellectual Skill Terminal Objective...................27
5. A General ICM for Learning of An Attitude toward Practical Use of a Newly Learned Intellectual Skill............33
6. Relationships among Motivation, Skill Acquisition, and Post-instructional Attitude toward Skill Use.............40
7. Bloom's Model of School Learning.............................................47
8. Hypothesized Path Diagram Showing lnterdomain Relationships.....................................................52
9. Path Diagram of Effects of Relevance Groups and Academic Self-concept on Attitudes and Performance..............75
10. An Alternative Path Model with All Path being Statistically Reliable.................................................................79
11. A Temporally Revised Path Model which has Marginal Statistical Reliability.
Numerous techniques and procedures in the area of
instructional design and development have been developed
with an emphasis on the cognitive aspects of learning
(Reigeluth, 1983). Little is known, however, about
introducing these cognitive skills so that they will be
utilized in practical situations. The learning of cognitive
skills is not automatically accompanied by a Positive
attitude toward their use (Krathwohl, Bloom, & Masia, 1964).
Many researchers (e.g., Bloom, 1976,. Mager, 1968) have
pointed out the importance of development of a positive
attitude toward what is learned. Such an attitude may
influence the future uses of the skills, as well as further
learning in the same subject area. Therefore, it is
important for designers to understand how strategies to
foster a positive attitude may be built into instruction.
Frameworks for integrating cognitive and affective
domains have been discussed in accounts of instructional
design models. Based on Gagne's (1985) taxonom of learning
outcomes and his notion of the conditions of learning,
Briggs and Wager (1981), Gagne and Briggs (1979), and Martin
and Briggs (19B6) have proposed conceptual frameworks for
integrating various domains of learning outcome. Recent
advancement of a motivational design model (Keller, 1983b)
for enhancing the appeal of cognitive instruction also
suggests integration of the affective and cognitive domains.
In the field of instructional research, very little
research has been conducted to examine the interactive
relationships among learning in various domains of outcome.
One exception is a study by Hurst (1980) that has
demonstrated hierarchical relationships among cognitive and
affective objectives. Using the framework of “interdomain
interactions” (Briggs & Wager, 1981), another study has been
conducted (Young, 1986) to examine the effectiveness of
sequencing among instructional units for intellectual
skills, verbal information, and attitude learning.
Reflecting on the lack of empirical support, Briggs (1982;
Martin & Briggs, 1986) has called for research in
interdomain interactions as one of the prospective fields of
instructional design research for the future.
Statement of the Problem
This study explored interdomain relationships among the
learning of an intellectual skill, information on its
potential usefulness, and attitudes toward learning of the
skill and toward its practical uses. More specifically,
using the conventional experimental paradigm, the purpose of
the study was to examine the effects of an introductory
presentation and use of concrete examples that make the
context of the skill learning ''relevant'' on motivation to
learn, skill acquisition, and attitudes toward the practical
use of the learned skill. A path model for causal
relationships among self-concept of ability, relevance group
assignment, skill acquisition, and attitudes was
hypothesized and tested in this study, With a special
interest on the relationship between initial success in skill
acquisition and attitudes toward practical use of the skill.
REVIEW OF THE LITERATURE
This study was drawn from research in instructional
strategies for both cognitive and affective objectives.
This review is presented in three parts: (1) the interaction
among the five domains of learning outcomes, (2) factors
influencing the effectiveness of instruction aiming at both
intellectual skills and attitudes, and (3) the use of path
analysis for causal inference.
Domains of Learning Outcomes and Their Interactions
The purpose of this subsection is to review Gagne's
five domains of learning outcomes and their interactions.
The relevance of Gagne's taxonomy as the framework for
designing instruction and the need for investigating
interactions among the domains will be discussed.
Gagne 's Taxonomy of Learning Outcomes
Because instructional research seeks to identify
factors that facilitate learning, any Proposed instructional
design strategies should have their basis in learning
theory. Current information-processing theories within
educational psychology have provided the strongest models
hypothesizing how human learning may take place.
Consequently, a number of instructional strategies that
support one or more stages of information-processing have
been proposed. Such instructional strategies, when
effective, may well be transferable to the teaching of a
variety of other similar skills.
Gagne's (1985) taxonomy of learning outcomes has been
widely employed in the context of instructional design for
its special attention to instructional design applications
and its firm underpinnings in current learning psychology.
It provides a way to categorize human capabilities for which
instruction may be designed. For each category, Gagne
hypothesizes a set of ideal conditions of leaning. These in
turn may be employed in instructional strategies for each of
the learning outcomes. In other words, each category of
Gagne's taxonomy represents a different kind of learning for
which different sets of external and internal conditions
would be optimal. According to Gagne, the categories of
human capabilities, or learning outcomes, are: Intellectual
skills, Cognitive strategies, Verbal information, Attitudes,
and Motor skills.
Among these five domains of learning outcomes,
instructional strategies for intellectual skills have been
the most widely Studied. The intellectual skills domain
consists of concept learning, rule or procedural knowledge,
and problem solving skills. Research of Gagne (1968), White
(1973), and White and Gagne (1974) supports rules for
sequencing intellectual skill objectives based on
prerequisite relationships among the subordinate enabling
skills. Designers employing these rules for sequencing
intellectual skills can develop more effective instruction
than those who are unaware of the hierarchical structure.
This emphasis of research on intellectual skills has
been partly because their importance as building blocks in
school learning (Gagne & Briggs, 1979). A more practical
reason may be that intellectual skills are more clearly
organized and easier to assess than capabilities in other
The verbal information domain requires a different set
of instructional strategies from the intellectual skills
domain. Verbal information consists of declarative
knowledge, presumably in the form of propositibns. Whereas
learning of an intellectual skill is facilitated by recall
of specific subordinate skills, the learning of verbal
information is facilitated by recall of a more general,
meaningful contextual knowledge structure, into which new
information may be subsumed. Instructional strategies for
verbal information learning include the use of advance
organizers (Ausube1, 1960) and various mnemonic strategies.
Attitudes are defined as “internal states that
influence the individual's choice of personal action [toward
some category of objects, persons, or events]” (Gagne, 1985,
p. 219). They constitute the learning outcome of the
affective domain in Gagne's taxonomy. Although an attitude
has been considered to have cognitive, affective, and
behavioral components (e.g., Rosenberg & Hovland, 1960),
little is known about the supportive functions of cognitive
skills for attitude learning. Gagne emphasizes the
behavioral component of an attitude for its importance in
instructional implications: “however important they may be
in understanding the essential nature of attitudes,
[cognitive and affective components] give few clues
regarding the function of attitudes” (p. 240).
Instructional strategies known to be effective for attitude
learning include the use of human modeling (Bandura, 1977)
and the experience of success in an activity.
Integration of Cognitive and Affective Domains
Expanding the notion of prerequisites for learning,
Gagne (1977,. Gagne & Briggs, 1979) has identified both
essential and supportive prerequisites for each domain of
learning outcomes. Table 1 lists the kinds of prerequisites
that indicate possible contributions across the domains when
one designs instructional strategies. As Gagne has already
pointed out, learning hierarchies deal only with essential
prerequisites for intellectual skill learning. For example,
attitudes, cognitive strategies, and verbal information may
aid in the learning of intellectual skills, but they are not
essential. Further examples include essential prerequisites
for attitude learning. Intellectual skills and verbal
information related to the choice behavior are sometimes
necessary as prerequisites for the learning of attitudes.
A more elaborate framework for integration of Gagne's Table 1
Essential ans Supportive Prerequisites for Five Kinds of Learning Outcomes
|Type of Learning|
Organized Sets of
From: Gagne,R.M. Analysis of objectives. In L. J.
Briggs (Ed.) , Instructional Design: Principles and
applications. 1977 Educational Technology
five domains of learning outcomes has been suggested by
Briggs and Wager (Wager, 1976,. Briggs & Wager, 1981) as a
matrix of domain interaction (Figure 1). Domain interaction
is defined as ''how the objectives in different outcome
domains may be expected to support each other'' (Briggs &
Wager, p. 89). These authors have also pointed out the lack
of attention to the affective domain in the context of
instructional design: ''It is probably Safe to say that very
little attitude instruction is consciously designed--if it
were, the nature of instructional materials would likely be
quite different from materials typically produced'' (Briggs &
Wager, p. 45). Use of “twin” objectives is recommended to
emphasize the importance of attitude objectives in designing
instruction, pairing an attitude objective with the terminal
intellectual skill objective of each unit of instruction.
The contribution of the work by Briggs and Wager was
strengthened by their operational tool for interdomain
interaction called Instructional Curriculum Maps (ICMs).
ICMs provide a way to describe how the objectives in
different domains may Support each other in the learning
process. By utilizing ICMs for designing a lesson, unit, or
course, interrelationships among the component objectives
within an instructional sequence can be diagrammed. The ICM
presumabl Serves as a visual analytic tool for
instructional design that facilitates planning for domain
interaction to make the instruction more effective.
Examples of ICMs can be found in the remaining parts of the paper.
From the perspective of motivation to learn, Keller
(1979, 1983a, 1983b,. Keller & Dodge, 1982,. Keller & Kopp, in
press,. Keller & Suzuki, in press) has provided instructional
designers with the conceptual framework of, and concrete
strategies for, enhancing the motivational property of
instruction. Motivation is defined as "the choices people
make as to what experiences or goals they Will approach or
avoid, and the degree of effort they will exert in that
respect'' (Keller, 1983b, p. 389). Keller's ARCS model
conceptualizes human motivation using four factors:
Attention, Relevante, Confidence, and Satisfaction. The
concepts and strategies included in the model aim to specify
the learning conditions so that instruction can be not only
effective and efficient, but also appealing.
The two psychological constructs, motivation and
attitudes toward learning, seem to share a great deal of
common external and internal conditions of learning, making
motivational research valuable in integrating cognitive and
affective domains. In other words, when motivation for
further learning (or “continuing motivation”, Maehr, 1976;
Martin & Briggs, 1986) is used as an accompanying learning
outcome of cognitive instruction, it should be treated as
an attitude objective, as Briggs and Wager (1981) suggested
by the use of “tvin objectives.” Keller's ARCS model
(1983a) includes procedures for specifying motivational
objective along with the cognitive objective, in order to
enhance the motivational property of cognitive instruction.
Thus, the ARCS model offers a useful conceptual and
practical framework for the integration of cognitive and
affective dornains in instructional design.
Another conceptual framework for integrating affective
and cognitive domains has recently been suggested by Martin
and Briggs (1986). It includes a new affective taxonomy for
instructional research and design, internal and external
conditions of learning of each component of the taxonomy,
and the ''capability verb,. (Gagne & Briggs, 1979) for each
component. The affective domain taxonomy ccnsists of well-
studied psychological constructs that may have implication
for instructional design. Self-development is placed at the
apex of the taxonomy that contains ten additional
subordinate components as shown in Figure 2.
Martin and Briggs (1986) consider their work an
expansion of Gagne's work in the cognitive domains (Gagne,
1985) to the affective domain, as the basis for integration
between the two. Martin and Briggshave listeda total of
230 external conditions that facilitate either a cognitive
or an affective objective, for each of their categories of
learning outcomes. Through the work of Martin and Briggs,
which has been derived primarily from the vast research in
social psychology and communications, instructional
researchers may now have an easy access to a list of
instructional strategies for affective learning. Their
taxonomy of the affective domain has also provided possible
interrelationships of major psychological constructs, each
of which has long been studied by different researchers in
an isolated manner.
Martin and Briggs's work has resulted in an expansion
of Gagne's (1985) work within affective domain in terms of
numbers of constructs covered and numbers of conditions of
learning listed. However, the nature of the affective
domain has not yet fully been revealed, including its
possible interaction with the cognitive learning. .It is
still uncertain, for example, whether instructional
strategies can be or need to be designed distinctively for
each subdomain in the Martin-Briggs's affective taxonomy.
It may well be that conditions for attitude learning
identified by Gagne are commonly applicable across most
affective learning normally covered by instructional design
effort. It should be kept in mind that dividing learning
outcomes into subdomains is necessary and useful only When
each divided subdomain requires a distinctive set of
internal and external conditions of learning, which can be
facilitated by different instructional strategies.
There are two empirical studies (Hurst, 1980,. Young,
1986), recognized to date in the field of instructional
design, that have examined the integration of cognitive and
affective domains. In Hurst-s (1980) study, both cognitive
and affective objectives were arranged hierarchically. The
terminal objective was to voluntarily implement Individually
Guided Education (IGE). Twenty-nine elementary school
teachers who were in the process of implementing a new
curriculum, IGE, were observed, interviewed, and tested
regarding attainment of the objectives related to the
terminal objective. No instruction took place to facilitate
the learning of the terminal objective; only the assessment
data were collected and analyzed in the study.
