An evaluation of humor as a motivational, cognitive, and affective enhancement to lean feedback and remediation strategies in computer-based instruction

Material Information

An evaluation of humor as a motivational, cognitive, and affective enhancement to lean feedback and remediation strategies in computer-based instruction
Teslow, James L
Place of Publication:
Denver, CO
University of Colorado Denver
Publication Date:
Physical Description:
viii, 238 leaves : illustrations ; 29 cm

Thesis/Dissertation Information

Doctorate ( Doctor of Philosophy)
Degree Grantor:
University of Colorado Denver
Degree Divisions:
School of Education and Human Development, CU Denver
Degree Disciplines:
Administration, Supervision, and Curriculum Development
Committee Chair:
Wilson, Brent G.
Committee Co-Chair:
Grabinger, R. Scott
Committee Members:
Duffield, Judith A.
Goodwin, Laura D.
Nuhfer, Edward


Subjects / Keywords:
Computer-assisted instruction ( lcsh )
Wit and humor in education ( lcsh )
Computer-assisted instruction ( fast )
Wit and humor in education ( fast )
bibliography ( marcgt )
theses ( marcgt )
non-fiction ( marcgt )


Includes bibliographical references (leaves 220-238).
Submitted in partial fulfillment of the requirements for the degree, Doctor of Philosophy, Administration, Supervision, and Curriculum Development, School of Education and Human Development.
Statement of Responsibility:
by James L. Teslow.

Record Information

Source Institution:
|University of Colorado Denver
Holding Location:
|Auraria Library
Rights Management:
All applicable rights reserved by the source institution and holding location.
Resource Identifier:
35327180 ( OCLC )
LD1190.E3 1995d .T47 ( lcc )

Full Text
James L. Teslow
B.S., University of Washington, 1970
M.B.S., University of Colorado, 1982
A dissertation submitted to the
University of Colorado at Denver
in partial fulfillment
of the requirements for the degree of
Doctor of Philosophy
Administration, Supervision, and Curriculum Development
School of Education

1995 by James L. Teslow
All rights reserved.

This dissertation for the Doctor of Philosophy
degree by
James L. Teslow
has been approved
Brent G. Wilson
R. Scott Grabinger
Judith A. Duffield
Laura D. Goodwin
Edward R. Nuhfer
^ / $9 Y'

Teslow, James, L. (Ph.D., Administration, Supervision, and
Curriculum Development)
An Evaluation of Humor as a Motivational, Cognitive, and Affective
Enhancement to Lean Feedback and Remediation Strategies
in Computer-Based Instruction
Dissertation directed by Associate Professor Brent G. Wilson
In this experiment a baseline for future research is established by
isolating humor effects in a simplified computer-based instruction (CBI)
design, referred to as a lean feedback and remediation model
(knowledge-of-correct-response feedback, learner control of review, and
remediation via return to previously-viewed material). Three
treatments included two forms of humor (content-related and
-unrelated) limited to use as a first screen in review loops, and a no-
humor control. The humor was intended as a lure to encourage a
second exposure to instructional content after practice. I hypothesized
that the humor treatments would increase voluntary reviews, enhance
posttest achievement due to increased reviewing, and engender positive
attitudes toward the CBI and content. I also expected achievement for
the related-humor treatment to surpass that of the unrelated-humor
treatment and the control, due to the elaboration provided by a
humorous perspective. Results were mixed for the sample of 92 teacher
education students randomly assigned to the three treatments.

Enroute tracking of review behaviors indicated that after an incorrect
practice response, students opted to review 13% of the time (evenly
spread across treatments), and 10% of the time after a correct response
(primarily by subjects in the related and unrelated humor treatments).
Statistical analysis yielded no significant difference among treatments for
the amount of reviewing that took place. Also, there was no significant
treatment effect on immediate or delayed achievement. There was an
overall significant treatment effect for four attitude measures (CBI fun,
CBI design, content interest, and content relevancy), with the related
humor treatment significantly surpassing the unrelated humor and
control treatments for the CBI fun subscale. The results suggest that
while limited exposure to humor in review loops may promote positive
attitudes toward CBI, its use as a lure is not provocative enough to
consistently affect review behaviors and subsequent performance.
This abstract accurately represents the contenj: of the candidate's
dissertation. I recommend its publicatic
Brent G. Wilson

1. INTRODUCTION..................................1
Statement of the Problem...................2
Purpose of the Study.......................5
Scope of the Study.........................7
Expectations for the Study..............8
Limitations of the Study................9
Implications of the Study.................11
2. REVIEW OF THE LITERATURE.....................13
Interactive CBI...........................13
How Do I Get to Carnegie Hall? .... 14
Feedback and Remediation...............16
Learner Control........................20
Intentional Learning Environments 22
Conceptual Model.......................25
Humor in Education........................27
Humor Theory and Research.................28
The Cognitive Connection...............30
Information Processing..............36
The Affective Connection...............39

3. METHODOLOGY.................................43
Design and Analysis.......................47
Research Questions.....................49
Additional Analyses....................50
Content Validity and Formative
CBI Features...........................55
Feedback Type..........................58
Humor Treatments.......................59
Dependent Variables.......................60
Enroute Data...........................60
4. RESULTS . . ............................68
Descriptive Statistics....................69
Preliminary Analysis......................70
Homogeneity of Subjects................71
Potential Covariates and Factors ... 71

Review Potential......................73
Enroute Data................................77
Research Question 1......................79
Review Potential as a Factor.............80
Achievement Data............................81
Research Question 2......................83
Research Question 3......................84
Research Question 4......................84
Attitude Data...............................84
Attitude Subscales.......................85
Research Question 5......................89
Open Response Analyses...................90
Chi-square Analyses...................90
Qualitative Review....................93
Additional Analyses........................ 101
Review Efficiency....................... 101
Average Review Time..................... 109
Summary.................................... 113

5. DISCUSSION............................. 116
Motivational Outcomes...................... 117
Cognitive Outcomes......................... 121
Affective Outcomes......................... 125
Conclusion................................. 129
A. Goal-orientation and Review Potential
Pre-measures............................... 133
B. Computer-based Instructional Materials . 135
C. Description of Humor Elements . . 164
D. Comparison of Practice and Posttest Items. 177
E. Achievement Posttest................ 184
F. Achievement Delayed Posttest .... 193
G. Attitude Survey..................... 198
H. Open Response Measure............... 201
I. Comparison of Subjects by Class . . . 202
J. Exploratory AnalysisIndividual
Difference Variables Correlated with
Dependent Variables.................. 208
K. Unedited Open-response ItemRelated
Humor Treatment...................... 211
L. Unedited Open-response ItemUnrelated
Humor Treatment...................... 215
REFERENCES....................................... 220

Oops, bad news! That is one way to look at certain forms of
instructional feedbacka notification of failure and a need for review,
redress, or retrial. For a user of computer-based instruction (CBI), this
feedback usually occurs during practice and may entail a range of actions
and treatments aimed at remediation of errorsfrom voluntary review
of previous instructional material to interactions with adaptive expert
systems and artificial intelligence. Since the classic works on
programmed instruction by Pressey (1927), Skinner (1958), and Crowder
(1961) there has been continuing interest in improving the effectiveness
of interactive feedback and remediation. The trend, lately, has been
toward increased CBI design complexity, especially with regard to
analysis and remediation of errors and misconceptions (Hannafin,
Hannafin, & Dalton, 1993). The current study reverses this direction
investigating the use of humor as a tactic for increasing learner
engagement with simplified, or lean, feedback and remediation strategies
characterized by knowledge of results (KOR) or knowledge of correct
response (KCR) and voluntary review of previously-viewed
instructional material.

Statement of the Problem
An important consideration for the CBI designer is the type of
feedback. During practice sequences, interactive CBI can provide
feedback immediately or adaptively with varying degrees of complexity,
ranging from simple "correct" or "wrong" messages to sophisticated
diagnostic treatments (Hannafin et al., 1993; Ross & Morrison, 1993).
However, over 50 studies of various forms of feedback have resulted in
no consistent pattern of results other than the general finding that some
feedback is better than no feedback (Kulhavy & Wager, 1993).
Recently, there have been calls for expanding the function of
feedback from its traditional operant role in the area of practice to a more
general information-processing approach that provides periodic input,
guidance, and coaching to the learner across all elements of the
instruction through use of the intelligent tutoring system (ITS) and other
emerging technologies (Kulhavy & Stock, 1989; Kulhavy & Wager, 1993;
Sales, 1993). Even though a great deal of effort has been invested in ITS
design, and the intelligent computer is clearly on the horizon (Burns &
Parlett, 1991), the "...lack of significant numbers of demonstrable,
practical ITSs does not encourage the idea that the methodology is an
appropriate one for efficient, economic production of courseware" (Tait,
1992, p. 128).
As for remediation treatments after feedback, few CBI systems
provide individually-tailored remedial frames based on student error
patterns. Intelligence in diagnosing error patterns and providing tailored

help is new to CBI (Criswell, 1989). The challenge involved in
determining faulty response patterns is exemplified by Burton's (1982)
instructional system called "Buggy," an intelligent CBI remediation
system designed to diagnose simple error patterns identified in the
responses of children learning elementary subtraction. A great amount
of effort has been invested in this program that deals with approximately
100,000,000 "bugs" (e.g., 0-n=n; n-0=0, etc.) resulting from all possible
combinations of approximately 110 simple error patterns. Because of
such complexities and the computing power required, most advanced
remediation treatments are developed in well-funded research
institutions and artificial intelligence (AI) laboratories (Brown & Van
Lehn, 1980; Burns & Parlett, 1991). There is currently a lack of AI
authoring systems and design tools for the rest of usthe proletarian
CBI designers without large budgets and limited access to main frame or
mini-computers (Dijkstra, Krammer, & Van Merrienboer, 1992).
Recent literature has been supportive of adaptive instructional
modeling and response-sensitive systems based on traditional
apprenticeship tutors that "watch" for opportunities to discuss
underlying concepts and provide post-failure explanations (Gott, Lesgold,
& Kane, in press; Grant, McAvoy, & Keenan, 1982; Lesgold, Ivill-Friel, &
Bonar, 1989; Riesbeck, in press); however, the jury is out on the
effectiveness of these systems due to a lack of research (Rosenberg, 1987;
Ross & Morrison, 1993). Sobering meta-analyses have indicated that we
know relatively little about how to individualize instruction (Kearsley,