The hypothesized hierarchical relationships and
empirically derived ones were “moderately similar to each
other” (p. 299), supporting that “cognitive skills and
attitudes were integrally related” (p. 299). More
specifically, empirically derived ordering of the objectives
indicated: (1) Affective skills that are not heavily
dependent on cognitive knowledge (e.g. , awareness, interest,
willingness to receive information) form the bottom layer of
the ordering. (2) Affective skills with substantial
cognitive components (e.g., voluntary engagement for learning
and application of the IGE) are built on the acquisition of
at least some of those cqgnitive components. Hurst's study
was a pioneering effort in the investigation of integrated
relationships among objectives in both cognitive and
Using the Briggs-Wager (1981) framework for the
interdomain interaction research, Young (1986) has conducted
an empirical study regarding the sequence of the objectives
of verbal information, intellectual skills, and attitudes in
her college physical education course. Young did not find
any significant difference in achievement over a five week
period among three treatment conditions: (1) verbal
information and intellectual skill were taught before
attitudes, (2) attitudes were taught before cognitive task,
and (3) the three domains were integrated in each lesson.
Young did not find any statistical difference of attitudes
toward physical education as measured by a semantic
differential scale and a behavioral differential scale among
the treatment groups at the end of the experimental unit.
It was found, however, that the attitude-first group
retained the positive attitude more than the cognitive-
skill-first group until the end of the course.
It was demonstrated by Young's (1986) study that the
Briggs-Wager's (1981) ICM can be utilized effectively to
plan and implement instruction for objectives in more than
one domain of learning. The "twin objectives" were used at
the unit level to plan not only the cognitive aspects of
instruction, but also the attitudes toward what was to be
learned. Figure 3 shows the ICM with the ''twin objectives''
produced for Young's study.
Although Young's (1986) study intended to apply the
Briggs-Wagerls framework of “interdomain interaction”, it
was weak in distinguishing the domains of verbal information
and intellectual skills. As may be seen in Figure 3, most
of the cognitive objectives in Young's study belonged to the
verbal information domain. Further examination of the
enabling objectives in each of the lessons reveals that
cognitive instruction was mainly targeted to the acquisition
of verbal information. Thus, it is unclear whether or not
her findings are also applicable to the interactive effects
of cognitive and affective domains when the emphasis of
instruction is an intellectual skill. Theoretically, the
conditions of learning for a verbal information objective
and an intellectual skill objective are distinctive. Thus,
it may well be that the nature of interdomain interactions
would be different betweeh domains of attitudes and
intellectual skills and those of attitudes and verbal
information. Further research is needed to investigate the
interdomain interaction between domains of intellectual
skills and attitudes.
In addition to the two studies mentioned above, which.
are directly concerned with the integration of cognitive and
attitude learning, assessment of attitudes has been included
as of secondary interest in some studies dealing with
instructional strategies for cognitive learning. For
example, Ross and his associates (Ross, 1983, 1984,. Ross &
Bush, 1980; Ross, McCormick, Krisak, & Anand, 1985) have
conducted a series of studies to adapt the context of
mathematical instruction to students' background. The major
concern of their studies was the posttest performance that
was designed to measure acquisition of mathematical
concepts. Additionally, they reported the participants'
“reactions to the learning task” (Ross, 1983, p. 521) as
measured by a semantic differential scale. Although the
manipulation of the instructional context was intended to
have, and actually did have, an effect on cognitive
learning, Ross and his associates have found that using
farniliar examples and practice questions may also have an
effect on the attitude toward what was being learned. This
type of research indicates that one instructional strategy,
or an external condition of learning, may have an effect on
both cognitive and affective learning, regardless of the
Instructional Strategies: Essential and Supportive
Prerequisites for the Learning ofIntellectual Skills
The purpose of this subsection is to review the
instructional strategies for intellectual skill acquisition
and attitude formation. Essential and supportive
prerequisites for learning of intellectual skills will be
reviewed with a focus on a concept classification skill,
which will be followed by a review of instructional
strategies for attitude learning.
Instructional Strategies for Intellectual Skill Learning
According to Gagne and Briggs (1979), only the
essential prerequisites for learning an intellectual skill
are subordinate intellectual skills that are included in a
learning hierarchy (Table 1). This indicates that the
domain of intellectual skills can be self-contained in and
by itself. It is the basic premise of Gagne's learning
hierarchy that all of the essential prerequisites can be
identified for any given intellectual skill objective by
constructing a learning hierarchy. Thus, design of
instruction for an intellectual skill may be built around
the hierarchy, covering all of the minimum requirements for
the skill acquisition. The limitation of components of a
learning hierarchy to intellectual skills was empirically
supported by white (1974b), where validation of a learning
hierarchy for a rule using skill in physics revealed that
verbal information objectives needed to be excluded except
at the very bottom of the hierarchy.
The most important instructional strategy for'
intellectual skill learning is, therefore, to assure that
all of the subordinate skills are brought into the learner's
working memory. Thus, it is critical to facilitate recall
of those prerequisite subskills before presenting a new
learning task. It is also effective to provide
instructional feedback that is tailored for remediation of
the missing subordinate skill(s) when an incorrect response
occurs. Thus, the use of instructional media, such as
computers, which have capability to provide precise feedback
depending on the learner's response are recommended (Reiser
& Gagni, 1983).
It is claimed that cognitive strategies and verbal
information have supportive relationships with skill
learning (Table 1). Verbal information may provide labels
for the concepts, contextual cues for retrieval, or a
linkage between concepts previously learned and the new
concept, thus facilitating the learning of a new
intellectual skill. Cognitive strategies, when initiated by
a learner at the right moment in skill learning, may speed
the learning process of the skill, aid recall of the skill,
or help transfer the skill to a new situation. Thus Gagne
and Briggs (1979) recommend the inclusion of these
supportive prerequisites when designing intellectual skill
A conceptual framework for the supportive relationships
of verbal information to learning of intellectual skills can
be found in instructional research. White and .Mayer (1980)
have proposed a classification system for knowledge
associated with a skill. According to this framework; a
learner has flan expanded capacity which is known as
understanding of the skill, and which is manifested in
greater ability to apply the skill in new situations" (p.
102) when verbal and visual knowledge of the productive type
is associated with the learned skill. They have listed the
possible types of ''productive knowledge'': analogies,
concrete examples, definitions of concept, and explanations
of rules. In contrast, they have suggested that formal
statements of the rule, historical facts, and computation
may be "unproductive" in aiding understanding of the skill.
Empirical studies have also demonstrated the supportive
effects of related verbal information on skill acquisition.
Beeson (1980) has reported that meaningful learning and
retention of most complex skills were enhanced by treatment
with relevant anchoring context. That is, participants
learned and retained the target intellectual skill better
when additional verbal instructions were given to provide a
more familiar context to the skill learning than when the
skill was taught in isolation. Similar results have been
obtained in the studies of advance organizers for the
learning of intellectual skills (e.g., Mayer & Bromage,
1980). Although the distinction between the two domains as
the learning outcomes remains to be important, adding
instruction for related verbal information may be an
effective instructional strategy for intellectual skills
Another interpretation regarding how these three
cognitive learning outcomes may interact with each other can
be found in the schema theory (e.g., Rumelhart, 1980) of
human memory structure. According to the schema theory,
human memory is the network of the units called ''schemata.''
A schema is formed around a central concept, which provides
a structure (containing ''slots'') for incoming information.
A schema is therefore composed of; (1) one or more concepts,
(2) cognitive strategies that indicate how the schema is to
be used or related to other schemata, and (3) accompanying
information. Thus, all three kinds of cognitive learning
outcomes may be found in a schema (Suzuki, 1987).
Although the schema theory is too young to fully
explain the learning process of a schema (Bransford, 1984),
several schema theorists (Rumelhart & Norman, 1978,.
Anderson, 1982) have proposed that there is more than one
stage in schema acquisition, which seem to be in
correspondence with Gagne's (1985) learning outcomes. That
is, incoming information may be learned as verbal
information (accretion; declarative stage) until schema
structure is forced to be altered or another schema is
formed (tuning or restructuring; procedural stage) as the
result of intellectual skill learning. Therefore, it may be
that not only Verbal information learning can be facilitated
by activating existing schema structure as in Ausubel's
advance organizers, but also intellectual skill learning
Would be facilitated by recalling relevant information or by
initiating related cognitive strategies, as well as by
recalling subordinate skills. This notion seems to have
been partially Supported by the studies that adapted context
to students backgrounds (e.g., Ross, 1983) and that trained
students in the use of cognitive strategies prior to skill
learning (e.g., Derry & Murphy, 1986).
It may be interpreted that the demonstrated
facilitative effects of advance organizers in learning of an
intellectual skill (e.g., Mayer & Bromage, 1980) Would have
been obtained by this “supportive” effect of schemata
activation. It should not be overgeneralized that domains
of intellectual skills and verbal information require the
same conditions for the instruction to be optimal.
Contextual clues may only be “supportive” in intellectual
skill learning, while they may be "essential" in verbal
information learning to be meaningful. only after all the
subordinate skills are considered in instructional design,
should other cognitive domains be considered as supportive
It is also noteworthy that the interdomain
relationships among the three domains of cognitive learning
outcomes are no less important than the relationships
between an attitude objective and an objective in one of the
cognitive domains. When integration of cognitive and
affective domains is planned, it is critical to classify the
cognitive objective into one of the three domains of
learning outcomes to examine the nature of the cognitive
task. Until this classification of the cognitive objective
is made, integration of the task and affective consideration
Attitudes have been recognized as supportive
prerequisites for skill learning (Table 1). Bloom (1976)
has identified various ''affective entry characteristics''
that may have an effect on cognitive school learning:
interests, attitudes, motivation, subject-related affect,
school-related affect, and academic self-concept. In his
review of related literature, Bloom concluded that “academic
self-concept is the strongest of the affect measures in
predicting school achievement...it accounts for about 25
percent of the variation in school achievement after the
elementary SChool period” (p. 95).
The strong correlation between academic self-concept
and cognitive learning may be because academic self-concept
is based on the cumulative effect of achievement in relation
to a learner'sv peer group. The more successful a learner
has been, compared to his or her classmates, the higher
academic self-concept is likely to be. Therefore, academic
self-concept may represent not only the affectlVe entry
characteristics of a learner, but also the cognitive
capability of the learner in general sense. Howeverl it has
a limited usage as far as instructional design at post-
secondary level goes because academic self-concept is
relatively stable once established through the experiences
The expectancy-Value theory (Atkinson, 1964; Keller,
1983b) of motivation to learn has postulated that the value
the learner sees ina learningtask is also an impbrtant
factor that determines how extensively a learner exerts hjLS
or her effort to learn the given task. Academic self-
concept is considered to represent the "expectancy" of the
learner with regard tohis other success in learning the
task. on the other hand, the learner needs to be convinced
that the task is worth learning (i.e., “value”) before
becoming fully involved. The expectancy-Value theory has
proposed relating the two factors in a multiplicative
manner, rather than an additive manner (Keller, 1983b).
That is, if either of the two factors were lacking, then
learner motivation would be zero. It is therefore
recommended to take both aspects of motivational strategies
into account in order to assist the cognitive learning
through supporting attitudes.
The trend in motivational research has been to place
the emphasis mainly on the expectancy aspect. A recent
review of instructional psychology has described that
attribution theory (Weiner, 1985) has been strongly
influencing the motivational research, but that “task value
has not been included in most attributional models”
(Pintrich, Cross, Kozma, & McKeachie, 1986). Some research
has nevertheless shown the effect of manipulating the
perceived task relevance on cognitive learning by adapting
the context to learners' backgrounds (e.g., Ross, 1983), or
connecting the new material to more familiar concepts (e.g.,
Mayer, 1979). Keller (1983a, 1983b,. Keller & Dogge, 1982;
Keller & Kopp, in press; Keller & Suzuki, in press) has
provided a list of instructional strategies to generate
relevance (value factor) as well as to build confidence
(expectancy factor) in various instructional situations.
Further empirical examination of the effects of task
relevance on the cognitive learning is still needed.
In summary, the essential and supportive prerequisites
for intellectual skill learning are reviewed in this
section. Subordinate intellectual skills are essential to
the learning of an intellectual skill. This has been well
established by the past research. Although verbal
information, cognitive strategies, and attitudes have been
identified as supportive prerequisites, further research is
needed to examine the effects of various instructional
strategies on acquisition of intellectual skills. Figure 4
depicts a generic ICM (Briggs & Wager, 1981) that shows
prerequisites relationships among the various domains for
intellectual skill learning.
Instructional Strategies for Attitude Learning
Attitude instruction, in contrast with intellectual
skill instruction, has obtained little attention in
instructional design literature (Briggs & Wager, 1981;
Reigeluth, 1983). Much of the rationale for the
instructional strategies in attitude learning has originated
from the fields of social psychology (e.g., Ajzen &
Fishbein, 1980; Bandura, 1977,. McGuire, 1969) and
communications (e.g., Rogers, 1983; Rosenberg & Hovland,
Martin and Briggs (1986) have reviewed the vast
literature related to attitude and attitude change, mostly
in the fields of social psychology and communications. They
identified the following eight external conditions that span
all types of affective learning:
1. Provide cognitive information that is new to the
learners, or that is presented in a new way.
2. Use successive approximations either to break a task
into smaller units so success can be achieved, or to
gradually increase the learners' cognitive base or
tolerance for an idea.
3. Model attitudes, values, emotions, etc., that are
consistent with the desired behavior.
4. Use group discussions or social interactions (one-
to-one, small groups, role plays, etc.) to assist
learners to (a) see another position, (b) take
anothervs perspective, (c) verbalize their own
position, and/or (d) solve problems.
5. Use direct reinforcement to (a) establish attitudes,
emotions, values, etc., when there is consensus on
the desired attitude, emotion, value, etc., and (b)
reward cooperation, participation, independence,
6. Match the learners' task to their abilities; strive
for a moderate level of difficulty.
7. Provide opportunities for learners to take an overt
8. Use the principle of contiguity to help learners
associate learning in general (school, training
sessions), and specific learning (affective or
cognitive knowledge and skills) to a pleasant,
stimulating environment. (p. 457 - 458)
Similar strategies and more have been included in
Keller's ARCS motivational model (Keller, 1983a, Keller &
Kopp, in press,. Keller & Suzuki, in press). Keller has
identified four factors that pertain to human motivation to
learn: Attention, Relevance, Confidence, and Satisfaction.
Keller has suggested the analysis of motivational
characteristics of learners accordir.g to the ARCS four
factors to match the motivational property of the
instruction with learner characteristics. Motivationally
enhanced instruction should not only facilitate cognitive
learning, but also help the learners attain higher level of
motivational characteristics at the end of instructional
activity. Thus, audience analysis using the four factors
and consequent formation of motivational objectives is
recommended as a partof theARCS model.
Based on the paradigm of expectancy-value theory,
Keller's relevance category includes such strategies as
making instruction familiar to the learner using a human
model, orienting the learner to a meaningful goal, and
matching instructional situation to learner's motives and
needs (i.e., needs for power, achievement, and affiliation).
The relevance factor is designed to create “value” of
instruction. In contrast, confidence category is
represented by “expectancy” building strategies. It
includes strategies to arrange the learning requirements
from attainable to more difficult to provide success
opportunities. Instructional strategies such as giving
personal control over the learning situation and attributing
success to learner's effort and ability are also considered
to have an effect on the ''expectancy'' aspect for continuing
Gagne (1985) has identified three major situations for
attitude learning: classical conditioning, perception of
success in behavior, and human modeling. Classical
conditioning for attitude learning refers to the pavlovian
type Conditioning of paired association learning, where
unconditioned unpleasant stimuli causes an affective
reaction. Reinforcement contingencies also provide attitude
learning because they let the learner perceive “experience
of success” with some object. Gagne has stated: “Positive
attitudes toward mathematics or English or public speaking
follow one or more experiences of success in these
activities. Conversely, attitudes of dislike result from
repeated instances of failure” (p. 231). Finally,
observational learning of an attitude can be facilitated by
vicarious reinforcement (Bandura, 1977), where a learner
observes a human model being reinforced for his or her
positive behavior toward the attitude object.
Based on these situations for attitude learning, Gagne
(1977, 1985; Gagne & Briggs, 1979) has identified
prerequisite capabilities for attitude learning. First, “the
learner must possess the concept of the class of object,
event, or person to which the new (or newly changed)
attitude will be directed” (Gagne, 1985, p. 347). Another
type of intellectual skill may be required so the learner
will be able to take an action of personal choice favoring
anattitude object. When a human model is employed as an
instructional strategy, the learner must recognize the model
as a credible source of information. Similarly, the learner
has to be able to comprehend the content of the persuasive
communication for the message to be effective.
As indicated in Table 1, intellectual skills and verbal
information are sometimes, but not always, essential
prerequisites for some type of attitude learning. In other
words, prerequisites for attitude learning are dependent
upon the kind of cognitive learning involved to comprehend .
attitude instruction and to be able to make a personal
choice in favor of the attitude object, person, or event.
Thus, when an attitude to be learned is to voluntarily
employ a new problem solving technique for motivation in
school learning, for example, the learner must be able to
recognize what motivation is (i.e., a concept learning). The
learner also has to realize the problem situations to which
the technique can be applied (i.e., verbal information
prerequisite). Then the learner must be able to use the
technique (i.e., the higher-order rule to use the problem-
solving technique.) A presentation of a school teacher who
has been successfully using the technique may be employed as
human modeling, in which case the learner must recognize the
model as a respective or familiar figure.
For the attainment of the “twin objectives” of an
intellectual skill and an attitude toward the use of the
skill, Briggs and Wager (1981) have proposed a sequence of
instruction among the prerequisite objectives..
“With regard to changing an attitude, the learner may
be stating the value of something, the rationale for
something, etc....The most efficient instructional
decision would be to elicit the verbal information
behavior first, since a more positive attitude should
facilitate attainment of the intellectual skills, and
it takes far less time for the learner toattain the
verbal information objective.” (p. 94).
This can be seen a situation where a verbal information
objective is serving as an essential prerequisite for
attitude learning, and at the same time as a supportive
prerequisite for intellectual skill learning. A generic ICM
can be drawn to represent this situation as in Figure 5.
Alternatively, if the attitude to be learned is heavily
social in nature (e.g., toward minorities, toward birth
control), the emphasis of the instruction may be different.
Ajzen and Fishbein (1980) have proposed the “theory of
reasoned action“, in which behavioral intention is affected
by the relative importance of attitudinal and normative
components. In other words, behavioral intention and actual
behavior are affected not only by the attitude toward an
action, but also by the way important referents are
perceived and by a motivation to comply with the referents.
For instruction of this type of attitudes, use of a human
model and social interaction should become a more important
determiner of the effectiveness of instruction than of
attitudes toward cognitive skill learning.
Another factor that may be operating to determine the
nature of attitude instruction is whether or not the learner
has already acquired the attitude in an undesirable
direction. It may require different instructional
strategies to try and change the existing attitude than to
merely form a new one. Wager (1975) has suggested that the
age of the learner and whether or not the attitude is tobe
established or changed may affect the selection of media.
To establish attitude in adult or change the attitude of
young people, alrnost any medium may be effective, but to
change attitudes of adult or establish attitudes in younger
people, enriched and more concrete messages may be more
effective. Wager's rationale for this distinction is that
the adult is mature enough to ''formulate attitude from
purely 'abstract' messages'' and that lIthe adult is more
critical of the credibility of the source of verbal message"
(p. 11). Wager has also pointed out that “'established'
attitudes will be, in general, more difficult to change than
recently formulated attitudes in adults or attitudes in
children (that have not had as long a history of
reinforcement)” (p. 11).
Although human modeling as an instructional strategy
for attitude learning usually includes a physical being,
human modeling can still be adopted in an abstract message.
For example, a description of human behavior may be used in
an anecdotal fashion. Reiser and Gagne (1983) have
recommended the use of motion-visuals for selecting
instructional media attributes for attitude learning so that
motion of a human model can be displayed. However, as Gagne
(1985) has pointed out, "[p]resumably, human modeling
represents the basic psychological process involved in the
acquisition of values from the reading of history and
literature'' (p. 239). Although forms of human modeling may
differ for different task-learner situations, human modeling
can be an effective strategy for any type of attitude
In summary, two general categories of effective
instructional strategies for attitude learning are (1) .
experience of success and (2) human modeling. An attitude
may be formed by generating relevance of the desired
personal action in relation to personal motives or needs
(i.e., value aspect) and by building confidence for
successful experience (i.e., competence aspect). Two
intervening factors for the effects of instructional
strategies are the learner's age or maturation and whether
ornot the attitude is to be newly established or to be
changed. Interrelationships among the prerequisites for
attitude learning are represented as an ICM in Figure 5.
Use of Path Analysis for
Interdomain Interaction Research
The purpose of this subsection is to review the
methodology for research into interdomain interaction.
ordering theory, multiple regression technique, and path
analysis will be reviewed in turn.
Ordering Theory for Hierarchy Validation
Ordering theory (Airasian & Bart, 1973,. Bart & Krus,
1973) has been used as a technique to validate a
hypothesized learning hierarchy. It was also used in the
study by Hurst (1980), Which demonstrated the hierarchical
relationships among the objectives in cognitive and
affective domains. Ordering theory was chosen by Hurst
because it “analyzes all intertask relationship (therefore
generating hierarchical sequences that were not
hypothesized) and has the special capability of defining
nonlinear sequences” (p. 299). Each pair of the subtasks is
examined for the percentage of “discomfirmatory” responses,
in which ahigher level task is passed without passing a
lower level task. When the ratio of discomfirmatory
response to the whole sample does not exceed the preset
“tolerance level” (e.g., 7% in Hurst's study), it is said
that a prerequisite relationship exists.
Ordering theory, however, has limitations as it was used
in Hurst's study. First, it uses only dichotomous data (i.e.,
passing or failing for each objective). Data would be lost
if more than one item was used for an objective by deciding
the passing criterion. Assignment of pass or fail seems to
become arbitrary especially for attitude objectives.
Second, no statistical test of significance is applied to
the data obtained from a sample, which makes the
interpretation of the results determin.istic (Hurst, 1980).
In other words, errors of measurement and sampling
distribution are not taken into account. Similar points
have previously been identified by white (1974a) in his
review of traditional techniques for hierarchy validation.
More importantly, hierarchical arrangement of the
cognitive and affective components of a terminal objective
may be misleading in terms of the nature of prerequisite
relationships among the components. A learning hierarchy
consists only of ''essential'' prerequisites, which are
“required” for the learning of a superordinate skill. Thus,
one can utilize a learning hierarchy to determine the
subordinate skills that, “the absence of which would make it
impossible [for a learner] to learn the new [skill]” (Gagne,
1985, p. 272). In contrast, it does not seem to be adequate
to conclude that “either cognitive or affective terminal
skills could be analyzed to determine the cognitive and
affective elements required to reach any given terminal
skill” (Hurst, 1980, p. 302, emphasis added), based on the
lack of disconfirmatory data without any statistical
inference. The elements that are analyzed by ordering
theory may have supportive relationships with each other,
but it is not necessarily the same relationships as
originally meant by the use of learning hierarchies. Thus,
it may be concluded that ordering theory as used in Hurst
study lacks appropriateness as the measurement model for
interdomain interaction research.
Use of Multiple Regression Analysis
An alternative to ordering theory to investigate
interdomain interactions is the use of multiple regression
analysis, which was adopted in the study by Young (1986).
Young selected the multiple regression analysis “because the
study contained a mixture of attribute and treatment
variables” (p. 59). By applying the multiple regression
technique, each of the identified “predictors” can be put in
an equation to predict the theoretically expected value of a
dependent variable, indicating how much variability of the
dependent variable is explainable by each predictor.
Multiple regression is, in this sense, parallel to analysis
of variance with covariates.
When the theoretical argument that all predictors that
have a direct effect on the dependent variable are included
in a multiple regression equation is defensible, not only
the correlational interpretation, but also a causal
interpretation of the results may be possible. For example,
a researcher might manipulate experimental conditions to
motivate learners to study an intellectual skill to see the
effects of motivational enhancement on skill acquisition.
If a task-specific pretest score is entered in a multiple
regression equation with the level of motivational
enhancement (i.e., 0 or 1), one can analyze the effect of
motivational treatment by controlling prior skill
differences. However, this technique may not be able to
provide theoretically adjustable causal interpretation
because it does not take into account “entry affective
characteristics” that have been argued to have a strong
effect on performance (Bloom, 1976). one way of accounting
for entry affective characteristics is to add a measure of
academic self-concept to the regression equation, then a
significant contribution of motivational treatment would be
hypothesized to have caused the skill acquisition.
One limitation of the multiple regression analysis is
that complex causal flows that are usually embedded in
interdomain interaction research cannot be considered due to
the lack of independence among predictor variables. For
example in the motivational treatment, an attitude outcome
is likely to be of interest, which concerns the learners'
attitude toward the use of the learned skill (i.e., Briggs-
Wager's (1981) twin objectives, Bloom's (1976) affective
outcomes). The attitude would be claimed to be formed by
skill acquisition as well as motivational treatment that
stresseis the usefulness of the skill as shown in Figure 6-a.
A regression analysis on this model, however, does not
address the fact that the skill acquisition is affected by
the motivational treatment. The actual relationships among
the three variable should be represented as shown in Figure
6-b, which cannot be analyzed using multiple regression per
Path Analysis for Causal Modeling
As an extension of the multiple regression analysis,
path analysis (Asher, 1983; Wolfle, 1980) has been developed
for causal modeling. By employing path analysis technique, a
factor may be perceived as both dependent and independent
variable simultaneously. For example, motivational treatment
may have an effect on the amount of skill acquisition, which
in turn may affect the extent of positive attitude toward
the use of the skill acquired. Furthermore, the motivational
treatment may inform the learners of how useful the to-be-
learned skill would be. If this is the case, then the
motivation level in itself may have a direct effect on the
post-instructional attitude toward skill uses, regardless of
the extent of actual skill acquisition (Figure 6-b). It can
be seen that the extent of skill acquisition is treated as
both dependent and independent variables, for which the path
analytic technique is most appropriate to be employed.