Hunter, & Seidel, 1983); and that interactions between feedback and
remediation treatments, learner characteristics, and task demands are
still not understood (Kulhavy & Stock, 1989). Some studies indicate that
complex feedback and remediation provides no significant improvement
over simple confirmation (e.g., KOR) or corrective (e.g., KCR) feedback
without remediation, but requires considerable development and
implementation costs (Merrill, 1987; Spock, 1987). The fact remains that
feedback and remediation complexity involves an increased level of
design and development effort, which many educational enterprises and
CBI developers cannot justify (Sales, 1993).
At the other extreme, some of the least complex and economically
feasible CBI designs are ones in which we ease program control
constraints, allowing learners to take charge of their own learning,
including remediation after feedback. This is consistent with
constructivist calls for acknowledgment of the unique set of
understandings, perspectives, and personal goals that each learner brings
to a learning experience (Duffy & Bednar, 1992). However, it is this
design philosophylearner ownership of his/her own learningthat
often leads to disappointments in the instructional design community.
Specifically, the problem that the current study addressed is that
motivating learners to voluntarily and mindfully use feedback and
remediation opportunities remains a design challenge (Morrison, Ross,
Gopalakrishnan, & Casey, 1995). Unfortunately, studies have shown that
learners often make poor decisions concerning their learning needs

when entrusted with interactive control over pace, content, sequence,
and remediation after feedback (Tennyson & Buttrey, 1980). The current
study addressed this issue by suggesting that one strategy for making lean
feedback and remediation designs more motivationally and cognitively
effective is through the exploitation of a powerful human attribute
humor appreciation.
Purpose of the Study
A recent review of the literature (Teslow, 1995) indicates that the
introduction of humor as an instructional strategy in computer-based
instruction (CBI) has not been studied as much as the effects of humor
use in the classroom and other media, even though general humor
research indicates potential for harnessing the appeal of humor for
affective, attentional, motivational, and information-processing benefit
(Kearsley et al., 1983; Snetsinger & Grabowski, 1993,1994). Thus, one goal
of the current study was to contribute to the understanding of the
relationships between two disparate areas of research: (a) interactive
feedback and remediation strategies in CBI, and (b) humor usage in
traditional educational settings. The closing of this gap in the literature
was suggested previously as one research area among several that
concern the potential use of humor in CBI (Teslow, 1995).
Specifically, the current project was designed to address a
particular feedback and remediation problem studied by Morrison et al.
(1995). Their study of the effects of feedback strategies and incentive

conditions found that even high-incentive subjects selected review
opportunities only 15% of the time after making an error, while
averaging less than four seconds per review. Moreover, they reported
that motivation for review was further reduced by the provision of
multiple trial, or answer until correct (AUC) opportunities, with subjects
reviewing only 12% of the time after an initial error. Morrison and his
colleagues recommended further research on coaching or advising
students on when and how to use feedback, and on making feedback
more attractive and useful to learners.
Following up on the latter recommendation, the purpose of the
current study was to investigate the use of humor in lean feedback and
remediation designs as a motivational attractor, or lure, that encourages
voluntary re-access of previously-viewed information after an error. It
was posited that, once accessed, content-related review screen humor
that provides additional information (e.g., a humorous example or non-
example, an alternate perspective, an anchor in the form of a humorous
real-world situation, a play-on-words using key phrases from the original
text, etc.) would make the feedback-remediation process more cognitively
An outcome mentioned frequently in the literature is that
content-related humor has a positive effect on recall and retention
(Chapman & Crompton, 1978; Hauck & Thomas, 1972; Kaplan & Pascoe,
1977; Vance, 1987); however, recent research by Snetsinger and
Grabowski (1994) did not find this to be the case in their CBI study. Also,

studies show that unrelated humor may interfere with performance and
adversely affect attitudes (Bryant, Brown, Silverberg, & Elliott, 1981;
Erickson, 1987; Hill, 1988). The current study continued the investigation
into this issue of the relatedness of imbedded instructional humor and
the impact on performance and attitude.
Scope of the Study
I studied the effects of three review screen treatments (no humor,
content-related humorous elements, and content-unrelated humorous
elements) on voluntary reviews (i.e., total number of review choices),
immediate and delayed achievement, attitude, and review efficiency (i.e.,
practice-to-posttest review effects). The treatments were applied to four
similar classes of teacher education students in two higher education
institutions on a common campus (see Chapter 3). Subjects from two of
the classes were graduate students beginning a combined elementary
teacher certification and Master's degree program. Participants from the
third class were students from a combined undergraduate secondary
teacher certification and Bachelor's degree program at another college.
Finally, the fourth class consisted of graduate students beginning a
combined secondary teacher certification and Master's degree program at
the same university as the subjects from the first and second classes.
Individual difference data were collected for each of the 92
subjects. Information on gender, age range (20s, 30s, 40s, 50s), academic
ability, task vs. learning goal orientation (based on Ng & Bereiter, 1991),

and review potential (intended to assess a subject's tendency to move
ahead or go back to review while studying) was gathered prior to the
treatments. Academic ability data consisted of individual Program for
Licensing Assessments for Colorado Educators (PLACE) Basic Skills
Test scores (based upon availability and subject consent). Goal
orientation and review potential scores were obtained from pre-
measures described in Chapter 3. Homogeneity of the subject pool (the
mix of undergraduate and graduate students from different settings) and
the potential use of individual difference data as covariates or factors
were assessed during preliminary analysis (see Chapter 4).
Expectations for the Study
Because of the lack of literature addressing the inclusion of humor
in CBI, and the equivocal nature of previous research findings in the
area of traditional educational uses of humor, the expectations for the
effects of treatment were necessarily modest. It was also recognized that
learner-controlled exposure to treatments within a review loop would
provide interesting data if a sufficient amount of reviewing takes place.
Nevertheless, an initial position had to be established based on the
review of the literature in the form of the research hypotheses presented
in Chapter 3. Summarizing these expectations, specific findings from the
project were predicted to show: (a) there would be increases in voluntary
review choices (total number of reviews) for content-related and
-Unrelated review screen humor treatments compared to the no humor

condition, and for the content-related humor treatment compared to the
content-unrelated humor treatment; (b) there would be increases in
immediate and delayed achievement for content-related and -unrelated
humor treatments compared to the no humor condition, and for the
content-related humor treatment compared to the content-unrelated
humor treatment; (c) of the three treatments, only content-related
humor would have a positive impact on retention of information
between immediate and delayed posttests; and (d) more positive attitudes
toward the CBI and the subject of instructional objectives would result
from content-related and -unrelated review screen humor treatments
compared to the no humor condition, and for the content-related
treatment compared to the content-unrelated humor treatment.
Limitations of the study
The importance of including individual difference variables and
learner characteristics (e.g., ambiguity/frustration tolerance, cognitive
flexibility, curiosity, field articulation, and serialist/holist learning style)
in humor research was stressed previously (Teslow, 1995). A limitation
of the current study was the focus on motivational and cognitive effects
of review screen humor usage without considering several of these
learner characteristics. The present design did include plans for an
assessment of the relationships between five individual difference
variables (gender, age range, goal orientation, review potential, and
aptitude) and the seven main dependent variables (voluntary review

choices, posttest score, delayed posttest score, and four attitude subscales);
however, most of these data were used only for descriptive purposes,
proving to be of limited value as factors or covariates due to a lack of
variability for the sample, or a lack of homogeneity across treatments (see
Chapter 4). Further factorial and correlational analyses that include these
and other individual-difference variables were deferred to follow-up
The instructional design (ID) literature is filled with theoretical
dialogue and evidence of a paradigm shift toward constructivist and
postmodern ideas (Jonassen, 1991); but alas, very few design models,
tools, or suggestions have been provided to the practitioner (Wilson,
Teslow, & Osman-Jouchoux, 1995). I applaud efforts to develop flexible
ID models that provide rich learning environments and opportunities
for multiple interpretations of "what is correct." However, many CBI
products will continue to be recognized as drill-and-practice platforms
whereby designwe limit interpretations due to specific skill needs
(Merrill et al., 1992). The study of humor in any of these diverse learning
environments is an inviting, natural research area (as evidenced by the
review of the literature). For example, a study of learner-generated
humorous elaborations and other humor-containing interactions with
expert systems would be quite interesting. However, in this research
domain (humor applications in CBI) we need to build a foundation.
Thus, the current project focus was limited to the mainstay, drill-and-
practice learning environment. In my opinion, it proved to be an

excellent starting point. Recommendations based on "lessons learned"
that will help other researchers improve upon the methods employed in
the current study are suggested in Chapter 5.
Implications of the Study
The results of this study of humor use as an incentive to
encourage learner-controlled access of review loops in a lean feedback
and remediation CBI environment provide additional data for the
community of scholars researching learner motivation. The current
study reports on findings that contribute to areas of interest to this
community, including effects of this unique form of incentive on
immediate and delayed achievement (see also, Hicken, Sullivan, &
Klien, 1992; Ross & Morrison, in press); on attitudes (see also, Morrison
et al., 1995); on voluntary selection of review screens (see also, Morrison
et al.); and on more mindful and efficient use of review opportunities
(see also, Salomon & Globerson, 1987).
The reactive design characterized by a relatively fixed linear
sequence and simplified feedback and remediation loops is usually in
need of a little affective, motivational, and cognitive enhancement
(Teslow, 1995). A bonus for this project was that it also provided a
simplified model that helped isolate the impact of embedded humor.
Even though many of the results of this baseline study indicated no
significant differences between treatment groups, the findings provide a
starting point for extensions of humor research into more complex

feedback and remediation strategies in richer, more constructivist
learning environments. The implications for findings from the study
are: (a) encouragement of further studies of strategies for increasing
learner engagement in feedback-remediation processes; and (b)
contributions to definitive prescriptions for humor applications that
reflect the affordances of modern instructional technology.

The purpose of this chapter is to review the theoretical and
research literature in two scholarly areas: (a) interactive strategies in CBI,
and (b) humor applications in education. First, a review of the literature
from three areas of interactive CBI researchfeedback and remediation,
learner control, and intentional learning environmentsresults in a
simplified conceptual model (a lean feedback and remediation design)
that provides a venue for baseline CBI humor research suggested by
Teslow (1995). Next, a review of humor research literature provides a
foundation for applications of humor that may benefit motivation,
information processing, and affect in CBIparticularly basic designs
represented by the conceptual model.
Interactive CBI
Schwier and Misanchuk (1993) refer to the most basic level of CBI
interaction as reactive. Reactive interaction involves a response by a
learner to presented stimuli or a response by the program to an action by
the learner (including coaching and tutorials). They contrast this basic
level with a proactive level of interaction that is characterized by designs
that allow the learner to generate unique constructions beyond designer-

imposed limits (e.g., learner-generated elaborations or reflections); and a
mutual level that is characterized by artificial intelligence and virtual
reality designs (still in their infancy) that provide an environment where
the learner and the system are mutually adaptive. These levels of
interaction relate directly to the range of complexity described in the
literature related to the design of practicea central theme of CBI theory
and research that will be discussed next.
How Do I Get to Carnegie Hall?
"Practice, practice, practice" is the usual answer to this question.
Practice is also a time-tested learning principle (Gagne, Briggs, & Wager,
1988). A traditional function of a learning environment is to require
learners to practice using a skill or new knowledge. One of the most
common and complementary features of different instructional design
theories is the importance placed on practice (Hannafin & Hooper, 1993).
Salisbury, Richards, and Klein (1985) point out that recommendations for
the design of practice may be rationalized or labeled differently among
theoretical camps, but they overlap significantly. Schwier and
Misanchuk (1993) note that Keller's (1983) Attention-Relevance-
Confidence-Satisfaction (ARCS) model speaks directly to the design of
practice events in instruction. These activities should: gain and
maintain interest (Attention); be directly related to the content and larger
instructional context (Relevance); provide a challenging but comfortable
level of difficulty with ample instructional support (Confidence); and