The implied correlation coefficients by the path model
may now be presented to clarify the difference between the
multiple regression model (Figure 6-a) and the path analysis
model (Figure 6-b). Since there are only direct effects in
the multiple regression model, the implied correlation
coefficients are the standard regression coefficients, which
are obtained by regressing the attitude on skill acquisition
and motivational treatment,.plus unanalyzed bivariate
correlation between the two independent variables. In
contrast, in the path model, indirect effects are taken into
consideration, which are added to the direct effects to
derive the implied correlation coefficients. Therefore, the
relationship between the motivational treatment and skill
acquisition is implied by thepath model, but not by the
regression model. The structural equations for multiple
regression runs and implied correlation coefficients for
both models are as follows:
Regression model Path model
X3 = P31 X1 + P32 X2 + R X3 = P31 X1 + P32 X2 + R
X2 : P21 X1 +R
Implied Correlation Coefficients
r13 = P31 + r12 X P32 r13 = P31 + P21 X P32
r32 = P32 + r12 X P31 r32 : P32 + P21 X P31
r21 = P21
Note: X1: Skill acquisition, X2: motivational treatment,
X3: attitude toward skill usage
p: standardized regression coefficients
r: implied correlation coefficients.
By employing path analytic technique, three types of
causal effects may be assessed (Wolfle, 1980). First,
direct effects are indicated by the unidirectional arrows.
For example, m.otivational treatment has a direct effect on
skill acquisition in Figure 6-b. Second, indirect effects
are the extent to which intervening variables account for
the relationships between predetermined and subsequent
variables. An example from Figure 6-b is the effect of
motivational treatment on the attitude toward uses of the
skill through the extent of skill acquisition. Finally,
spurious effects are the extent to which antecedent
variables account for relationships between other variables.
The effect of motivational treatment on the relationship
between skill acquisition and the attitude is an example.
Thus, more complex causal flows can be examined by using a
path analytic model than a cumulative use of multiple
Although path analytic techniques were originated in
non-experimental research traditions (e.g., Anderson &
Evans, 1974; Werts & Linn, 1970,. Wolfle, 1980), applications
have been seen in the experimental paradigm. For example,
Covington and omelich (1984) have utilized path model to
determine the effects of retesting opportunities and
criterion- versus norm-referenced grading on end-of-semester
achievement, taking students' motivational reactions into
account. One of the hypothesized causal links was supported
by the path analysis, including the effects of retesting on
the end-of-semester achievement directly and indirectly
through motivational reactions. Thus, they concluded that
“the performance superiority of a task-mastery system arises
not only because of its inherent instructional properties
(retesting), which account for some 62.7% of the explained
variation in tend-of-semester] performance, but also because
of an enhanced motivational climate (14.0%)...” (p. 1046).
In summary, path analysis would seem to be a powerful
tool in interdomain interaction research, which involves
complex flows of interrelationships among objectives in
various domains of learning outcomes. Not only the direct
effects, but also indirect and spurious effects can be
included in the causal inference using path analysis.
This study was based on the conceptual framework of
interdomain interaction research proposed by Briggs and
wager (1981). Instructional Curriculum Maps (ICMs), which
are to be referred in the design of instructional events,
represent a way to identify possible interdomain ,
relationships. By integrating various domains of learning
outcomes, it is the goal of interdomain interaction research
to provide instructional strategies that facilitate learning
to its fullest extent.
Intellectual skills have been advocated as the building
blocks of school learning (Gagne & Briggs, 1979). This was
also the basic assumption of this study. Intellectual
skills have some advantages as the building blocks of school
learning. First, they Cannot be simply looked up as verbal
information, and they are “typically learned over relatively
short time period” (Gagne & Briggs, p. 13). The
hierarchical structure within the domain of intellectual
skills is also advantageous in forming school curricula. It
is easier, compared to other learning outcomes, to determine
if an intellectual skill has been learned. The structure of
the intellectual skill domain has been clarified by the work
of Gagne (1985), White and Gagne (1974), and others. It was
a natural next step to integrate the intellectual skills
domain with other domains, extending the notion of utilizing
prerequisite relationships for better design of skill
A further premise of this study Was that the integration
among the domains of learning outcomes might contribute to a
shift from a narrow viewpoint of instructional design to a
longer and broader perspective. Typically, designers
consider the end of lesson objective a "terminal" objective
in the design effort. In contrast, Briggs and his
associates (Briggs & Wager, 1981,. Martin & Briggs, 1986)
have stressed the linkage of lesson objectives to unit,
course, and life-long objectives. In this perspective, the
end of lesson objective is only a “target”, not “terminal”,
leading to other objectives at other levels.
Martin and Briggs (1986) have pointed out that
integration of affective objective to the cognitive
counterpart may require the planning of instruction in a
longer range. Learning cognitive strategies may also take a
long period of instruction, which is possibly incidental, or
indirect, in nature (Derry & Murphy, 1986). Whereas the
unit of design effort may still be a lesson level, a
designer should take into account that the “terminal”
objective of a lesson must serve as an enabling objective
for the next lesson. Thus, it is important that the learner
leave an instructional unit with not only a newly learned
skill, but also a positive attitude toward what has been
In order to conceptualize a unit of instruction in the
flow of school learning, Bloom (1976) has offered a process
model of school learning as shown in Figure 7. He conceives
a unit of instruction as having such student characteristics
as input variables, learning task and quality of instruction
as process variables, and learning outcomes as output
variables. It is parallel to the framework of interdomain
interaction in that Bloom has identified both cognitive
entry behaviors and affective entry characteristics as the
input, as well as level and type of cognitive achievement
and affective outcomes as the output of an instructional
Bloom (1976) has also recommended investigating the
causal relationships among the input, process, and output
variables so that a more precise prediction of the output
variables can be made. He has stated:
“The theory attempts to make explicit the ways in which
the learners' previous characteristics and the quality
of instruction determine the outcomes of the learning
process whether the outcomes be different levels of
learning, different rates of learning, or such
affective outcomes of learning as attitudes toward the
learning or attitudes about the self. The causal
system described by these variables is amenable to
modification at a number of points with consequent
changes in the outcomes of learning” (p. 202 - 203).
Thus, in terms of the research methodology, path
analytic procedures were used to examine the flow of causal
relationships. path analytic procedures were employed to
provide an empirical support of Bloom's model of school
learning at the general level, as well as to validate the
ICMs derived from the theoretical assumptions and past
research. Where complex interdomain relationships are
expected, it should be advantageous to use path analytic
procedures to examine the model and the ICMs as a whole.
In this study, the path analytic procedures were
combined with an experimental manipulation of the
''relevance'' of a skill to be learned. This experimental
manipulation enabled the researcher to construct another
level of interpretation of the results, namely, a
traditional experimental paradigm. Should the data fail to
support the hypothesized causal model as a whole, the data
could still be interpreted in terms of direct effects of the
“relevance” treatment, as in a traditional two group
analysis of covariance. Considering the two levels of
interpretation of the results, hypotheses were derived as
This study explored interdomain relationships among the
learning of an intellectual skill, information of its
potential usefulness, and attitudes toward learning the
skill and toward its practical uses. Following the
traditional experimental paradigm, the purpose of the study
was to examine the effects of “relevant” context of the
learning of a skill on motivation, skill acquisition, and
attitudes toward the use of the learned skill. It was also
the purpose of this study to examine a hypothesized path
model that depicted relationships among supporting
instructional objectives from various domains of learning
outcomes. Among the relationships included in the path
model, the effect of success in skill acquisition on the
formation of positive attitude toward practical use of the
skill was of special interest in this study.
The Experimental Manipulation and its Hypothesized Effects
Two levels of the to-be-learned skill--Relevant and
Irrelevant--were introduced in this study. This was done by
adapting the context of the instruction (Ross, 1983) to the
learners' prospective profession (classroom teachers) and to
an irrelevant profession (business managers). In each
version of the instruction, two concrete examples of the
skill's application were described to provide a modeling
condition. The usefulness of the skill and potential
benefits of learning the skill were discussed as a
preinstructional introduction. Further, each version used
practical examples from typical situations in classroom
teaching (Relevant treatment) and business training
(Irrelevant treatment) when the learning of the skill took
place. Therefore, the Education version was expected to be
more meaningful and attractive, thus more “relevant,” to the
participants of the study (preservice teachers) than the
It was hypothesized that the Education version would be
more effective than the Business version in facilitating
motivation to learn the model, acquisition of the skill, and
formation of an attitude toward the use of the skill. In
Ross's study (1983), where the emphasis was to adapt the
context to student background knowledge, the adaptive
context (classroom) was more effective than either a non-
adaptive (nursing) or an abstract (numerical) context in
learning of probability rules by preservice teacher
education students. Schema theory (Rumelhart, 1980) and
research in advance organizers (e.g., Mayer, 1979) have also
indicated the effects of a more meaningful context on
cognitive learning. Increasing relevance of instruction has
alsobeenclaimed tobe one of the most effective
motivational strategies (Keller, 1983b) for aiding both
cognitive and affective learning. With respect to the
effects on attitude learning, knowing “the situations in
which choices of action are likely to be made” would
facilitate the learning of an attitude (Gagne, 1985, p.
237). Furthermore, the perceived usefulness of the skill by
the learner may provide expectation of reinforcement
available for actually utilizing the skill (Briggs, & Wager,
The Hypothesized Causal Model
In order to make a causal interpretation of the
results theoretically defensible, a measure of self-concept
of ability was included as an additional predictor that
presumably has a direct effect on some of the dependent
variables of interest in the study. The importance of
causal inference in research of affective learning has been
pointed out (e.g., Bloom, 1976) so that the process by which
an attitude is developed and the way in which it influences
learning may be explored. Self-concept of ability has been
claimed by many researchers (e.g., Bloom, 1976; Byrne, 1984;
Hansford & Hattie, 1982; Uguroglu & Walberg, 1979) to have
the strongest predictability among measures of affective
entry characteristics of learners.
The interrelationships among the variables of interest
are shown in Figure 8 as a hypothesized path model. Causal
hypotheses are indicated by the unidirectional arrows
showing the various pathways of influence to be tested.
Using path analysis allows for the relative contribution of
all of the determining factors as specified in the model
(direct, indirect, and spurious) to be estimated with regard
to variations in both cognitive and affective outcomes.
Although path analytical techniques were originated in non-
experimental settings (Asher, 1983; Wolfle, 1980),
applications in experimental research have been increasingly
reported (e.g. , Covington & Omelich, 1984).
Two causal networks were of interest in this study.
The first network concerned the various pathways of
influence of the "relevance" of instruction, which was
manipulated in the study. The Education version was
expected to influence motivation to learn the skill (i.e., an
attitude toward learning of the skill), which in turn Was
likely to have an effect on skill acquisition. The use of
relevant context was also assumed to affect skill
acquisition directly, which would in turn aid formation of a
positive attitude toward the skill's practical uses. By
providing information about the contexts for which the skill
could be applied, the relevant version was expected to have
a direct effect on attitude formation concerning its
The second network stemmed from self-concept of
ability. Representing the general expectancy for success in
college learning, self-concept bf ability was expected to
have a direct effect on learner motivation for learning the
skill. Moreover, because self-concept of ability had been
built upon a learner's perception of the academic
performance relative to his or her peers (Bloom, 1976), it
was likely to have adirect influence on the skill
acquisition. Because practical use of the skill would be
out of the college context, no direct effect of self-concept
of ability on the attitude toward practical uses was
expected; the effect was thought to be only indirect through
motivation for learning and actual skill acquisition.
Among the bivariate relationships in the hypothesized
path model, a special attention was given to the effect of
the learner's initial success in skill performance on the
formation of positive attitude toward practical uses of the
skill. The more skillful a learner becomes at using the
newly-presented technique, the more likely he or she is to
develop a positive attitude toward its uses. This may be
because the attainment of the skill is, in itself,
reinforcing (Briggs & Wager, 1981), and because a skillful
learner who feels comfortable using the skill may exhibit
stronger intention of actual uses (McClelland, 1985).
The specification of the causal model was defensible
using Bloom's (1976) process model of school learning.
Bloom has identified two input variables having direct
effects on the process of school learning: cognitive entry
skill and affective entry characteristics. It was assumed
in this study that cognitive entry behavior does not have a
direct effect on either motivation to learn or acquisition
of the to-be-learned skill because the instruction assumed
no specific prerequisite skill and because the to-be-learned
skill was new to the learners. Affective entry
characteristics were represented by the self-concept of
ability, Which has been claimed to be the most powerful
predictor among the affective measures of the school
learning. Although general cognitive capability of the
learner (e.g., that measured by IQ and other demographic
factors) has also been identified to have a direct effect to
the effectiveness of instruction (Dick & Carey, 1985), this
was assumed to be represented by the self-concept of
ability. Thus, it was claimed that important identifiable
causes of the dependent variables of interest in this study
were included in the hypothesized path model.
In order to avoid the bias in model specification, all
variables having direct effects must be included in the
model. If any direct cause were left out, the hypothesized
model would be questioned for its specification, and any
causal interpretation using the model would be invalid.