reward successful performance (Satisfaction). Practice not only motivates
by providing opportunities for active participation by learners, but it also
provides a means for assessing their progress so that corrective feedback
and remediation can be provided if needed (Smith & Ragan, 1993). The
latter corrective function has engaged the instructional design
community since the 1950s, when B. F. Skinner extended Sidney
Pressey's (1927) ideas on self-paced, auto-instructional devices and
Key components of successful CBI design according to Skinner
(1958) include: clear instructional objectives, the teaching of substeps as a
way to attain mastery of larger units, allowing students to progress at
their own rate, and carefully programmed (or sequenced) instruction.
Many instructional designers still refer to Skinner's lesson frames, which
consist of: (a) a small amount of instructional information, (b) questions
that deal specifically with the information, (c) student response to the
questions, and (d) feedback on the response (Skinner, 1958). Under the
principles of Skinnerian behaviorism, almost error-proof practice is
created that motivates learners through success (Smith & Ragan, 1993).
Several CBI design strategies have evolved from the continuing
influence of Skinner and the seminal work of Robert Gagne.
Gagne (1985) influenced the design of instructional events by
stressing the importance of internal cognitive processes such as gaining a
student's attention or motivation, and retrieval of information from
memory during practice. In contrast to an error-proof behavioral

approach, designers working under cognitive principles tend to create
practice that evokes "...any misconceptions about the new information
the learners might have developed....[and] seems to pique learners'
interest even more than successful experience" (Smith & Ragan, 1993, p.
78). This is the type of interaction that human teachers do wellask a
learner several questions, determine what misconceptions he/she holds,
then provide guidance as to what the problem seems to be (Criswell,
Similar feedback and remediation features in CBI would provide
opportunities to increase its power by further individualizing and
adapting instruction, but a current limitation of technology is that it is
generally not intelligent. Emerging technologies that utilize artificial
intelligence, intelligent tutoring systems, and advanced presentational
features could further improve the effectiveness of adaptive or
conditional feedback and remediation (Hannafin et al., 1993); however,
the designer of CBI must recognize that programming complexity
significantly affects design and development effort and associated costs
(Hannafin & Peck, 1988; Sales, 1993; Smith & Ragan, 1993). In the
following sections these trade-offs are explored in the light of research
Feedback and Remediation
"Feedback" is a term widely used to refer to the communication
with a learner to inform him/her of the accuracy of a response to a

practice question or a set of questions (Cohen, 1985; Sales, 1993). Kulhavy
and Wager (1993) point out that, as with practice, reasons for using
feedback are still with us that also reach way back in time. Feedback is
viewed as: (a) a motivator or incentive for greater effort in the future by
letting people know how well they are performing on a task (Brown,
1932), (b) a reward or reinforcement for a correct response (Thorndike,
1913,1927), and (c) a unit of information that allows learners to adapt
after error responses (Pressey, 1927). The feedback-as-information
interpretation has been heavily researched since the 1960s when a shift
occurred from a behavioral view of feedback as a reinforcing function to
one reflecting an interest in how feedback influenced cognitive and
metacognitive processes (Dempsey, Driscoll, & Swindell, 1993).
The feedback literature separates feedback types into two strategy
levels: simple and complex (Morrison et al., 1995; Schimmel; 1988).
Simple forms include confirmation feedback, commonly referred to as
knowledge of results (KOR), that informs the learner of the accuracy of a
response (i.e., "Correct" or "Wrong"). Confirmation feedback is generally
effective for reinforcing correct answers. For incorrect answers it
provides little information from which the student can deduce the
correct answer (Hannafin & Hooper, 1993). Another simple form is
corrective feedback, which supplements incorrect responses with
knowledge of correct response (KCR) immediately after a response, or in
the case of delayed feedback, for all questions at once. Proponents of
delayed feedback suggest that it creates a second learning trial that is

more effective than an immediate feedback condition, because the
temporal separation causes less interference between incorrect answers
and corrective feedback (Kulhavy & Anderson, 1972). Simplified forms
of feedback may also provide remediation after feedback by suggesting or
requiring a return to previously-viewed (or recast) material.
Complex types of feedback involve additional interaction with the
learner through either reading or responding (Morrison et al., 1995). The
most common form is answer-until-correct (AUC) feedback, in which
the learner simply responds until the correct answer is given. More
sophisticated interactions combine feedback and remediation. They are
aimed at repairing flawed knowledge, beliefs, and faulty logicoften
very durablethat create learning inhibition (Spiro, Feltovitch, &
Coulson, 1990). Explanatory feedback supplies relevant information and
events central to the desired response (Hannafin & Hooper, 1993). The
goal is to explain why the given answer is incorrect, drawing attention to
misconceptions that require attention (Ross & Morrison, 1993).
Diagnosticor prescriptivefeedback goes farther, attempting to
identify the source of the misconception by comparison with common
errors and recommending a solution for the amelioration of deficiencies,
such as basic or extra instruction on a specific topic (Hannafin et al., 1993).
Another complex level, elaborative feedback, is based upon the same
cognitive principles as elaboration in general. It supplements or extends
the knowledge assessed in the practice question, attempting to establish

connections between the new content and the learner's prior knowledge
(Hannafin & Hooper, 1993).
As one might expect, research indicates that alternative feedback
forms may not be equally effective. A review of the literature by Clariana
(1990) indicated that several studies have shown KCR to be superior to
KOR, and KOR to be superior to no feedback, and some studies have
favored AUC and elaborative feedback over KCR feedback. Reviews by
Bangert-Drowns, Kulik, Kulik, and Morgan (1991); Kulik and Kulik
(1988); and Schimmel (1983) concluded that none of these sorts of
relationships are well-established. A review by Kulhavy and Stock (1989)
uncovered over 50 studies of various forms of feedback that had been
conducted since Crowder (1961) experimented with "branching"
programming styles that presented information in addition to simply
telling students the correct answers. Several studies indicate that
complex feedback forms containing more information generally produce
higher learner performance (Albertson, 1986; Collins, Carnine, &
Gersten, 1987; Grant et al., 1982; Hannafin, 1983; Roper, 1977); however,
others report no significant effect on performance due to complex
feedback (Corbett & Anderson, 1990; Hodes, 1985; Kulhavy, White, Topp,
Chan, & Adams, 1985; Merrill, 1987). As Kulhavy and Wager (1993) put
it, "There appears to be no consistent pattern of results in this sizable
literatureother than the general finding that some feedback is better
than no feedback" (p. 12).

Successes have been reported in the development of systems
designed around the traditional apprenticeship training model that
provide authentic practice activities and representations of expertise as
targets of instruction (Brown, Collins, & Duguid, 1989; Collins, Brown, &
Newman, 1989; Palincsar & Brown, 1984; Resnick, 1987; Scardamalia &
Bereiter, 1991; Schoenfeld, 1985). Feedback and remediation in these
designs provide external support or scaffolding in the form of "...ideal
modeling of performance, hints, reminders, explanations, or missing
pieces of knowledge...that are sensitive to changing student needs at
different stages of skill acquisition" (Gott et al., in press). As indicated in
Chapter 1, these and other new directions for feedback and remediation
design require intelligent tutoring systems that can be very expensive
and time-consuming to develop (Merrill, 1987; Smith & Ragan, 1993;
Spock, 1987). It may be some time before integration of existing research
and new studies provide evidence that the effort is worthwhile.
Learner Control
Another factor that affects the complexity of CBI design is locus of
instructional control, which exists on a continuum from program
(external) control to learner (internal) control. Most instructional design
and research discussion of this topic refers simply to learner control, the
degree to which the learner controls the pace, navigation, and
instructional choices in CBI. Bereiter (1973) has written that, "given an
ample choice of activities deemed worthwhile and some help in making

up their minds, [learners] will tend to engage themselves in worthwhile
activities" (p. 88). Following this philosophy, some instructional
designers allow significant learner control of navigation, including the
decision to review material or continue to the next topic after an error.
On the other hand, Dewey (1938) maintained that education involves the
selection of experiences for others that present possibilities for future
development. This perspective is accomplished in some CBI designs via
"preordained" interactions and selections (Schwier & Misanchuk, 1993).
As with feedback and remediation research, learner control studies
have been inconclusive. Some studies have indicated no significant
differences between learner control and program control treatments
(Balson, Manning, Ebner, & Brooks, 1985; Klein & Keller, 1989; Mayer,
1976; Reiser & Sullivan, 1977). However, other findings have indicated
significantly enhanced achievement due to learner control (Kinzie,
Sullivan, & Berdel, 1988; Gray, 1987). The literature indicates that learner
control benefits high achievers (Borsook, 1991; Gay, 1986; Hannafin &
Colamato, 1987); provides motivation (Santiago & Okey, 1990; Steinberg,
1977, 1991); and provides learners with regulation of presentations with
respect to text density (Ross, Morrison, & O'Dell, 1988).
Some researchers have shown that learner control may have a
negative impact on achievement (Fisher, Blackwell, Garcia, & Green,
1975; Fry, 1972); may increase time spent learning (Santiago & Okey,
1990); and may permit learners to make poor decisions about how much
practice they need (Ross, 1984; Tennyson & Buttrey, 1980). The literature

also indicates that naive learners or low achievers benefit more from the
structure of program control (Borsook, 1991; Carrier & Jonassen, 1988;
Higginbotham-Wheat, 1990; Kinzie et al., 1988; Schloss, Wisniewski, &
Cartwright, 1988); and that learner control with advisement is superior
for enhancing achievement, curiosity, on-task behavior, and self-
challenge (Arnone, Grabowski, & Rynd, 1994; Hannafin, 1984; Mattoon,
Klein, & Thurman, 1991; Merrill, 1980; Milheim & Azbell, 1988; Ross,
1984; Santiago & Okey, 1990; Tennyson & Buttrey, 1980). Moreover,
Johansen & Tennyson (1984) found that with continued practice students
gradually improve in their ability to make strategy decisions from
advisement information.
Intentional Learning Environments
Adaptive control systems can certainly make use of modern
computing power to provide strategy advisement and artificial tutoring
(Burns & Parlett, 1991); and program-control constraints, upon detecting
an error, can require the learner to repeat an information presentation
and/or retry questions before continuing (Hicken, Sullivan, & Klein,
1992). However, these technologies do not take into consideration the
individual's responsibility for intellectual development and learning
with understanding (Tennyson & Buttrey, 1980), and the importance of
the student's own desires and intentions as they encounter instructional
tasks (Bereiter & Scardamalia, 1992).

The complex interaction of learner characteristics and task
demands was studied recently by Ng and Bereiter (1991). Using think-
aloud protocols, they identified three levels of goals in a group of 16
adults learning the BASIC programming language. The first goal, task-
completion, was characterized by students who relied on persistence and
exercise, and looked at learning as identifying what needs to be done.
The second goal level, an instructional orientation, included students
who saw learning as a prescribed set of cognitive gaps to be filled, and
responded to the instruction in a way calculated to improve their
performance on an evaluation. Students characterized by the third level,
a knowledge orientation, actively constructed learning agendas for
themselves, concentrated on making the most of the instruction that was
available, and treated difficulties as problems to be solved. The authors
stress that recognizing a hierarchy of goal orientations should have an
impact on the design of instruction. The Ng and Bereiter study
illustrates why integrative models of motivation and cognition are
receiving increased attention (Pintrich & Garcia, 1994). Researchers have
shown, for example, that the motivational construct anxiety is linked to
information processing (McKeachie, 1951). Intrinsic and extrinsic goal
orientation (Ames, 1992; Brophy, 1987; Weiner, 1980) and self-referenced
systems of motivation, such as self-efficacy and self-determination
(Bandura, 1986; Deci & Ryan, 1985), are associated with cognitive
engagement and metacognitive learning strategies (McKeachie, 1990).
According to one motivation model (Brophy, 1987; Weiner, 1980) value

(the benefits the activity will bring to a student) works hand-in-hand
with expectancy (the degree to which a student expects to be able to
perform a task).
Hoska (1993) points out that most CBI lessons overlook an
important responsibility of any learning environmentto motivate so
that learners invest the effort required to gain knowledge and skill. This
includes structuring a learning situation to foster a learning goal
orientationespecially important for performance-oriented learners
who do not associate effort with success (Dweck, 1986; Dweck & Leggett,
1988). Learning environments that encourage mindful behavioractive
examination and exploration of the information presentedallow the
learner opportunities to reflect, analyze meanings, and consider
alternative interpretations before making a cognitive commitment to an
idea (Salomon & Globerson, 1987). Unless learners are motivated to try
to gain deeper understanding, they may not achieve more than simply
surface-level processing (Morrison et al., 1995; Schiefele, 1991). Feedback
and remediation that is too "helpful" may actually promote this limited
processing, inhibiting mindful behavior, and diminishing active
learning (Bangert-Drowns et al., 1991). Likewise, locus of control that
does not encourage learners to act on their knowledge of their own needs
will result in incidental rather than intentional outcomes (Bereiter &
Scardamalia, 1989).

Conceptual Model
The main challenge for instructional designers has been to
stimulate more active engagement with learning materials (Ross &
Morrison, 1993). A review of the literature on just two areas of CBI
research (feedback-remediation strategies and learner control) leaves one
awed by the complex interplay of affective, behavioral, and cognitive
components of human learning that is most likely the root cause for
equivocal findings. In frustration, it seems, we in the instructional
design community try to improve our repertoire of strategies and tactics,
turning to more and more powerful computing techniques as our
awareness and appreciation of the complexity of learning grow. But, is
this necessarily the only road to travel? Previously, it was suggested that
" purveyors of research and development we owe the users of our
output nothing less than the thorough exploration and exploitation of
all powerful human attributes" (Teslow, 1995, p. 23). One of these
humormay provide a means of achieving CBI instructional successes
in simplified (lean), rather than more complex (full) learning
Let us return to the basics for a fresh start. Cognitive theory
suggests that performance is unreliable without substantial practice and
that success can occur in multiple ways (Lesgold et al., 1989). In CBI the
pathways that lead to success may be determined by second chancesthe
interactions that are available to the learner after practice and feedback.
From an economic point of view that considers the effort and resources

that go into the development of CBI, a lean feedback and remediation
design that provides the simplest context for intentional learning
through learner control is represented by the upper row of Table 2.1; and
contrasted with full, more complex feedback and remediation (lower
row). A lean configuration simply asks an embedded question (practice
item), tells the learner if his/her answer is right or wrong, provides the
correct response if an error occurs, and suggests a review of previously-
viewed material at the discretion of the learner.
Feedback Locu3 of Control Remediation
Lean KCR Feedback Learner control Same/Recast
Full Complex Feedback Program control Adaptive Remediation
Table 2.1 Lean Feedback and Remediation Conceptual Model with
Knowledge of Correct Response (KCR) Compared to Full Model
The following review of the theory and research literature on
instructional humor will focus on this simplified conceptual model of
CBI as a point of departure for studies of humor applications in CBI. The
model's validity as a "tried-and-true" auto-instructional platform, dating
back to Pressey (1927), and its simplicity will facilitate the isolation of the
effects of humor variables in the current study.

Humor in Education
Most of us would agree that much of what we each understand
about human nature and society is constructed through a powerful
forcehumor (Cousins, 1979). Indeed, one of the most powerful "ah-
ha" experiences is the "oh-I-get-it" experience (Teslow, 1995). Writers,
speakers, entertainers, and educators who exploit humor recognize that
it seems to be uniquely valuable to humans (Morreall, 1983). Humor is
an authentic component of most work, play, and educational settings
an ubiquitous feature of our culture and media (McGhee & Goldstein,
The American Heritage Dictionary (Soukhanov, 1992) provides a
modern definition of humor: "The quality of being amusing or comical"
(p. 627). Morreall (1983) defines amusement as the feeling produced by a
pleasant psychological shift, and laughter as the physical activity that is
sometimes caused by this feeling. Aspects of humor and laughter have
also been associated with aggression, defense mechanisms, and mild fear
(Ziv, 1984). It is well-known that styles of humor are culture-dependent.
In fact, scholars report that some of the greatest differences between
cultures are in the contents and situations of humor (Guegan-Fisher,
1975; Ziv, 1988). Acknowledging the cultural basis of humor is vital for
any scholarly inquiry. Accordingly, this review of the literature on
educational applications of humor will focus on contemporary humor as
practiced in North America.

Educators generally accept that humor is qualitatively valuable
(Bryant, Gula, & Zillmann, 1980; Ziv, 1976), and anecdotal evidence seems
to support the belief that "What is learned with laughter is learned well"
(Robinson, 1983, p. 121). Yet humor seems to be absent in most
technology-based education and training products. The relative paucity of
references to humor in instructional technology literature would lead one
to conclude that the field considers it to be irrelevant. Instructional
designers may be wary of humor for many reasons, including: sensitivity
to the dangers of humor, a lack of exposure to theories and research
findings, and a need for definitive prescriptions for humor applications
that reflect the affordances of modern technology. Teslow (1995) addressed
these concerns, providing cautions, theoretical foundations, findings from
humor research, and an agenda for future research aimed at providing
guidance for humor applications in CBI. In the following sections I draw
upon this previous work for a discussion of cognitive and affective aspects
of humor and implications for lean feedback and remediation designs.
Humor Theory and Research
Anyone who surveys the theoretical literature associated with
humor will soon uncover a repetition of three main themes upon which
modern hypothetical work is based (see Chapman & Foot, 1976;
Coleman, 1992; MacHovec, 1988; Morreall, 1983; Paulos, 1980, 1985). The
first is superiority theory, put forth by English philosopher Thomas
Hobbes in 1651. This construct relates humor to the "sudden glory" (his

term), self-satisfied superiority, or gloating a respondent feels toward
others being disparaged or depreciated (Zillmann, 1983). Superiority
theory represents the potentially dangerous side of humor in varying
degreesfrom fairly innocent sarcasm to disparagement. The physically
and mentally challenged, ethnic and racial groups, and women have
borne the brunt of this unkind humor (Anderson, 1989). The consensus
in the literature is that ridicule, insult, sarcasm, and ethnic, racist,
violent, or sexist jokes and cartoons offer little benefit for instruction
(Anderson, 1989; Lodish, 1993; Rareshide, 1993; Surlin, 1978). The use of
even the mildest forms of this type of humor in CBI does not serve the
interests of the instructional design community; thus, the rest of the
review will not dwell on it.
The second theme is incongruity theory, originated by James
Beattie in 1776 (Morreall, 1983). The idea behind this popular framework
is that two or more odd, inappropriate, inconsistent, or unsuitable parts
or circumstances are juxtaposed, with resulting surprise or
unexpectedness (Rothbart, 1976). The notion that incongruity is the basis
for humor was further developed by Beattie's contemporary, the
philosopher Immanuel Kant. His famous formulation equated humor
with the "sudden transformation of a strained expectation into nothing"
(Paulos, 1980). Much later, Max Eastman (1936) referred to this
interpretation of humor as derailment theory, pointing out that our
normal flow of thoughts is sometimes derailed by an encounter with

The third explanation for the effects of humor is relief theory,
which relates to surplus energy that overflows as humor, amusement, or
laughter when "consciousness is transferred from great things to small,"
as Herbert Spencer put it in the early 1800s (Paulos, 1980). Darwin's
similar interpretation of laughter as the draining, or venting, of
unnecessarily generated energy greatly influenced Freud (1905/1960),
who treated wit and humor in his Jokes and their Relation to the
Unconscious. He maintained that emotional censors in the mind form
powerful barriers that suppress forbidden thoughts. Jokes, with their
compact ambiguities, elude the simple-minded censorswhich react to
only innocent surface meaningsenabling a person to vent aggressive
or sexual feelings and anxieties in a disguised manner (Minsky, 1984;
Morreall, 1983).
The Cognitive Connection
Humor designed around the incongruity theme benefits learners
via enhancement of interest and arousal in the forms of attention,
curiosity, and incentive (Keller, 1983). Also, advance organizers, which
facilitate learning and retrieval, need not be serious prose passages.
Effective organizers have taken the form of games and visuals (Jonassen,
1982) which could contain humor. Moreover, links between
incongruity, problem solving, and creativity point to the potential for
using humor to enhance higher-order cognition as well. Comedian
George Carlin puts it this way: "No one is ever more him or herself than

at the moment they are laughing...all defenses are is at a
moment like that when a new idea can best be implanted" (as cited in
Grace, 1991, p. 21). In the following sections humor research is reviewed,
implications for lean feedback and remediation designs are discussed,
and gaps in the literature are identified for attention, motivation, and
information processing.
Attention. Keller (1983) notes that humor is an effective method
for varying the presentation of instructional elements for the
maintenance of learner attention. However, research indicates that the
effectiveness of jokes and humorous anecdotes depends upon the age of
the learners, and on how the humor relates to the material being taught
(Bryant et al., 1981; Hannafin & Hooper, 1993; Hill, 1988). Zillmann,
Williams, Bryant, Boynton, and Wolf (1980) found that fast-paced,
unrelated humorous inserts (i.e., irrelevant to the content to be learned)
in televised instruction improved attention for 5- to 7-year-old subjects.
However, a summary of 33 studies on humor use in higher education
found that unrelated humor interfered with immediate recall of
concepts taught in a college classroom (Erickson, 1987). Researchers have
also noted that when humor causes arousal above a moderate level,
learning performance declinestoo many cues detract from the
processing of a particular message (Brockner, 1979; Brown & Itzig, 1976;
Vance, 1987). We have all experienced the distracting nature of humor
in the extreme. Sometimes when a thought is perceived to be
humorous, "...further reasoning is drowned in a flood of activity