Three direct causes that were identified, but that were
excluded were: (1) teaching experiences in the public
education settings, (2) perceived likelihood of becoming a
classroom teacher, and (3) knowledge about motivational
design of instruction.
These variables were excluded based on the assumption
that they were constant throughout the population of this
study. That is, it was assumed that the target audience had
little experience in teaching, and that they intend to
become a teacher (i.e., pre-service teachers). The
instructional treatment consisted of CAI courseware
containing an introductory lesson, which required no prior
knowledge about the topic, motivational design of
instruction. Although these three factors presumably had
direct effects on the variables in the model, they did not
have tobe inthe model if they were constant across the
members of the population. To ensure the consistency, data
were obtained in the proposed study so that the participants
were checked against these criteria to be included in the
subsequent analyses. This enabled the causal interpretation
of the results using the hypothesized path model.
One hundred and twenty-two pre-Service teacher
education students were recruited from two sections of an
educational psychology Course at the Florida State
University. The students were awarded extra credits toward
their grades for their participation in the experiment.
Sixteen students were excluded because they indicated that
they do not intend to become classroom teachers. Also
excluded were two students assigned to the business group
(irrelevant treatment) who had business backgrounds. Two
other students were excluded due to insufficient data. A
total of 102 participants remained in the analyses.
Statistical power of this study was calculated to be .72 for
the preset alpha = .05 and a medium effect size (eta
square -- .06, Cohen, 1969, using the table of two group
For the results of this study to be generalized to a
wider population than the pre-service teacher-education
students at The Florida State University, Some biographical
data of the participants were described as follows. Among
the 102 participants analyzed in this study, 82 were females
and 20 were males. Fourteen were majoring in special
education, seven in early childhood education, 47 in
elementary education, 29 in secondary education, and five ln
other categories such as physical education. Most of the
participants had either no teaching experience (49) or
limited experience through a pre-internship (42). Majority
of the participants were naive in studying via a computer,
due to their limited exposure to this type of learning
The materials for this study were an introductory
lesson on the ARCS model for motivational design (Keller,
1983a; Keller & Dodge, 1982; Keller & Kopp, in press; Keller
& Suzuki, in press). The instructional goal of the lesson
was to be able to classify motivational strategies used in
instructional plans into one of the four categories of the
ARCS model: Attention, Relevance, Confidence, and
Satisfaction. The lesson consisted of two types of
materials: (1) A two page introduction of the ARCS model,
which was included in a booklet, and (2) A computer-assisted
instruction (CAI) courseware as a main body of instruction
Which had 41 instructional screens.
Two versions of the lesson were systematically
developed (Briggs & Wager, 1981) and formatively evaluated
(Dick & Carey, 1985) with two different target audiences in
mind: (1) preservice teacher education students (Education
version), and (2) undergraduate business major students
(Business version). Although the target objective and the
basic structure of the module were identical in both
versions, the two versions were different in (1) the
preinstructional introductory “positioning” of the ARCS
model, which was included in the two page introduction, and
(2) the examples and practice items, which were used in the
Each version was designed so that the ARCS model maybe
perceived as “relevant” by each target audience. That is,
business students would find the Business version relevant,
and the education major students would find the Education
version relevant. The version that was designed to be I
relevant for business students was assumed to be irrelevant
for the education students. Specifically, human modeling
technique (Bandura, 1977; Gagne, 1985) was employed in the
two-page introduction of the ARCS model, using classroom
teachers as examples in the Education version, and business
people as examples in the Business version. Two anecdotes
of successful application of the ARCS model in on-the-job
situations were used to encourage the target audience to
learn the ARCS model. The two versions of the introduction
are shown in Appendix A, with other pre-instructional
assessment tools used in this study.
The main body of the lesson was CAI courseware, using
a PLATO instructional computer system. After a four-page
introduction to the courseware, a menu screen was presented,
from which a learner selected five instructional sections
and two assessment sections. Each of the instructional
sections consisted of five to seven screens, explaining
motivational problems and the four ARCS categories. In
order to explain the ARCS four categories, examples and
practice items were drawn from typical situations in
business settings for the Business (Irrelevant) version, and
in classroom settings for the Education (Relevant) version.
Some sample screens are presented in Appendix B.
The following measures were employed in this study to
operationalize the hypothetical path diagram shown in Figure
8. Two preliminary Collections of data (N = 27 and 16) were
conducted in order to examine the measurement devices. The
results of the preliminary data collection are also
presented in this section.
Academic Self-Concept Scale
Self-concept of ability Was measured using the Academic
Self-Concept Scale (ASCS) developed by Reynolds, Ramirez,
Magrina, and Allen (1980). The ASCS consisted of 40 opinion
statements concerning college-related personal expectancy of
academic success, to each of which the participant selected
either strongly disagree, disagree, agree, or strongly
agree. Reliability of the ASCS has been reported to be
alpha = .91 inone study (Reynolds et al., 1980), and the
construct validity of the scale has also been supported
(Reynolds, 1981; Reynolds et al., 1980). The preliminary
data collection has also supported the reliability of the
ASCS to be high (alpha = .91 in both pilot data
collections). The ASCS scale is shown in Appendix A as a
part of the pre-instructional booklet.
Criterion-Referenced Measure of Skill Acquisition
Acquisition of a defined concept was measured by a
criterion-referenced posttest. The objective was that the
students would be able to classify motivational strategies
into the ARCS four categories. Four different sets of
fifteen true-false items, one set for each of the ARCS
categories, were given as the posttest as the first
assessment section of the CAI courseware. Each set
contained items indicating a true or false representation of
a motivational strategy. Appendix C consists of the 60
posttest items used in this study.
Adifferent form of the posttestwas used in a pilot
study, which turned out to have low reliability. The test
asked the learner to classify sentence-long statements of
motivational strategies into the ARCS's four categories.
Reliability coefficients Were -.111 for a difficult version
and .372 for an easy version. There was a ceiling effect to
which the low reliability is attributed. Thus, this old
assessment scale was replaced with a new one as described
above. Item specification procedures (popham, 1980) were
employed to ensure the content validity of the test in terms
of the congruency to the instructional objective of the
module. The new test was formatively evaluated in the
second pilot study, in which items with less than fifty
percent correct responses were examined for their clarity
and were rewritten accordingly. A high reliability was
obtained in the second pilot study (alpha = .93, standard
error of measurement = 3.2 with 16 participants).
Three types of attitude measures were employed to test
the participants attitudes toward learning the ARCS model
and toward its practical uses: (1) Continuing Motivation
Scale, (2) Behavioral Intention Differential Scale, and (3)
Semantic Differential Scale.
Maehr's (1976) scale of Continuing Motivation (CM) was
used to measure the attitude toward learning the ARCS model.
The CM scale consisted of “three sequentially and
hierarchically ordered items” (Maehr, 1976, p. 449) that
asked willingness of the respondent “to sign up to work on
the task at some point in the future” (p. 449). The CM was
included in the pre-instructional booklet (Appendix A).
The first preliminary data collection showed high
correlations of the CM scale withboth the SDS and the BID
scale (r = .501, p < .01 and r= .535, p < .01,
respectively), indicating the CM scale may be valid when
employed as a measure of attitude outcomes. Further, those
who read the Educationversion of the ARCS model (Mean =
5.5, SD = 0.9) scored significantly higher on the CM scale
than those who read the Business version (Mean = 4.2, SD =
1.3), F (1,23) = 6.753, p< .05, eta square = .227).
Behavioral Intention Differential Scale (BID, Triandis,
1964) was employed to measure the participants' attitude
toward practical uses of the ARCS model in terms of their
intended uses as classroom teachers. The BID consisted of
20 four-point scales describing intent to apply the ARCS
model in hypothetical school situations. Each of the items
measures the extent to which the participant agrees or
disagrees with the intent of such a personal action. The
BID was administered in the second assessment section of the
CAI coursevare. The 20 items can be found in Appendix D.
The first preliminary data collection revealed that the
initial form of the BID scale, which consisted of 10 items,
had a low reliability Coefficient (alpha = .483). Further,
the BID scale did not relate to the extent of skill
acquisition, or to the relevance of instruction.
Nonetheless, the BID scale as used in the first preliminary
data collection was highly Correlated with the SDS (r =
・517, p < .01). Thus, the use of BID was decided to be
valid as a measure of attitude toward the use of ARCS model.
The BID scale was lengthened to consist of 20 items in the
second pilot study, where the reliability increased to be
alpha = .92.
A Semantic Differential Scale (SDS; Osgood, 1965) was
developed to measure the participants' feelings toward the
ARCS model. The SDS intends to measure the meaning that
people give to a concept. Thus, it was considered to be an
indirect measure of the participants' attitude toward any
action involving the ARCS model in this study. The SDS was
used to supportwhat theCM scale and the BID scale were
intended to measure. The SDS consisted of 18 bi-polar
adjectives with a seven-step scale in between each pair of
the adjectives, which can be found in the Appendix A.
Reliability Coefficient of the SDS was .87 in the first
preliminary data collection, and was .82 and .90 when
administered in the second pilot study, before and after the
participants went through the CAI courseware, respectively.
In the first pilot study, it was also found that the
participants' attitude measured by the SDS was high when
they SCored high on the skill acquisition test (r = .45,
p < .01).
Two sessions of a third and fourth year undergraduate
Educational psychology class were asked for volunteers to
participate in this study. The students made individual
appointments With the experimentor for a block of two hours
to come to a learning resource center. There were usually
five to ten participants in each block of the time. Upon
arrival, each participant was given a booklet, which
included the demographical data collection sheet, the ASCS
measure, one of the two versions of the introductory
statementof the ARCS model, the SDS measure, and the
Continuing Motivation (CM) measure. The assignment of the
two versions of the introduction was made at random with the
restriction that the two groups would have equal number of
participants. The SDS and the CM measures assessed the
affective entry characteristics prior to the intellectual
skill instruction, but after the introduction of the skill
in order to manipulate the relevance generating property.
When each participant finished with the booklet, he or
she was introduced to the next activity, which was given
regardless of his/her intention to study the ARCS model as
expressed in the CM measure. The participants were directed
to study the ARCS model using computer courseware. Based on
the version of the introduction to the ARCS model, which was
just shown in the booklet, each participant was assigned to
either the Eduction or Business version of a CAI module to
study the four categories of the ARCS model. An on-line
posttest was administered when each participant reached the
end of the CAI module at his or her own pace. The posttest
was followed by a screen that showed how well the
participants did on the skill test. Finally, the SDS
measure and the BID measure concerning attitudes toward the
ARCS model and toward its practical uses were administered
as the last part of the computer courseware.
Two kinds of statistical techniques were employed in
order to answer the two types of research questions in this
Analyses for the Experimental Manipulation
Analyses of covariance were used to examine the effects
of the experimental manipulation, the relevance group
assignment, controlling for the initial difference among the
participants as expressed by academic self-concept. Three
separate analyses of covariance were employed to test the
group differences on: (1) Attitudes toward learning of the
skill as measured by the Continuing Motivation Scale, (2)
Acquisition of the skill as measured by the posttest, and
(3) Attitudes toward practical use of the skill as measured
by the Behavioral Intentional Differential Scale. The score
on the Academic Self-Concept Scale was used as a covariate
in these analyses.
Analyses for the Path Model
This study employed path analysis techniques to examine
the hypothesized model of causal inference shown in Figure
8. The hypothesized model was a recursive model (Asher,
1983), in which the causal flow of influences between
variables was unidirectional. Multiple regression runs were
combined to confirm the hypothesized path model and to
deterrnine individual effects of each of the determinants.
Since the path model employed in this study combines
observed data (e.g., self-concept of ability) and
experimental manipulation (i.e., relevance of instruction),
results obtained by analyses of covariance using the
experimental manipulation would still be valid even if the
path model were not confirmed.
The approach for path analysis in this study was a
mixed approach of confirmation and exploration, due to the
limited availability of the theoretical support for the
hypothesized model. Although the theoretical rationale for
the model specification has been presented in the previous
chapters of this paper, support from past research was
limited to individual relationships among the adjacent
variables. No empirical support could be identified for
making adequate rationale for constructing the proposed
model. Thus, when the proposed model was not confirmed by
the collected data, alternative models, based on the
empirical data, were explored.
The overall procedure of the data analysis used was to:
Step 1: Calculate observed bivariate correlation
coefficients among all variables in the model.
Step 2: Run multiple regression for each of endogenous
variables in the model using all direct causes,
which resulted in the structural equations.
Step 3: Calculate the implied correlations by the model,
using correlation decompositions.
Step 4: Determine the model fit by comparing the implied
and observed correlations, by examining
empirical support for each of the paths in the
model, and by checking assumptions for the
specified functional form.
Step 5: Modify the original model to look for an
alternative model if the original model was not
confirmed by the data.
This study was designed for two levels of
interpretations of the results as expressed in the two
subcategories of the hypotheses: (1) traditional
experimental paradigm with two levels of the manipulated
variable, relevance of instruction, and (2) causal inference
of the interrelationships among the predictor variables
included in the hypothesized path diagram. Thus, the
results of the study are presented in three subsections in
this chapter: (1) descriptive statistics and verification of
assumptions for inferential statistics, (2) the effects of
the experimental manipulation, and (3) an examination of the
hypothesized path model.