furious motions of thorax, abdomen, head, limbs and face, accompanied
by loud barking, wheezing, and choking noises" (Minsky, 1984, p. 189).
Thus, the inverted U graphical relationship between arousal level and
performance first suggested by Yerkes and Dodson (1908) should be an
important consideration when designing humor into instruction.
Since both humor and curiosity are related to cognitive conflict and
uncertainty resulting from reactions to new, strange, incongruous, or
mysterious elements, the inverted U relationship between anxiety level
and performance would also appear to be important. Optimally, we would
like our learners to enter a pleasurable attentional level described by Day
(1982) as a zone of curiosity. Low uncertainty caused by, say, humor that is
easily interpreted, results in low motivation and disinterest (short of the
optimal zone of curiosity); whereas, a stimulus such as subtle humor that
presents complexity, unfamiliarity, or high uncertainty levels, results in
frustration, avoidance, and disinterest again. Learners in the latter
condition have entered a zone of anxiety with resulting negative impact
on motivation and performance (Day, 1982).
Hannafin (1987) warns, "Simply elevating arousal, in the absence of
an appropriate orienting perspective, is unlikely to improve, and may
actually hamper, learning" (p. 214). Thus, for adults (the focus of the
current study) using lean feedback and remediation CBI, attentional
humor is probably most beneficial when it complements the instruction
by directly illustrating a point (Bryant et al., 1981). The review of the
literature indicates a lack of research concerning: (a) relationships between

attentional humor treatments in CBI and zones of arousal and anxiety,
and (b) relationships between learner characteristics and attending when
humor is present in CBI (Teslow, 1995).
Motivation. Think of a time when you did not get a joke because
of a lack of knowledge. The social coherence function of humor (i.e.,
wanting to be in the clique that gets the joke) may have motivated you to
pretend understanding, but subsequently investigate a content area that
you would not normally value. This phenomenon illustrates that
besides the attention effect discussed above, humor motivates
exploration of an environment in order to resolve social or cognitive
conflict caused by complex stimuli (Arnone, Grabowski, & Rynd, 1994;
Berlyne, 1960; Nerhardt, 1976).
There is little documentation on the motivational effects of humor
in CBI. Roman (1985) found that humor as a teaching tool in CBI reduced
adult trainee resistance to change and maintained interest. Fiderio (1986)
reported that the judicious use of humor in computer-based training was a
factor in adult students' decision to stick with the instruction. These
findings are promising, but one wonders if the interactive computer
interface, which presents an opportunity for action, is being fully exploited
as a means for enhancing motivation through humor. Laurel (1990, 1991)
maintains that the use of drama theory borrowed from the theatrical
domain and applied to on-screen anthropomorphic agents can make
human-computer interactions more "...enjoyable, invigorating, and
whole" (Laurel, 1991, p. 120).

In an attempt to harness this enjoyment, Snetsinger and Grabowski
(1994) used a humorous agent (Count Tickula) in an undergraduate CBI
science lesson about tick identification. They reported that, while humor
did not significantly impact learning and retention, the more dramatic
humorous reinforcement of concepts did result in significantly more
worry about ticks and tick-borne diseases than non-humorous treatments.
The authors presumed that the students in the humor group were
engaged to the extent that, at least, this differential outcome was achieved.
Similar engagement potential is evidenced by the Jasper Series produced
by the Cognition and Technology Group at Vanderbilt (1990). These
videodisc-based adventures encourage fifth- and sixth-grade students to
collaboratively resolve the ending of a story by solving a complex
mathematical problem. The instruction is anchored in contexts that are
complex and realistic enough to provide reasons for the usefulness of
information. The Vanderbilt Group has made a design decision to deliver
much of the instruction in a humorous way, using a comical-looking
character named Larry who knows lots of interesting information
(Cognition and Technology Group at Vanderbilt, 1992). It is probable that
significant disappointment would occur among previous Jasper Series
users if Larry were deleted from a future adventure.
The literature indicates that the motivational construct, incentive,
is related to performance (Hicken et al, 1992; Sullivan, Shultz, & Baker,
1971), and may provide opportunities for interesting extensions of
previous research through the addition of humor variables. Clabby (1979)

examined humor elements as reinforcing tools in grammar lessons.
Experimental group subjects who correctly selected nouns were shown
humorous cartoons. Non-noun selection was followed by non-humorous
drawings. Results indicated that humor significantly facilitated
intentional learning for low-creative groups. This idea of providing
reward-like humor as part of a target activity presents opportunities for
research into the motivational constructs, value and expectancy (Brophy,
1987; Weiner, 1980). Value-enhancing strategies in lean feedback and
remediation designs could include opportunities for humorous
interpretations of content (e.g., the viewing of humorous perspectives or
cartoon-like fantasy during remediation). Also, as mentioned above,
some humor places heavy social demands on us, and we value
information that helps us perform socially. Expectancy for success in lean
feedback and remediation designs may be enhanced if supportive
humorous feedback messages are used occasionally to ease tension, lessen
anxiety, and increase a learner's confidence (Weinberg, 1973).
The review of the literature indicates a lack of research
concerning: (a) the effects of the addition of humorous elements as a
means of enhancing the attractiveness of review screens (compared to
serious treatments), or as lures that encourage access of these screens; (b)
relationships between the presence of motivational humor elements in
CBI (e.g., humorous on-screen anthropomorphic agents, or humor-as-
reward treatments) and learner engagement in learning activities; (c) the
effects of humor elements in CBI on learners' perceptions of the value

and relevancy of content; (d) the effects of humor-resolving activities
provided as a reason for learning new information; and (e) relationships
between learner characteristics and motivation when humor is present
in CBI (Teslow, 1995).
Information processing. Incongruity theory represents humor as a
reaction to something perceived that results in surprise. This receptive
aspect is discussed in the literature using the rubric humor
appreciationthe ability to understand and enjoy messages containing
humorous situations that are incongruous but not menacing (Ziv, 1984).
Another perspective that is pertinent to information processing is the
expressive aspect of humor. Ziv (1984) defines this dimension as humor
creativitythe ability to perceive relationships between people, objects,
or ideas in an incongruous way, as well as the ability to communicate
this perception to others. Since the scope of this literature review is
limited to reactive levels of CBI interaction, only research on
information processing aspects of humor appreciation will be discussed
in this section. See Teslow (1995) for implications of humor generation,
which would be possible in proactive and mutual levels of CBI
Suls (1972) proposed a two-stage cognitive model of humor that
characterizes the comprehension and appreciation of cartoons and jokes.
During the first stage, perceivers find that their expectations are
disconfirmed (an incongruity is experienced) in the last panel of a comic
strip, by the ending punch line of a verbal joke, or by an extraordinary

situation or caption in a cartoon. The second stage is a form of problem
solving that involves finding a cognitive rule that reconciles the
incongruous parts. Thus, to appreciate humor we must both identify
and resolve the incongruity. The more implicit the resolution, the more
varied the perspectives and interpretations. Verbal or textual jokes
engage the subject in differentiating, visualizing, and semantic problem
solving. Visual humor is unique in that it encourages comparing,
contrasting, and hypothesis-generation (Teslow, 1995).
The implication for lean feedback and remediation strategies is
that humorous text and visual images can be useful as a basis for
forming thought processes and mental images, which can facilitate
information acquisition and retrieval (Brody, 1982; Snetsinger &
Grabowski, 1993) that would assist learners in repairing their
misconceptions (Brown & Van Lehn, 1980). However, the implicit
nature of many cartoons may make it difficult to predict a learner's
interpretation, thus limiting their usefulness in instructional products.
In any modalityverbal, textual, or visualthe designer should provide
opportunities for humor resolutions that are not too obscure. Lacking a
resolution, or at least a reduction in cognitive dissonance, the
respondent is puzzled or even frustrated (LaFave, Haddad, & Maesen,
1976). Formative evaluation would help gauge the level of resolution
difficulty for representatives from the target population (Tessmer, 1993).
A review of the literature by Ziv (1988) indicated that much of the
early research showed that the introduction of humor into the teaching

process did not improve learning. For example, Bryant et al. (1981)
found that pictorial humorous illustrations had no effect on information
acquisition and negative effect on persuasibility. More recently,
Snetsinger and Grabowski (1994) found no significant difference in
retention between serious and humorous treatments for an
undergraduate CBI science lesson. Some studies indicated that humor
based on distortions, exaggerations of reality, and irony lead to perceptual
disorientation or faulty impressions in children (Weaver, Zillmann, &
Bryant, 1988; Zillmann et al., 1984). Countering these findings, other
researchers reported that humor has a positive effect on recall and
retention (Chapman & Crompton, 1978; Hauck & Thomas, 1972; Kaplan
& Pascoe, 1977; Vance, 1987).
Several weaknesses and limitations in former studies are
discussed by Vance (1987). He posits that negative conclusions in past
research may be due to an inaccurate understanding of humor as a
complex theoretical construct. Ziv (1988) notes that many studies were
conducted in artificial, experimental settings with no resemblance to real
educational situations, and no consideration of the time element for
cognitive processes. It is common, he points out, for published findings
to be based on experiments with durations of 10 minutes or less. The
review indicates gaps in the research literature concerning: (a)
relationships between content-related and -unrelated humor elements
and information processing in CBI, (b) relationships between humor-
resolving activities and problem-solving strategies, (c) relationships

between selected learner characteristics and information processing
when humor is present in CBI, and (d) the effects of different humor
presentation modes (verbal, textual, visual) on information processing
in CBI (Teslow, 1995).
The Affective Connection
Relief Theory is operationalized when a speaker warms up an
audience with a joke (Paulos, 1980). It has long been recognized that
humor is a beneficial strategy in education as a reliever of tension and
expectation. Sensitively-used humor and laughter in the classroom
opens avenues of communication (Hill, 1988), loosens fixed positions,
and can enable students and teachers to take risks and perceive ordinary
information in unusual patterns and connections (Bryant & Zillmann,
1988; Cornett, 1986; Goor, 1989; Warnock, 1989). Several research studies
have concentrated on the types of teachers' use of humor in the
classroom, focusing on taxonomies and typologies of teacher humor,
teacher effectiveness, and teacher evaluations (see Goor, 1989; Neuliep,
1991). Generally, teachers who use humor are rated as more
approachable by students, develop a positive rapport with learners, and
receive higher evaluations (Neuliep, 1991).
Does this putative rapport between student and instructor develop
if the humorous teacher appears on a video or computer screen?
Wetzel, Radtke, and Stern (1994) report that the general direction of
findings regarding learning via video and instructional television is that

humor is most effective for children, becoming less effective as learners
grow older, with negligible effects for adults. Also, some studies indicate
that the enjoyment experienced with humorous presentations may
lessen the satisfaction with non-humorous material (Chapman &
Crompton, 1978; Singer & Singer, 1979; Wakshlag, Day, & Zillmann,
1981; Zillmann, 1977). This contrasts with results from studies indicating
that positive affect created by humor increases the probability that
material will be remembered and recalled (Kaplan & Pascoe, 1977;
Monson, 1968).
Many scholars are calling for increased emphasis on the affective
domain of learning in instructional technology (Dick, 1992; Lebow, 1993;
Martin & Briggs, 1986; Reigeluth, 1989) where humor has a role to play
(Keller, 1983). Even though affective objectives are operationally
challenging, socially complex, intangible, and many times controversial
(Martin & Briggs, 1986), they support the move toward learning
environments that "...extend, amplify, and enrich our own capacities to
think, feel, and act" (Laurel, 1991, p. 33). Humanizing the learning
experience is especially important for CBI, which may be branded as
high-tech, machine-oriented, serious business for some learners;
especially for adults in training settings who are more familiar with
classroom learning modes. The challenge for designers of lean feedback
and remediation designs is to look for opportunities for technological
extensions of the classroom model of humor usage by skilled teachers.
This includes putting students at ease and creating an enjoyable learning

environment by: setting the tone with introductory humor, defeating
preconceived notions about "dry" content with humorous examples,
breaking passive learning habits or monotony with humorous
interludes, and releasing built-up tension during practice and assessment
activities with humorous scenarios.
Markiewicz (1974) and Gruner (1978) observed that studies of the
effects of humor have typically yielded insignificant results and are
characterized by poor methodology. The review of the literature indicates
that this appears to be true for the few studies of the affective impact of
humor in educational settings, including CBI. Specifically, there is a gap
in the literature concerning: (a) the conditions under which humor
causes negative affective associations with CBI; and (b) relationships
between learner characteristics and affect when humor is present in CBI
(Teslow, 1995).
A mainstay of CBI design practice is the reactive interaction
learning environment characterized by a relatively fixed linear sequence,
simplified feedback for single errors, and learner control over the pace of
instruction and the use of remediation loops (Criswell, 1989; Kulhavy &
Wager, 1993; Schwier & Misanchuk, 1993). Simplified remediation after
an error may involve rerouting the student through the same
presentation of information; it may add prompts or suggest strategies
that would help the student attend to a repeated presentation; or it may

recast the lesson by presenting the information in a different way
(Hannafin & Peck, 1988). A review of research findings on feedback and
remediation strategies, and locus of control indicated that simplified
designs appear to be just as valid as more complex systems, which claim
significant amounts of the CBI designer's time and budget. Thus, a
simplified conceptual model labeled a lean feedback and remediation
strategy (KCR feedback-learner control-voluntary review of previously
viewed material) was chosen as the standard against which humor
research findings were compared. Several possibilities for research into
the effects of humor variables in simplified CBI designs emerged from
the review of theoretical and research literature concerning
motivational, cognitive, and affective aspects of humor appreciation.

This chapter describes the subject sample, research design, data
analysis techniques, formal research questions and hypotheses, CBI and
other materials, and procedures used for the experiment proposed in
Chapter 1 to address the gap in the humor research literature identified
in Chapter 2. Formal variable names are capitalized to facilitate the
connection between discussions of research design and data analysis.
A sample of 110 subjects was drawn from graduate and
undergraduate students enrolled in elementary and secondary teacher
certification programs at two higher education institutions on the
Auraria Campus in Denver, Colorado, during Fall Semester, 1995. The
majority of the graduate students participating in the experiment were
enrolled in the Initial Teacher Education (ITE) Programa combined
teacher certification and Master's degree trackat the University of
Colorado at Denver (UCD). Subjects from three ITE program classes
participated in the study. Also included in the subject pool were a mix of
graduate and undergraduate secondary teacher education students from
one class at Metropolitan State College of Denver (MSCD), who were

enrolled in a combined undergraduate secondary teacher certification
and Bachelor's degree program. Specific course and section information
is provided in Appendix I. Subjects were randomly assigned in equal
numbers to one of three review screen treatments (a control with no
review screen humor, a condition with initial review screens containing
content-related humor, and a condition with initial review screens
containing content-unrelated humor). All other aspects of the
computer-based lesson on the topic, Instructional Objectives, were
identical across treatments. The CBI was designed to match the lean
feedback and remediation conceptual model described in Chapter 2.
Discussions with instructors for the four courses indicated that
participation in the proposed study was appropriate for their students.
Also, all instructors determined that the content was consistent with the
objectives and syllabi for their courses, and that the time allotment
(approximately one hour) was satisfactory. Also, the mix of
undergraduate and graduate students from two institutions was
determined to not be a concern after discussions with the MSCD
instructor. This professor indicated that he observed no differences in
maturity level, intellectual ability, motivation, or general classroom
behavior for his class as a whole compared to graduate courses he had
recently taught; and he noticed no such differences with respect to
undergraduates versus graduates. Expected characteristics of the entire
subject pool included the following: (a) predicted lack of variability in

gender with the majority of subjects being female; (b) predicted lack of
variability in age with most subjects expected to be in their twenties;
(c) predicted variability in academic ability due to diverse backgrounds,
education levels, and time out-of-school; and (d) predicted variability in
specific learner characteristics pertinent to the experiment (Goal
Orientation and Review Potential) for which instruments were designed
(see section on Materials below).
The procedure was administered to Classes 1 and 2 (the UCD
elementary education students) on two days in the same week, Fall
Semester, 1995. Class 3 (the MSCD students) and Class 4 (the UCD
secondary education students) received a slightly modified instructional
procedure three weeks later. This adjustment occurred at the midpoint
of the experimentbetween sessions for the first two classes and the last
two classes. After administering treatments to approximately half the
subjects, the first outcome of the experiment became evident. Initial data
analysis indicated that the KCR feedback environment resulted in
limited reviewing by subjects in all treatments. It was apparent from
review of open response data that several subjects were not aware of the
nature of the review loop elements (some containing humor), even
though an introductory screen provided a flow diagram and encouraged
reviewing (see Appendix B).
Since at least some reviewing was necessary for the review screen
treatment to take effect, a change was made to the CBI program that
forced all subsequent users to review at the first opportunity (after

Practice Quiz 1.1). The modification insured that all treatments provided
the subject with a chance to become procedurally familiar with the
nature of the review loop early-on (especially important for the Related
Humor and Unrelated Humor groups). Normalization for the
modification was based on Voluntary Review Choices. For all analyses
described in the next section, this involved removal of all subjects in
Classes 1 and 2 who did not review at all during the CBI session. For
Classes 3 and 4 normalization for some enroute analyses involved the
exclusion of the first (forced) review. Since the Class 1 and 2 subjects
"surviving the cut" all experienced the review process within the first
lesson segment (the first 20 percent of the CBI), it was determined that
effects of the CBI experience on overall achievement and attitude would
be the same for all four classes. Thus, the final experimental sample
resulted from a normalization due to the CBI modification and two
special cases. This involved the removal of 18 subjects from the original
pool of 110, including the following participants: 16 subjects from Classes
1 and 2 who did not review at all during the CBI session; a subject in
Class 2 who did not complete the CBI session; and a subject in Class 3 due
to an irrecoverable computer disk malfunction. These adjustments
resulted in a total subject pool of 92, with 31 subjects in the No Humor
treatment, 30 in the Related Humor treatment, and 31 in the Unrelated
Humor treatment.

Design and Analysis
Process outcomes (i.e., review behaviors) for each treatment were
compared using oneway analysis of variance (ANOVA). This involved
enroute data gathered by tracking routines built into the CBI that
recorded which review screens were accessed by the learner and the time
spent viewing each review screen. Oneway ANOVA procedures were
performed for Voluntary Review Choices (i.e., total number of review
loops activated by the learner), Average Time per Serious Review Screen
(for all three treatments), and Average Time per Humor Review Screen
(for Related Humor and Unrelated Humor treatments).
A repeated measures design was used to address another focus of
the studythe effects of review screen humor on achievement in the
form of comprehension and retention. Specifically, a two-way
(treatment x time) repeated measures analysis of variance (ANOVA)
tested the effects of three review screen treatments (no humor, content-
related humor, and content-unrelated humor) on immediate and
delayed achievement posttests (see Table 3.1).
Repeated Measures
Treatment n Posttest Delayed Posttest
No Humor 31
Related 30
Unrelated 31
Table 3.1 Two-way repeated measures ANOVA design

The design also included multivariate analysis of an affective
measure in the form of an attitude survey consisting of two scalesone
designed to measure attitudes toward the CBI, and one designed to
measure attitudes toward the content (Instructional Objectives). These
two scales in turn were each made up of two subscales. The CBI attitude
subscales assessed feelings regarding Fun and Design. The content
subscales were designed to assess feelings about Interest and Relevance.
These four subscales were analyzed using oneway multivariate analysis
of variance (MANOVA) with each subscale constituting a separate
dependent measure (see Table 3.2).
Dependent Variables Attitude Subscales
Treatment n Fun Design Interest Relevancy
No Humor 31
Related 30
Unrelated 31
Table 3.2 Oneway MANOVA design for Attitude Subscales
Attitude data were also gathered in the form of an open response
question for the Related Humor and Unrelated Humor treatments.
Written responses stored by the CBI were examined for commonalties.
These categorized and coded data resulted in two themes (remarks about
reviewing and comments regarding distraction caused by humor) that

were analyzed using a chi-square test. All statistical tests were performed
using an alpha of .05.
Research Questions
Major research questions and associated hypotheses (at a = .05)
addressed by the true experimental study of review screen treatments for
the sample were as follows:
1. Is there a review screen treatment effect for enroute behavior?
Research Hypothesis: There will be a significant main
treatment effect for Voluntary Review Choices, such that
means will be significantly higher for both Related Humor and
Unrelated Humor groups compared to the No Humor group;
and for the Related Humor group compared to the Unrelated
Humor group.
2. Is there a review screen treatment main effect for combined
immediate and delayed achievement?
Research Hypothesis: There will be a significant review screen
treatment main effect, such that combined means for
immediate and delayed posttests will be significantly higher for
both Related Humor and Unrelated Humor groups compared
to the No Humor group; and for the Related Humor group
compared to the Unrelated Humor group.
3. Is there a time-of-testing main effect?
Research Hypothesis: There will be a significant time-of-
testing main effect, such that means for combined immediate
posttests will be significantly higher than means for combined
delayed posttests.