Table 2 shows possible ranges, group and grand means,
standard deviations, and Cronbach's alpha reliability
coefficients of the predictor variables for the sample of
102 included in the subsequent analyses. A check against
scatterplots for normality of the score distribution within
each group has revealed no major violation for any of the
variables included in the table. Further, the F max tests
confirmed that all measures had similar variances in the two
Other assumptions for inferential statistics were also
tested. First, for the subsequent analyses of covariance,
the product term of the Academic Self-concept and the groups
was entered in multiple regression runs to test the
homogeneity of slopes or lack of interaction effect on each
of the dependent measures. It was found that no covariate
by factor interaction was significant (F (1,98) : 0.00,
1.14, 0.00, 0.02, 0.12 on CM, 1st SDS, posttest, BID, and
2nd SDS, respectively). Therefore, the Academic Self-concept could be used as a covariate. For multiple
regression runs, which were used for the path analysis, the
assumptions of a normal distribution and homoscedasticity of
the residuals were examined. A scattergram plot (the
residuals by the predicted values of the dependent measure)
was depicted for each of the two relevance group whenever a
multiple regression analysis was conducted.
The bivariate correlation coefficients among the
predictor variables are shown in Table 3. First, the
bivariate correlation coefficients were calculated for each
of the two relevance groups to examine as to whether any
large discrepancies exist. No major difference was detected
between the group correlation coefficients. Thus, it was
decided to pool the two groups in order to arrive at the
bivariate correlation coefficients shown in Table 3.
The relevance of the instruction, which was manipulated
in this study, was expected to have both affective and
cognitive effects. The learners who studied Education
(relevant) version were hypothesized to score higher on the
Continuing Motivation (CM) Scale, posttest, and Behavioral
Intentional Differential (BID) Scale than those who studied
Business (irrelevant) version. one-way analyses of
covariance were used to control for the initial difference
of affective entry characteristics as measured by ASCS.
As expected, the relevance manipulation in the
introduction had a statistically reliable effect on the
attitude toward studying more about the ARCS model as
measured by the CM scale, when the Academic Self-concept was
used as a covariate, F (1,100) = 4.618, p < .05, eta square
= .04. Although the difference was not as large as Cohen's
(1969) definition of medium effect size (eta square = .06),
the adjusted means of the Education group (5.0) was
approximately four tenths of a standard deviation higher
than the Business group (4.5). (Note that due to the small
group difference on the ASCS measure, the adjusted and
unadjusted means were approximately the same.) The 95%
confidence interval (95% CI) of the adjusted group mean
difference was from 0.04 to 1.02. That is, the group
difference on the CM scale may have been 3% to 78% of the
standard deviation (a less than small effect to large effect
in Cohen's definition). In contrast, such a group
difference was not found tobe reliable on thepre-instructional feeling toward the ARCS model as measured by
the SDS scale, F (1,100) = 0.792, p > .05.
Although the FJducation version was expected to be more
effective than the Business version, the difference between
average performance on the posttest was not statistically
reliable when Academic Self-concept was used as a covariate,
F (1,100) = 0.585, p > .05. Also found to be not
significant was the group difference on the postl
instructional attitudes as measured by BID and SDS scales,
with the Academic self-concept as a covariate.. F (1,100) =
1.12,p >.05 for the BID scale, F (1,100) = 0.35, p > .05
for the SDS scale. Therefore, in this study, the effect of
the relevance group difference was found to be limited to
the pre-instructional, but postlintroductory, motivation for
further learning as measured by the CM scale.
The hypothesized path diagram (Figure 8) was examined
by combining multiple regression analyses, producing path
coefficients shown in Figure 9. The path coefficients were
obtained by multiple regression analyses including only the
predictor variables that were assumed to have direct effects
on the dependent variable in each analysis. All the
hypothesized direct effects are shown as unidirectional
arrows, of which statistically significant paths are
indicated by thicker arrows in the diagram.
In order to determine if the hypothesized path model
fits the data obtained in this study, empirical support for
the correlation among the predictor variables, each model
path, and the functional form (i.e., residuals of the
regression equation) of the relationships among the
variables were examined. It was found that all but one of
the observed correlation coefficients were similar (i.e.,
the difference smaller than .05) to the implied correlation
reproduced by tracing paths of the model. That is, the data
supported all the bivariate relationships between each pair
of the variables in the model except the one between the
Academic Self-concept (ASCS) and Behavioral Intention
Differential Scale (BID). The absence of the direct causal
path from ASCS to BID was speculated to be related with this
discrepancy. The direct and indirect effects in the
hypothesized path model are summarized in Table 4.
The functional form was also examined by looking for
any peculiar pattern in the residual plots. No systematic
pattern was found in the scatterplots of residuals in each
of the regression analyses, supporting the functional form
of the hypothesized path model. In contrast, as indicated
in Figure 9, several of the hypothesized paths of direct
effects were not statistically Significant.
Thus, it was not confirmed that the hypothesized path
model fit the actual data in this study. The hypothesized
path diagjram, as a whole, did not represent the way
attitudes toward the newly learned skill were formed, in a
statistically reliable manner.
The effect of actual success in learning, as expressed
by the performance on the posttest, was of special interest
in this study as it was related to the formationof positive
attitudes associated with the ARCS model. As expected,
bivariate correlation between the posttest performance and
the scores on the SDS was found to be of moderate magnitude
and statistically reliable (r = .205, p < .05). A multiple
regression analysis also revealed that a reliable and
moderate portion of the change in the feeling toward the
ARCS model between the pre and post administrations of the
SDS measure was attributed not to the version of the
material studied, but to the actual degree of success in the
posttestperformance (F (1, 98) : 15.302, p < .001, eta
square : .083). That is, controlling for their groups and
intial ratings of the model, thosewho scoredhighon the
posttest expressed more positive feelings about the ARCS
model than those who scored low.
The magnitude of this effect can be more concretely
expressed by the shift of the SDS score by changes in the
posttest. The SDS score increased 15 points, or 1.2 sigma,
while the posttest score only changed from 35 points (mean
posttest score - 2 sigma) to 58 points (mean + 2 sigma).
This increase of the SDS score by 15 points may be
considered to be a large effect., when compared with Cohen's
(1969) definition of a large effect size (8/10 of a standard
deviation). Ninety-five percent CI for the average increase
of the SDS score Was from 7.4 points to 22.5 points,
representing six tenths and 1.8 times as large as a standard
In contrast, the relationship between the posttest
performance and the score on the BID scale was not found to
be statistically reliable in this study, although they Were
positively related (r = .16, p > .05). In a similar
multiple regression analysis, the effect of the posttest
performance on the attitude shift between pre and post
instruction measures (using the CM and BID scales) was
not found to be statistically reliable, F (1,98) = 2.0, p
Although alternatives to the hypothesized path model
were explored as planned, no temporal path model could be
identified, in a statistically reliable manner, using the
components in the hypothesized model. The direct effect
paths that were found to be nonsignificant were deleted, one
atatime, to identify a model with all direct effect being
statistically Significant. Although a diagram was
identified with every path being significant (Figure 10),
the model suggested implied (or reproduced) correlation far
apart from the observed correlation among the components.
Similarity between correlation reproduced by tracing the
paths of the model and the observed correlation is one of
the necessary conditions for the temporal model if it is to
fit the data. Thus, no alternative model, which fit the
data, could be found without changing the components in the
hypothesized path model.
As the next step, alternative models were explored by
adding some components to the hypothesized model, and
deleting some paths from the model, againoneatatime.
when the post-instructional feeling toward the ARCS model
(i.e., the second administratioh of the SDS) was added, a
temporal model could be identified with a statistically
marginal reliability. Figure 11 shows the alternative path
model, which has all the direct effects being significant
(one path at a marginal probability), and which fits the
data under a more lax criterion for correlation similarity
(i.e., the difference between observed and implied
correlation among the components being .10, instead of .05).
It should be noted that this revised model needs
further investigation for its confirmation. Also,
theoretical adjustment for inclusion and exclusion of the
components, as well as the direct pathways, are in order.
However, this model may be useful to discuss the findings
of this study in the next chapter.
The hypothesized path model as a whole was not
confirmed to represent the complex causal flows of the
formation of attitudes toward practical uses of a newly
learned intellectual skill. Results indicated, however,
that some of the individual hypotheses were supported in
this study. A temporal alternative to the hypothesized path
model was also identified. In this section, the results
will be first discussed in light of the hypotheses: (1) the
experimental manipulation, and (2) the path model. Then,
the limitations of the study, suggestions for future
research, and instructional implications of the study will
The Effects of Relevance
The two versions of the materials were employed in this
study in order to create cognitive, as well as affective,
group differences. These differences were expected to be
due to the difference in “relevance” of the instruction,
which was a function of whether the instruction related to
the specific career orientation of the target audience. For
the pre-service teacher participants in the study who
expressed classroom teaching as the first preference of
their career, the Education version was expected to be more
effective in aiding both cognitive and affective learning
than the Business version.
Looking at the results that indicate such effects of
relevance groups to be limited to the pre-instructional
motivation (as measured by the CM scale), it can be
concluded that the effects of relevance groups were created
only in the introduction via the use of human models. Note
that there were two parts in the relevance manipulation: (1)
different success stories using the human model technique,
and (2) different examples and practice items in the
courseware. The CM measure was administered after the
introduction, but before the courseware. Only after the
administration of the CM were examples and practice items
introduced from different contexts. The difference,
however, did not have any systematic effects on cognitive
performance. In contrast, the initial motivational
difference created by the human modeling technique was
persistent until the end of instruction, which was the
single reliable source of variance in the BID measure.
Why were the effects of relevance groups limited to the
beginning phase of instruction? probably because that was
where the relevance was being established for subsequent
events of instruction. It is consistent with instructional
design models (e.g., Dick & Carey, 198S,. Gagne & Briggs,
1979) suggesting the utility of the target task emphasized
at the beginning of an instructional activity. Once
established at an early stage of learning, perceived
relevance of the new skill may have been difficult to alter
by the subsequent experiences. The expectancy-value theory
(Atkinson, 1964; Keller, 1983b) suggests that the perceived
relevance can be highly independent of the perceived
confidence with the skill.
An alternative explanation is that the Business
Introduction did not avert the participants from further
learning, but merely had a weaker impact on the formation of
positive attitudes toward learning the skill than the
Education Introduction. The concrete success stories in the
Business version may have contributed to the credibility of
the ARCS model, rather than persuading the participants that
the model was irrelevant to them. On the average, the
participants in both groups were willing to learn about the
ARCS model, but the Business group was not sure if they
would do this on their own time, whereas the Education group
was (i.e., Means of 4.5 and 5.0 on the CM scale,
If a strong negative feeling toward learning the ARCS
model was formed by the Business introduction, the
participants in the Business group would have created a
“perceptual screen (Briggs & Wager, 1981)”, which prevented
them from effective learning. Because the participants were
still “pre”-service teachers, the irrelevant treatment group
may have not been bothered by the business orientation of
the materiali. A case of in-service teachers may be quite
different. The in-Service teachers may have already
acquired a belief that business motivation models do not
apply to the school classroom settings.
The absence of a direct effect of the relevance group
on final performance and attitudes may also be attributed to
the relevant content of the learning task. The content
(i.e., about motivation) was carefully chosen so that the
learning task could be potentially perceived to be relevant
for the pre-service teachers. However, it might be so
attractive to the participants that the difference in
relevance groups did not have visible effects. In Ross'
study (1983; Ross & Bush, 1980), where the adapting contexts
was effective, the learning task was basic statistics. The
subject matter was relatively difficult to relate to the
students' background without the contextual help built into
the instruction. In this study, on the other hand, it may
have been relatively easy for the participants to utilize
their existing knowledge, even when the new skill was
presented in an irrelevant context. Recently, Ross (Ross &
Morrison, 1987) made a point that much of theeffects of
adapting the instructional context to student background
could be attributed to the novelty effect, and that the
place to use this strategy should be carefully selected.
The characteristics of the CAI courseware used in this
study may also have weakened the effects of the relevance
group. Although the exarnples and practice items were drawn
from different contexts, both versions had the same, well-
designed, structure. Regardless of initial differences in
motivation, the courseware may have required all the
participants to be "on task" via motivational properties
such as frequent use of questions and the menu driven
structure with short segments (Keller & Suzuki, in press).
An average study time of about an hour may have also been
too short to create any reliable effects of motivational
differences among the participants. High averages and small
standard deviations on the posttest performance indicate
that both versions of the courseware were effective.