4. Is there a review screen treatment by time-of-testing
Research Hypothesis: There will be a significant review screen
treatment by time-of-testing interaction, such that the
difference between immediate and delayed posttest means will
be significantly less for both Related Humor and Unrelated
Humor groups compared to the No Humor group; and for the
Related Humor group compared to the Unrelated Humor
5. Is there a review screen treatment effect for attitudes regarding
the CBI lesson and the content?
Research Hypothesis: There will be a significant overall
multivariate treatment effect for the attitude survey; and
univariate analyses of variance will indicate significant
treatment effects on all four subscales, such that means will be
significantly higher for both Related Humor and Unrelated
Humor groups compared to the No Humor group; and for the
Related Humor group compared to the Unrelated Humor
Additional Analyses
Supplemental data gathered independent of treatment were used
to identify preexisting differences among groups that might warrant the
introduction of a covariate or factor. Possible modifications to the
planned design through addition of independent variables were based
upon preliminary analysis. The supplemental data set included Gender,
Age Range (20s, 30s, 40s, 50s), aptitude measures, and an experimenter-
designed instrument to measure Goal Orientation and Review Potential.
The aptitude measures chosen were common to all the teacher education
students in the study. Either the Program for Licensing Assessments for

Colorado Educators (PLACE) Basic Skills Test or the California
Aptitude Test (CAT) was required as part of admission procedures at
both institutions. A total of 48 subjects (16 from each treatment group)
granted permission for use of their PLACE Basic Skills records,
consisting of score for three subscales: Reading, Math, and Writing.
Some of the subjects had taken the CAT instead of the PLACE, and
consented to use of these records; however, there were too few subjects
in this situation (8 No Humor, 7 Related Humor, and 9 Unrelated
Humor subjects) so CAT scores were not used.
A Goal Orientation instrument was included as a pre-measure in
the hope that it would inform analyses of the amount of reviewing
conducted by all subjects. The measure was based on the work of Ng and
Bereiter (1991) mentioned in Chapter 2. In that study of 16 adults
learning BASIC programming, results identified roughly one-third of the
subjects as task oriented, one-third as learning oriented, and one-third
in-between (instruction oriented). Subjects in the previous study picked
sentences that contained verb cues related to a task or a learning
orientation. With the assistance of both Ng and Bereiter, a simple
measure was designed for the present study that used just the verb pairs
from their sentences in the form of semantic differential items. The verb
pairs were judged to have face validity in the previous study. The
instrument was administered prior to the instructional sequence in the
CBI, and consisted of 9 paired phrases (separated by an unbroken line)

that completed the statement, "Generally, I prefer to..." For example, one
item from the measure was:
Complete Find out
In this example, the verb complete would be a cue for task orientation;
find out would be a cue for learning orientation.
In CBI terms Review Potential might be thought of as "learner
control momentum." The reason for including the Review Potential
item was to gain information prior to treatments on a subject's feelings
about moving ahead or going back to review while studying. This
tendency was measured with a single semantic differential item at the
end of the CBI-administered Goal Orientation instrument (as a tenth
item). This item was judged to have face validity; however, since it
stands alone, it has no reliability. The itemdepicted belowalso used
the stem, "Generally, I prefer to..."
Press on Review
The ten-item pre-measure for Goal Orientation and Review
Potential is included as Appendix A. Unfortunately pilot testing of the
instrument containing the Goal Orientation and Review Potential items
did not take place before the experiment due to scheduling problems.
Details of the performance of the instrument during the actual
experiment are provided in Chapter 4. Of the individual difference
variables examined in preliminary analyses, only one subscale of the

PLACE Basic Skills (Reading) was considered as a covariate for analyses
of achievement, and only Review Potential was considered as a factor
during analysis of Voluntary Review Choices. The outcome of these
preliminary analyses and the impact on statistical results related to the
research questions are discussed in Chapter 4.
In addition to supporting statistical analyses of review behaviors,
enroute data were also used for descriptive statistics. The identification
of which review loops were accessed, combined with the recording of
practice and posttest item outcomes, allowed for reporting of review
efficiency for the sample and comparison across treatments. This
entailed informal investigation into two areas of interest: (a) how often
reviewing occurred after a correct practice item response versus how
often reviewing occurred after an incorrect practice item response, and
(b) the relationship between the outcome of a practice item response and
the outcome of an associated posttest item response when reviewing
occurred. The former analysis simply involved an accounting of the
correctness of the practice item responses for each review. The latter
analysis involved the coding of practice-posttest item outcome pairs for
each review episode, according to the following scheme:
Wrong Wrong Incorrect practice response-incorrect posttest response
RightWrong Correct practice responseIncorrect posttest response
WrongRight Incorrect practice responseCorrect posttest response
RightRight Correct practice responseCorrect posttest response
The results of preliminary and additional analyses using the
techniques described in this section are presented in Chapter 4.

The learning material used in the study was adapted from a self-
paced CBI lesson on instructional objectives used in a previous study
(Morrison et al., 1995). The instructional unit contained approximately
3500 words, and was divided into five sections with the following topics:
(a) Aims, goals, and objectives, (b) Observable verbs, (c) Specifying
conditions, (d) Specifying criteria, and (e) Time.
Content Validity and Formative Evaluation
The researchers who conducted the previous study converted
lecture and print material from a regular course unit on instructional
objectives to a CBI program with only minor content modifications. The
content was based on the objectives specified for a required introductory
course in teacher education on basic applications of technology. Prior to
the previous study, three instructors who lecture on behavioral
objectives judged the CBI unit to be an appropriate replacement unit for
teacher education students (Morrison et al., 1995). For the current study,
the content validity was judged by the instructors of the four
participating classes, and was again deemed to be valid and appropriate
for a similar sample of subjects. Acting upon the advice of one
instructor, the clarity of some examples used in the original instruction
was improved.
Prior to the experiment, the lesson materials, practice questions,
and immediate posttest instrument supplied by Morrison and his

colleagues were incorporated into a HyperCard-based shell, along with
an experimenter-designed attitude survey. During June, 1995, formative
evaluation (FE) of the basic CBI configuration took place. Participants
included a content expert, a CBI usability expert, and three volunteer
learners (one for each review screen treatment in the design).
Modifications to the CBI were made according to the results of the FE
(i.e., comments of the two experts and the three learners), with particular
attention paid to the content-relatedness (or -unrelatedness) of the
review screen humor elements. Later that month and extending into
July, 1995, a pilot of the experimental materials and instruments took
place. The 25 volunteer learners were drawn from a population of
education students attending courses during the Summer semester (i.e.,
similar subjects but not from the full-scale study sample). The results of
the pilot are discussed under the Dependent Variables section below.
CBI Features
The initial section of the CBI program contained tutorial
instructions for using the program. The assumption was made that
providing practice in the use of a mouse tool was not necessary; and
indeed, all subjects in both the pilot and the full-scale study were
familiar with this device. The tutorial screens included: (a) explanation
that navigation was via HyperCard buttons, and that a Help button
(always available) would most likely answer their questions during the
session (see the second screen in Appendix B); and (b) a screen that

provided coaching on the importance of reviewing and a "hint"
regarding the presence of humor in the review loops for the Related and
Unrelated Humor treatments. The basic coaching statement (same for
all treatments) was as follows:
This lesson includes "Review Loops" that allow you to return to
the page of the instruction that relates to a Quiz Question.
Whether you answer a Quiz Question correctly or incorrectly, you
are encouraged to review for a second exposure to specific
information (which is similar to the way you would use a
This coaching statement and the flow diagram that represented
the review as taking the learner back to previously-viewed material are
shown as they appeared to the No Humor treatment in the third screen
of Appendix B. This indication of the nature of the review loop was of
crucial importance to the experiment, for this was the "clue" given to the
subjects as to whether or not their lesson contained review screen
humor elements. Thus, for the Related Humor and Unrelated Humor
treatments, the following sentence was added to the end of the above
basic coaching statement: "Note: each review loop contains a little
humor." Also, the flow diagram graphically depicted humor with a jack-
in-the-box replacing the magnifying glass in the review loop of the No
Humor treatment. It turned out that these clues may not have been
strong enough during the experiment, even though they seemed to
work well during formative evaluation and pilot testing. The resulting
lack of reviewing resulted in the modification requiring review for the
first practice question discussed previously.

Embedded within the instruction were 16 multiple-choice practice
questions on the material covered in immediately-preceding sections.
The original practice questions were designed by Morrison et al. (1995) to
assess learning at the comprehension level by changing the phrasing of
the original text or going beyond the specific text information to address
ideas not explicitly stated (e.g., classifying a new example as a particular
concept). The number of questions per section varied from one to five,
depending on the amount and type of content covered. After the subject
responded to a practice question, knowledge of correct response (KCR)
feedback was presented, followed by an opportunity to either review the
corresponding lesson materialregardless of the correctness of the
responseor move on to the next practice question or new material.
Copies of all CBI screens used for all treatments (basic linear instruction
and practice screen sequence) are included as Appendix B.
When the subjects in the No Humor treatment chose to review,
they were routed back to a screen or multiple screens (as many as three)
that covered the material pertaining to the practice question. Subjects
could not leave the review loop until all review screens were accessed.
Subjects in the Related Humor and Unrelated treatments were routed
first to a screen containing humor, then forced to continue to the same
review screens that the No Humor treatment would experience.
Copyright restrictions preclude the publication of the actual humor
screens used in the Related and Unrelated Humor review screen

treatments; however, a description of each humor element along with
its associated practice item and content "topic" is found in Appendix C.
Feedback Type
The Knowledge of Correct Response (KCR) feedback design was
based on previous studies by Clariana et al. (1991), Dempsey (1988), and
Morrison et al. (1995). An on-screen message informed the learner of the
correct answer after each response (see Appendix B, Quiz Number 2 after
the first content section). A correct response resulted in knowledge of
results (KOR) with display of the word "CORRECT." An incorrect
response resulted in KOR with the message "SORRY" along with display
of the correct response (KCR). The original question and the answer
choices remained in view for both correct and incorrect responses with a
"check mark" indicating the subject's correct response, or an "X"
indicating the subject's incorrect response. After either a correct or an
incorrect response and the associated feedback, a prompt along with
"Review" and "Move on" buttons provided the option to either review
the preceding material or move on to the next question (or the next
section in the case of the last practice question). After review, the subject
was branched to the next question or the next section of content, as
applicable, when either the "Go To Next Quiz Question" or "Go To Next
Section," button (each available only during review) was activated.