Finally, the research setting may have been another
factor. That is, the obtrusive experimental situation,
together with the voluntary Participation, may have demanded
the participants to exert more effort than they Would
normally in a learning situation such as in a class. If the
experiment were unobtrusively conducted as a part of course
activities over a longer period of time as was originally
planned, the demand characteristics in the setting could
have been lower. The participants were explicitly told
that, regardless of the posttest score, they would be given
the same amount of extra credits for the course. It was
observed, however, that most of the participants studied the
material very seriously. The interactive feature of the
courseware was also likely to increase learning from both
In summary, it can be concluded that the unique
combination of the learner, learning task, and situational
characteristics seem to create less than anticipated effects
of the relevance manipulation. Although the human modeling
technique in either context was found to be effective, the
effects of contexts themselves were likely to be different
from situation to situation. one determiner may be the
degree towhich the learningof the skill is felt tobe
urgent and the application of the skill in the specific
context is perceived to be immediate. Intrinsic interest
value of the task may be another factor.
The Causal Model
The hypothesized path diagram was not supported by the
data obtained in this study. The path diagram did not
sufficiently explain the causal flows among the variables
included in the model. The model fit was determined
statistically by the magnitudes of bivariate correlation
between each pairof thevariables, as well as the
specifications of the relationships among the variables.
Thus, the reason why the hypothesized model could not be
confirmed was either (1) the expected bivariate
relationships between pairs of variables were different from
the true relationships, or (2) the specification of the
model was incorrect, or (3) both.
The absence of relevance group effects on both the
posttest and post-instructional attitudes greatly influenced
the causal flows in the model, which can be considered to be
a discrepancy between the expected and actual bivariate
relationships. In order for a path model to be confirmed,
all of the hypothesized direct effects must be statistically
significant. Among the three hypothesized direct effects of
the relevance groups, only one (on the CM scale) was found
to be significant. The unanticipated weak effects of the
relevance group difference was sufficient to disconfirm the
Was the specification of the model incorrect? The
answer to this question seems tobe unknown at this point.
Theoretical argument for each of the bivariate relationships
exists, although the model as a whole may not be defended by
prior research. If the magnitude of the effects of
relevance group differs, the current model may be confirmed
as a whole. The bivariate relationship between the academic
self-concept and the attitude toward learning the ARCS model
was hypothesized, but found to be nonsignificant in this
study. A stronger difference between the relevance groups,
however, may change this relationship, because the attitude
was regressed by both the academic self-concept and the
relevance group in the model. Thus, until different degrees
of the magnitude of the group difference are tested, the
hypothesized path model should not be determined to be
A distinctive difference between experimental and non-
experimental uses of the path model is the manipulation of a
component in the path model when employed in an experimental
setting. In non-experimental situations, the bivariate
relationships between the components of a causal model are
determined for a researcher. Thus, the main concern is to
specify the causal model correctly, by determining which
direct paths to include in the model. On the other hand, in
an experimental setting, a researcher has a certain degree
of control in the effects of the manipulated variable. Even
if a model were confirmed in an experimental setting, the
same model would not necessarily work when the magnitude of
experimental manipulation changes. A special caution is
required when generalizations of the results are made in
terms of the causal inference. Similarly, when a proposed
model fails to be confirmed, additional experiments are in
order before any conclusion can be drawn in relation to the
specification of the model.
The effect of successful performance on the formation
of positive attitudes toward the newly-learned skill and its
applications was of specific interest in the path model,
which was partially supported in this study. On one hand,
the posttest score was the only reliable source of variance
explaining the shift between pre and post administrations of
the SDS measures. On the other hand, however, the direct
effect of the posttest performance was not statistically
reliable when the variation of the BID scores was examined.
That is, whereas the initial success or failure in the
learning of the model had a reliable and moderate effect on
the shift of the participants' impression of the model
itself (eta square = .083), the effect on the attitude
toward its actual use was not found to be reliable.
Instead, the attitude toward model's actual use was strongly
influenced by the initial motivation to learn the model (eta
square = .170).
It seems, at least in this particular case, that the
introductory nature of the learned skill may in part explain
the limited effects of posttest performance on the attitude
formation. The instructional objective of the courseware
used in this study concerned a classification skill, which
is necessary to utilize the ARCS model to solve motivational
problems. The courseware, however, covered only a
prerequisite skill in order to be able to actually use the
model in practical situations. Although successful
performance in the courseware might have perceived as a good
start in learning the model, less successful performance
might also have perceived as "not a bad start". Because the
linkage between the successful learning with the given
material and the successful application was less than
obvious, those who scored relatively low on the posttest
could still say that they would use the model when they
If the learners of the courseware had required a more
immediate need for skills dealing with motivational
problems, the effects of posttest performance on their
behavioral intentions may have been stronger. This may be
the case for in-service teachers. The in-service teachers,
rather than their pre-service counterparts, are more likely
to have urgent demands for a technique such as the ARCS
model to solve their on-going problems. Thus, their
evaluational decisions would likely be either negative or
positive, reflecting their subsequent actual behaviors
regarding the uses of the skill. In contrast, for the pre-service teachers, who were still outside of everyday
teaching, all they could do was guess at future utility.
The BID method may be better used in situations where an
administration of the BID scale is followed by immediate
actions rather than describing actions in distant future.
Looking at the alternative to the hypothesized path
model, some arguments for refining the model seem to be
worth mentioning here. The effects of relevance groups were
found only on the pre-instructional motivation. It can be
temporally Proposed that the direct effects of relevance
manipulation may be limited to the beginning of an
instructional activity. This would mean possible exclusion
of the direct effects of the relevance manipulation on the
posttest performance and post-instructional attitudes as
Figure 11 suggests. The relevance manipulation may still
have an effect on cognitive performance and the end-of-lesson attitudes, but it may be only indirect through the
learner motivation. This is consistent with Keller's
(1983b) position that the effects of perceived relevance and
perceived confidence on performance are only indirect
through motivation as measured by the amount of effort.
The other causal flow of interest stemmed from the
academic self-concept. The academic self-concept has been
claimed to be the most powerful predictor of cognitive
performance among the affective entry Characteristics. This
notion was supported in this study in that the ASCS was the
only reliable predictor when the posttest performance was
regressed. It was, however, found that the academic self-concept was not a reliable predictor for the pre-instruction
motivation as measured by the CM scale.
It is possible that the academic self-concept, which
presumably represents confidence or expectancy aspect of
motivation, may not have a direct impact until experience
with the new skill has been well underway. At the time of
the CM measurement, the participants had read only an
introductory description of the new skill. The solo source
of the motivation at that point may have been how attractive
the new learning task was in terms of perceived relevance.
Thus, it is possible that the effect of confidence aspect of
motivation does not become apparent until the learning with
the new skill has been sufficiently experienced. In the
case this position was taken, the direct effect of the
academic self-concept on the attitude toward learning the
new skill would be deleted from the path model.
In sum, the results regarding the causal flows are, at
this point, inconclusive. Careful consideration should be
given either to reject or accept the hypothesized model
after additional empirical investigation. Although the
alternative model could be identified only under a more
lax statistical criterion, several modifications for the
current causal flows are temporally Suggested with
Limitations of the study and suggestions for Future Research
Conclusions drawn from this study are limited to the
individual relationships between the pairs of variables in
the model, because the hypothesized causal model as a whole
was not confirmed. Some interesting results were, however,
found as to whether the hypothesized individual
relationships may exist. Although it was not as strong as
anticipated, the experimental manipulation did have a
significant effect on the attitude toward learning the
skill. In order to further examine the causal relationships
among the predictor variables included in this study, as
well as other related variables, more research studies are
It is suggested for future research that the current
path model be tested as it now exists, by re-defining the
differences between the two versions of the relevance
manipulation. As discussed earlier in this section, the
magnitude of the effects that the experimental manipulation
has determines the causal flows in the path model to a high
degree. Before the model is discarded, the experimental
manipulation should be altered so that the relevant
treatment can be perceived as something positive, and that
the irrelevant treatment can be perceived as something
negative. When such a difference is observed and the
current model fails to be confirmed, then modifications of
the model may be explored, taking the suggested refinement
into account (note the previous discussion under the path
model in this chapter).
In order to strengthen the manipulation in the study,
replication of the present study with different target
audiences is suggested. Ross (1983) replicated his study of
adapting contexts of statistics so that his treatments could
be couterbalanced. Employing nursing and education
versions, Ross used teacher education students in the first
study, then he used nursing. students in his second study
when he found opposite results. The same technique can be
employed to this study if both Business and Education
versions are tested using business major students in the
subsequent study. Because the view of business students
toward education version may be different from that of
teacher education students toward business version,
different attitudes may be formed, which are likely to
result in a different flow of effects.
Another possible target audience is in-service, instead
of pre-service, teachers. As discussed earlier in this
section, in-service teachers are more likely to have strong
attitudes toward a new technique or model than pre-service
teachers, since their judgement are based on their own
experiences in educational settings. The in-service
teachers probably have a stronger identity as teachers than
the pre-service counterparts, which may influence the
acceptability of business oriented human models. A stronger
difference might be expected between the two versions,
leading to the effects hypothesized in the model.
Another modification for a future study is the content
of the learning task. Any model dealing with motivational
problems may be perceived to be useful by the pre-service
teachers, because of their anticipation of motivational
problems in school classrooms. Taking Ross' (Ross &
Morrison, 1987) comment into account, the hypothesized path
model may be re-tested with a less salient subject matter.
If the learners have more difficulty relating the learning
task to their interest or prior knowledge, the relevance
manipulation may have a stronger effect.
It may be beneficial to break down the relevance
manipulation in order to seek a way to make the difference
more powerful. In this study, relevance manipulation was
combined with the use of human models to create the largest
possible difference between the two levels, based on the
assumption that the human models in the irrelevant context
would avert participants' interests to the material. The
study Was further designed to determine which part of the
relevance manipulation (either the introduction, the
courseware, or both) had an effect on the posttest
performance and post-instructional attitudes. Since this
study Showed that only the introduction, which contained the
human modeling technique, had an effect on the pre-instructional motivation, the effects of the introduction
may be explored further. One possibility is a 2 x 2 design
with the versions (relevant vs. irrelevant) of the material
as one factor and the introduction (presence vs. absence, or
human models vs. no human models) as the other factor.
Using the same target audience, it may be interesting
to examine the effects of the irrelevant version without the
human modeling technique in the introduction, or without the
introduction at all. In such a case, the irrelevant version
would be less likely to increase the credibility of the new
skill, Which seemed to be the problem in the present study.
It should be noted, however, that the counterbalancing
technique, using business students as well as education
students, cannot be employed if the human modeling
technique, or the introduction itself, is excluded from one
version. The difference of the study time would also be a
confounding factor in such design. Nonetheless, altering
the relevance manipulation is needed, in one way or another,
to test the adequacy of the hypothesized path model.
The measurement of attitudes has been the area lacking
adequate development of research methodology and
operationalization. The findings of this study are as Valid
as the measurement instruments employed in this study.
Based on Gagne's (1985) operational definition, personal
choices of action, as represented by the CM and BID scales,
were used to measure the attitudes toward the learning and
practical application of the ARCS model in this study.
Further, feelings toward the ARCS model, as expressed by the
SDS scale, was employed as a supplement to the behavioral
indications of the attitudes. Although the attitude scales
were carefully created based on the currently available
resources, and although the reliability coefficients were
high, the attitude scores may have not represented the
internal status of attitudes adequately.
Although it was confirmed that the three attitude
measures (i.e., the CM scale, BID, and SDS) were highly
correlated to each other, it can be speculated that the
SDS may represent somewhat different aspects of an attitude
From either the CM or BID, because each was related
differently to other variables. For example, the posttest
performance explained a moderate portion of the shift
between the pre and post SDS measures, whereas such an
effect was not found on the shift of the attitudes as
measured by the CM and the BID scales. Further research is
called for in order to examine the validity issue of each of
the attitude instruments as it is related to the operational
definition of an attitude (e.g., Gagne, 1985). The validity
of the BID scale, when employed outside of the context of
actual behaviors, also remains to be an issue for further
An additional limitation may be the nature of the
skills posttest. The original format required the student
to classify the statement given in to one of the four
categories of the ARCS model. Because it is possible, out
of context, that a particular statement may be correctly
classified into more than one category, a forced choice test
(yes or no) was developed where there would be only one
correct answer. However, in doing this, it is possible that
categorizations may have been made too obvious, and that
student may have performed as well regardless of the
instruction. This possibility should certainly be
investigated if further research is done with these
Instructional Implications of the Study
Although this study failed to confirm the hypothesized
path model that describes how the attitudes are formed in
relation to the learning of an intellectual skill, several
findings may be of interest for instructional designers.
The findings include (1) the effects of the human modeling
technique, (2) roles of the entry affective characteristics
in learning, and (3) importance of motivational properties
of the instructional material.
First, the human modeling technique was found to be
effective in the introduction of the new skill. The human
modeling was employed in the form of success stories, using
concrete names in a context relevant to the learners' career
orientation. The success stories included positive
consequences of the skill's application (i.e., vicarious
reinforcement, Bandura, 1977). Although the introduction
consisted only of two pages of text, it was found to be
effective in creating a strong commitment for further
leaning. Therefore, an effective instructional strategy may
be to include a brief example of a successful application of
the to-be-learned skill at the beginning of an instructional
activity. This may help make the instruction more relevant.
Second, the affective entry characteristics (Bloom,
1976) played an important role in cognitive learning. The
academic self-concept was found to be related to the
posttest performance. The more confident a learner was
about him or herself, the more likely he or she performed
better on the posttest. On the other hand, the academic
self-concept did not affect the pre-instructional
motivation, probably because the introduced model was new to
the learners. Although the self-concept itself was likely
to be beyond an instructional designer's control, it may
affect the learning substantially. Therefore, it is
recommended to take the affective entry characteristics of
the target audience into account when one designs an
instructional activity. The extent and nature of past
experiences related to the target task may be the
determining factors of the affective entry characteristics,
which includes the confidence aspect of learner motivation.
Finally, the use of the Instructional Curriculum Maps
(ICMs, Briggs & Wager, 1981) was an effective way to
structure the hypothesized path diagram, which presumably
represents effective flow of instructional activities.
The ICMs can be employed in finding a way that the affective
components of a lesson influence the target objective of the
lesson, when it belongs to the intellectual skill domain.
It is recommended the ICMs and the notion of interdomain
interaction design in general be applied in designing
various instructional activities.
An equally useful tool was the notion of motivational
design of instruction (Keller, 1983b) and the ARCS
strategies as applied to CAI courseware design (Keller &
Suzuki, in press). The extent to which motivational
enhancements are included in a instructional material should
be based on the analysis of both the learners and the
learning task. A systematic process should be followed
making the instruction motivational at the optimal level.
It is especially necessary to address both expectancy (Or
confidence) and value (or relevance) aspects of motivation.
Without increasing either of these two factors. motivation
may not be enhanced.
It does not seem to be an overgeneralization to say
that the affective learning (i.e., motivation and attitudes,
however defined) influenced cognitive learning. The
cognitive learning also influenced the affective learning.
It is the researcher's hope that the important roles that
interdomain interactions play be recognized by instructional
designers and be planned for in instruction. It is also
evident that more research is still called for in this area.
Airasian, P. W., & Bart, W. M. (1973). Ordering theory: A
new and useful measurement model. Educational
Technology, 13(5) , 56 - 60.
Ajzen, I., & Fishbein, M. (1980). Understanding attitudes
and predicting social behavior. Englewood Cliffs, NJ:
Anderson, J. G., & Evans, F. B. (1974). Causal models in
educational research: Recursive models. American
Educational Research Journal, 11, 29 - 39.
Anderson, J. R. (1982). Acquisition of cognitive skill.
Psychological Review, 89, 369 - 406.
Asher, H. B. (1983). Causal modeling (2nd Ed.). Beverly
Hills, CA: Sage Publications.
Atkinson, J. W. (1964). An introduction to motivation. New
York: Van Nostrand.
Ausubel, D. P. (1960). The use of advance organizers in the
learning and retention of meaningful verbal material.
Journal of Educational Psychology, 51, 267 - 272.
Bandura, A. (1977). Social learning theory. Englewood
Cliffs, NJ: Prentice-Hall.
Bart, W. M., & Krus, D. J. (1973). An ordering-theoretic
method to determine hierarchies among items. Educational
and Psychological Measurement, 33, 291 - 300.
Beeson, G. W. (1980). Influence of knowledge context on the
learning of intellectual skills. Manuscript submitted
Bloom, B. S. (1976). Human characteristics and school
learning. New York: McGraw-Hill.
Bransford, J. D. (1984). Schema activation and schema
acquisition: Comments on Richard C. Anderson's remarks.
In R. C. Anderson, J. Osborn, & R. J. Tierney (Eds.),
Learning to read in American schools: Basal readers and
content texts. Hillsdale, NJ: Lawrence Erlbaum
Briggs, L. J. (1982). Instructional design: Present
strengths and limitations, and a view of the future.
Educational Technology, 22(10), 18 - 23.
Briggs, L. J., & Wager, W. W. (1981) Handbook of procedures
for the design of instruction (2nd Ed.). Englewood
Cliffs, NJ: Educational Technology Publications.
Byrne, B. M. (1984). The general/academic self-concept
nomological network: A review of construct validation
research. Review of Educational Research, 52, 123 - 142.
Cohen, J. (1969). Statistical power analysis for the
behavioral sciences. New York: Academic Press.
Covington, M. V., & Omelich, C. L. (1984). Task-oriented
versus competitive learning structures: Motivational and
performance consequences. Journal of Educational
Psychology, 76, 1038 - 1050.
DerrY, S. J., & Murphy, D. A. (1986). Designing systems
that train learning ability: From theory to Practice.
Review of Educational Research, 56, 1 - 39.
Dick, W., & Carey, L. (1985). The systematic design of
instruction (2nd Ed.). Glenview IL: Scott, Forseman and
Gagne, R. M. (1968). Learning hierarchies. Educational
Psychologist, 6, 1 - 9.
Gagne, R. M. (1977). Analysis of objectives. In L.J.
Briggs (Ed.), Instructional design: Principles and
applications. Englewood Cliffs, NJ: Educational
Gagne, R. M. (1985). The conditions of learning and theory
of instruction (4th Ed.). New York: Holt, Rinehart, and
Gagne, R. M., & Briggs, L. J. (1979). Principles of
instructional design (2nd Ed.). New York: Holt,
Rinehart, and Winston.
Hansford, B. C., & Hattie, J. A. (1982). The relationship
between self and achievement/ performance measures.
Review of Educational Research, 52, 123 - 142.
Hurst, B. M. (1980). An integrated approach to the
hierarchical order of cognitive and affective domains.
Journal of Educationa1 Psychology, 72, 293 - 303.
Keller, J. M. (1979). Motivation and instructional design:
A theoretical perspective. Journal of Instructional
Development, 2(4), 26 - 34.
Keller, J. M. (1983a, June). Development and use of the
ARCS model of motivational design. Enschede, The
Netherlands: Vakgroep lnstruktietechnologie, Toegepaste
Onderwijskunde, Technische Hogeschool Twente.
Keller, J. M. (1983b). Motivational design of instruction.
In C. M. Reigeluth (Ed.), Instructional-design theories
and models: An overview of their current status.
Hillsdale, NJ: Lawrence Erlbaum Associates.
Keller, J. M., & Dodge, B. (1982). The ARCS model of
motivational strategies for course designers and
developers. Ft. Monroe, VA: The Army Training
Keller, J. M., & Kopp, T. (in press). Application of the
ARCS model of motivational design. In C. M. Reigeluth
(Ed.), Instructional theories in action: Lessons
illustrating selected theories and models. Hillsdale,
NJ: Lawrence Erlbaum Associates.
Keller, J. M., & Suzuki, K. (in press). Use of the ARCS
motivational model in courseware design. In D. H.
Jonassen (Ed.) , Instructional designs for microcomputer
courseware. Hillsdale, NJ: Lawrence Erlbaum Associates.
Krathwohl, D. R., Bloom, B. S., & Masia, B. B. (1964).
Taxonomy of educational objectives Handbook II affective
domain. New York: David Mckey Company.
Maehr, M. L. (1976). Continuing motivation: An analysis of
a seldom considered educational outcome. Review of
Educational Research, 46, 443 - 462.
Mager, R. F. (1968). Developing attitude toward learning.
Palo Alto, CA: Fearon Publishers.
Martin, B. L., & Briggs, L. J. (1986). The affective and
congnitive domains: Integration for instruction and
research. Englewood Cliffs, NJ: Educational Technology
Mayer, R. E. (1979). Can advance organizers influence
meaningful learning? Review of Educational Research,
49, 371 - 383.
Mayer, R. E., & Bromage, B. K. (1980). Different recall
protocols for technical texts due to advance organizers.
Journal of Educational Psychology, 72, 209 - 225.
McClelland, D. C. (1985). How motives, skills, and values
determine what people do. American Psychologist, 40,
812 - 825.
McGuire, W. J. (1969). Attitudes and attitude change. In
G. Lindzey, & F. Aronson (Eds.),Handbook of social
psychology (Vol. 2). Reading, MA: Addison-Wesley.
Osgood, C. E. (1965). Cross cultural comparability in
attitude research via multilingual semantic
differentials. In I. Steiner, & M. Fishbein (Eds.),
Current studies in social psychology. New York: Holt,
Renehart, & Winston.
Pintrich, P. R., Cross, D. R., Kozma, R. B., & McKeachie, W.
J. (1986). Instructional psychology. Annual Review of
Psychology, 37, 611 - 651.
Popham, M. J. (1980). Domain specification strategies. In
R. A. Berk (Ed.), Criterion-referenced measurement: The
state of the art. Baltimore, MD: The John Hopkins
Reigeluth, C. M. (Ed.). (1983). Instructional-design
theories and models: An overview of their current status.
Hillsdale, NJ: Lawrence Erlbaum Associates.
Reiser, R. M., & Gagne, R. M. (1983). Selecting media for
instruction. Englewood Cliffs, NJ: Educational
Reynolds, W. M. (1981). Measurement of academic self-concept in college students. Paper presented at the
Annual Meeting of the American Educational Research
Association, Los Angels.
Reynolds, W. M., Ramirez, M. P., Magrina, A., & Allen, J. E.
(1980). Initial development and validation of the
academic self-concept scale. Educational and
Psychological Measurement, 40, 1013 - 1016.
Rogers, E. M. (1983). Diffusion of innovations (3rdEd.).
New York: The Free Press.
Rosenberg, M. J., & Hovland, C. I. (1960). Cognitive, affective,
and behavioral components of attitudes. In C. I. Hovland,
& Rosenberg, M. J. (Eds.) , Attitude organazition and
change. New Haven: Yale University Press.
Ross, S. M. (1983). Increasing the meaningfulness of
quantitative material by adapting context to student
background. Journal of Educational Psychology, 75,
519 - 529.
Ross. S. M. (1984). Matching the lesson to the student:
Alternative adaptive designs for individualized learning
systems. Journal of Computer-Based Instruction, 11,
42 - 48.
Ross, S. M., & Bush, A. J. (1980). Effects of abstract and
educationally Oriented learning contexts on achievement
and attitudes of preservice teachers. Journal of
Educational Research, 74, 19 - 22.
Ross, S. M., McCormick, D., Krisak, N., & Anand, P. (1985).
Personalizing context in teaching mathematical concepts:
Teacher-managed and computer-assisted models.
ECTJ, 33, 169 - 178.
Ross, S. M., & Morrison, G. R. (1987, February-March).
Computer-based instructional research: Impliations for
design. Paper presented at the Annual Meeting of the
Association for Educational Communications and
Technology, Atlanta, GA.
Rumelhart, D. E. (1980). Schemata: The building blocks of
cognition. In R. J. Spiro, B. C. Bruce., & W. F. Brewer
(Eds.) , Theoretical issues in reading comprehension.
Hillsdale, NJ: Lawrence Erlbaum Associates.
Rumelhart, D. E., & Norman, D. A. (1978). Accretion,
tuning, and restructuring: Three modes of learning. In
J. W. Cotton, & R. L. Klazky (Eds.), Semantic factors in
cognition. Hillsdale, NJ: Lawrence Erlbaum Associates.
Suzuki, K. (1987, February-March). Schema theory: A basis
for domain integration design. In W. Wager (Chair),
Application of schema theory to instructional design.
Symposium conducted at the Annual Meeting of the
Association for Educational Communication and Technology,
Triandis, H. C. (1964). Exploratory factor analsis of the
behavioral component of social attitudes. Journal of
Abnormal and Social Psychology, 68, 420 - 430.
Uguroglu, M. E., & Walberg, H. J. (1979). Motivation and
achievement: A quantitative synthesis. American
Educational Research Journal, 16, 375 - 389.
Wager, W. (1975). Media selection in the affective domain:
A further interpretation of Dale's cone of experience for
cognitive and affective learning. Educational
Technology, 15(7) , 9 - 13.
Wager, W. (1976, April). Instructional Curriculum Mapping.
Paper presented at the Annual Meeting of the American
Educational Research Association, San Fransisco
(ED 141 264).
Weiner, B. (1985). An attribution theory of achievement
motivation and emotion. Psychological Review, 92,
548 - 573.
Werts, C. E., & Linn, R. L. (1970). Path analysis:
Psychological examples. Psychological Bulletin, 74,
193 - 212.
White, R. T. (1973). Research into learning hierarchies.
Review of Educational Research, 43, 361 - 375.
White, R. T. (1974a). Indexes used in testing the validity
of learning hierarchies. Journal of Research in Science
Teaching, 11, 61 - 66.
White, R. T. (1974b). The validation of a learning
hierarchy. American Educational Research Journal, 11,
121 - 136.
White, R. T., a Gagne, R. M. (1974). Past and future
research on learning hierarchies. Educational
Psychologist, 11, 19 - 28.
White, R. T., & Mayer, R. E. (1980). Understanding
intellectual skills. Instructional Science, 9,
101 - 127.
Wolfle, L. M. (1980). Strategies of path analysis.
American Educational Research Journal, 17, 183 - 209.
Young, J. F. (1986). The effects of three different
sequences of instruction on student performance and
attitudes in a teacher education course. Unpublished
doctoral dissertation. Tallahassee, FL: Florida State