Humor Treatments
Three levels were incorporated into the design of review screen
treatmentsa control with no humor, humor related to content, and
humor not related to content. As mentioned above the control (No
Humor) treatment presented the original screen(s) of material associated
with a practice question when the review option was selected for all
feedback treatments. The humor treatment strategy was designed to
facilitate the differentiation of time spent on humorous review screen
elements versus time spent on serious review screen material. This was
accomplished by placing either the content-related or -unrelated
humorous element on a separate screen just before the beginning of the
serious review material.
The content-related humor treatment provided visual humor in
the form of a cartoon, or in one case textual humor in the forin of a
funny scenarioMager's well-known shark fable (Mager, 1984)as an
initial screen when the review option was selected. This humorous
element was designed to be relevant to the content of the review
material. It provided a humorous example or non-example, an alternate
perspective, an anchor in the form of a humorous real-world situation,
or a play-on-words using key phrases from the original text. A prompt
on the initial humor-containing screen provided the subject with
instructions to continue to the next screen (containing original lesson
material). The content-unrelated humor treatment also provided a
cartoon as an initial screen when the review option was selected;

however, this "tacked on" humorous element was unrelated to the
content of the review material. The same prompt that was used with
content-related humor screens provided the subject with instructions to
continue to the next screen (containing original lesson material).
Dependent Variables
Measures for this study included process parameters (enroute
tracking of which specific review screens were viewed and time spent
viewing each review screen), an immediate and delayed posttest, and an
attitude survey. All data except for the paper-and-pencil delayed posttest
answers were automatically collected by the CBI program.
Enroute Data
The CBI included tracking routines for the recording of which
review screens were viewed and the amount of time spent per review
screen. These data were primarily used to analyze the dependent
variable, Voluntary Review Choices. The program also recorded
responses to practice items, providing additional information that was
used for descriptive statistics concerning review efficiency. This
information included: (a) number of reviews after correct and incorrect
practice responses, and (b) the relationship between the outcome of a
practice item response and the outcome of an associated posttest item
when reviewing occurred. Also, average time per screen for serious and
humor review screens was analyzed; however, these data were subject to

possible contamination by off-task behavior during the review loop.
This problem was handled by screening the data for outliers (see Chapter
4). Enroute data were recorded directly on computer disk while subjects
were working on the instructional unit.
The posttest was a 16-item instrument designed to measure
achievement on the content of the CBI program. In order to investigate
the impact of review screen treatments on both immediate and delayed
achievement, the posttest was administered as a repeated measure. The
immediate posttest occurred at the end of the CBI session, and was
administered electronically. The delayed posttest was a paper-and-pencil
version of the immediate posttest (identical questions), and was
administered two weeks after completion of the CBI. The posttest items
for the current study were used in the previous study (Morrison et al.,
1995), and were all designed as paraphrased-transformed versions of the
16 embedded practice questions, thus providing some measure of
comprehension. The following example excerpted from Morrison et al.
(1995) illustrates a paraphrased-transformed posttest item (note that the
practice question is paraphrased by substituting the word "show" for the
word "demonstrate," and transformed by reversing the stem and the

[Original practice question] Which of the following is an "observable"
verb for use in an objective?
a. Demonstrate (correct answer)
b. See
c. Determine
d. All of the above
[Paraphrased-transformed posttest item] What would the verb
"show" be considered when used in an objective?
a. Observable (correct answer)
b. Non-observable
c. Passive
d. None of the above
A side-by-side comparison of the 16 practice questions and their
associated paraphrased-transformed posttest item is included as
Appendix D. Also, the screens used for the electronically-administered
immediate posttest are shown in Appendix E, and the paper-and-pencil
delayed posttest is included as Appendix F.
Pilot data for the set of practice items and posttest items were
tested using the Item and Test Analysis Module (ITAM) of the
MicroCAT Testing System (Assessment Systems Incorporated, 1995).
Internal consistency results using Cronbach's alpha were .23 for the 16
practice items and .34 for the 16 posttest items. Also, the ITAM indicated
that several structural problems existed for these instruments.
Subsequently, the practice and posttest instruments were extensively
checked for content validity, consistency, and weaknesses in test design
by two of the professors from the classes participating in the study. A few
changes were made to items that contained statements inconsistent with

the content. Also several format flawssuch as lengthy correct response
compared to detractorswere corrected. Cronbach's alpha improved to
.57, .49, and .52 for the practice, immediate posttest, and delayed posttest,
respectively, for the actual experiment.
An attitude survey consisting of two semantic differential scales
was designed to assess subjects' reactions to the CBI instruction and the
content (instructional objectives). Attitude Scale A consisted of 14 paired
phrases (separated by an unbroken line) that completed the statement, "I
feel that, overall, this computer-based lesson was..." Seven items from
Attitude Scale A were designed as a subscale that had face validity for
attitudes related to the CBI being fun. The other seven items in Scale A
were designed to similarly assess attitudes regarding the design of the
CBI. The examples below are from Subscale Fun (top) and Subscale
Design (bottom), respectively.
Drudgery 1 Fun 1
Well- Poorly-
designed designed
Similarly, Attitude Scale B consisted of 10 paired phrases that
completed the statement, "As a result of this lesson, I tend to view

instructional objectives as..." Five items from Attitude Scale B were
designed as a subscale that had face validity for attitudes related to the
content being interesting. The other seven items of Scale B were
designed to similarly assess attitudes regarding the relevancy of the
content. The examples below are from Subscale Interest (top) and
Subscale Relevancy (bottom), respectively.
Boring Interesting
Relevant Irrelevant
The attitude survey subscales were administered electronically to
each subject after completion of the instruction and before the
immediate posttest. Subjects were asked to move a "slider" along on a
four-inch continuous line on the screen similar to the above examples
for each attitude item. The score for each item, based on the position of
the slider ranged from -5.0 to +5.0 with negative scores representing a
negative attitude, zero indicating no opinion, and positive scores
representing a positive attitude. The attitude survey screens are included
as Appendix G.
The empirically-derived items for the attitude subscales were
reviewed for content validity by two faculty members, and determined to
be satisfactory. Internal consistency reliability was assessed for the
attitude survey prior to the experiment using Cronbach's alpha formula.

Data from the administration of the survey to the 25 subjects during the
pilot resulted in coefficients of .78 for Scale A (14 items related to the CBI
lesson) and .87 for Scale B (10 items related to the content). Analysis of
the four subscales resulted in coefficients of: .60 for the Fun subscale, .65
for the Design subscale, .72 for the Interest subscale, and .88 for the
Relevancy subscale.
An open response item was added to the CBI for the two humor
treatments as a method for gathering supporting data on the effects of
review screen humor elements. Of particular interest were subject
comments regarding motivational aspects (e.g., did the review screen
humor cause subjects to review more or less?) and cognitive effects (e.g.,
did the humor cause subjects to reflect, did it assist in recall of
information, or did it cause an unwanted distraction?). The actual open
response screen that solicited a typed-in response from the subjects in the
Related Humor and Unrelated Humor treatments is provided as
Appendix H. Chapters 4 and 5 show that the open response measure
proved to be a useful source of information for the analysis of attitudinal
data and review behaviors. All attitude data were recorded automatically
by the CBI on computer disk.
Data analyzed from the pool of 92 subjects making up the
experimental sample were gathered using similar environmental
conditions and procedures. Subjects were all randomly assigned to one

of three review screen treatments (No Humor, Related Humor, and
Unrelated Humor). For Classes 1 and 2, groups of 7 to 16 subjects at a
time reported to a teaching laboratory that contained 16 identical
monochrome Macintosh LC II computers. The same laboratory was used
for all of Class 4, except that eight students occupied the same type of
computer in an adjoining section of the same computer laboratory. The
administration of the CBI to Class 3 took place in another, very similar,
Macintosh computer laboratory. All laboratories except for the adjoining
laboratory section that the eight students in Class 4 utilized were
reserved for the experiment and scheduled for a time when they were
otherwise unoccupied. Little impact was evident for the eight students
in Class 3 who overflowed into an adjoining section. An equal number
of nondescript, unlabeled 3 1/2-inch computer disks for each treatment
were randomly placed and thoroughly shuffled on a table in the
laboratory for each session. After subjects entered the laboratory, they
randomly selected one of the disks, and each subject was seated at an
individual computer station.
Standard verbal instructions described the general nature of the
task (a unit on instructional objectives), and the procedures .for
responding (unlimited time, closed book, individual work). The
standard instructions included a reminder that the subjects were
participating in an experiment and that different kinds of activities could
be going on around the room. This statement was intended to mitigate
possible confusion that a No Humor treatment subject may experience

due to overt responses from subjects participating in Related Humor or
Unrelated Humor treatments. A proctor remained in the lab, answering
navigational questions, quelling disruptions, and monitoring progress
until all subjects finished the instructional unit.
Lesson presentation, KCR feedback, review screen treatment,
administration of the attitude survey and immediate posttest, and
enroute data-gathering Were accomplished automatically by the CBI
program, and controlled by subject response to prompts. Contact time for
participating in the brief tutorial, reading instructional material,
answering practice questions, reviewing, and completing attitude and
posttest instruments averaged approximately one hour. Only data sets
for subjects who complete all tasks in one continuous session were

The following sections provide the results of quantitative and
qualitative analyses of data obtained during the procedure described in
Chapter 3. Specifically, this chapter presents: descriptive statistics,
exploratory analyses; inferential statistics that address the research
questions; and summaries of open-response data. It should be noted that
most of the results presented in this chapter involve the comparison of
parameters across the three humor treatments. For the sake of
consistency and to facilitate comparison of tables and figures the data are
presented with treatment on the vertical (i.e., the ordinate of a graph or
the rows of a table), and placed from top-to-bottom in the following
order: No Humor treatment, Related Humor treatment, then Unrelated
Humor. It is hoped that the similarity of this scheme to the research
design tables in Chapter 3 will enhance the continuity of the reporting of
results. Another presentation feature for this chapter (carried over from
Chapter 3 and intended to improve the connection between text and
graphics) is the capitalization of all formal independent and dependent
variable names used in the analyses (e.g., Posttest).

Descriptive Statistics
A total of 110 subjects participated in the study. However, as
discussed in Chapter 3, the final experimental sample resulted from
adjustments that involved the removal of data gathered for 18 subjects.
This involved exclusion of data from the 16 subjects in the first two
classes who did not review at all during the CBI session, data from the
subject who did not complete the CBI session, and data lost to a computer
disk malfunction for one subject. These adjustments resulted in a total
data set representing the subject pool of 92. Tables 4.1 and 4.2 show
thateven after these adjustmentsthe random assignment of subjects
to treatments resulted in a fairly even distribution by gender and age.
However, examination of a cross-tabulation of these subject variables in
Table 4.3 indicates that the sample consisted, primarily, of females in
their twenties (55.4%).
Female Male Totals;
No Humor 24 7
Related 23 7
Unrelated 23 8
Totals: 70 22
Table 4.1 Observed Frequency for Gender

Twenties Thirties Forties Fifties
No Humor 23 4 3 1
Related 23 4 3 0
Unrelated 21 7 3 0
Totals: 67 15 9 1
Table 4.2 Observed Frequency for Age Range
Twenties Thirties Forties Fifties
Female 51 13 5 1
Male 16 2 4 0
Totals: 67 15 9 1
Table 4.3 Observed Frequency for Gender vs. Age Range
The lack of variability in gender and age was expected for a sample
drawn from a population of first-year teacher education students. The
data are provided here for descriptive purposes onlyinadequate cell
sizes precluded the use of gender or age as factors in the statistical
analyses described below.
Preliminary Analyses
In addition to assessment of the above demographic data,
exploratory analyses were conducted to investigate the existence of other
group differences prior to treatments. This section describes the

investigation into two areas: homogeneity of subjects, and the potential
for use of individual difference data as covariates or factors.
Homogeneity of Subjects
Chapter 3 described the source of subjects as four separate classes of
teacher education students from two institutions, the University of
Colorado at Denver (UCD) and Metropolitan State College of Denver
(MSCD). Factorial 4 (class) x 3 (humor treatment) ANOVA analyses were
conducted for each of the main dependent variables in the study to
investigate the similarity (or dissimilarity) between classes. Means and
ANOVA tables presented in Appendix I indicate that there were no
significant main effects for the class variable, or interactions between
class and treatment for any of the achievement, attitude, or enroute
dependent variables. This outcome had a favorable effect on the
generalizability of the findings from the current study, and helped to
confirm the qualitative assessment in Chapter 3 that the subject pool
used in the study was fairly homogeneous.
Potential Covariates and Factors
Appendix J provides the complete results of correlations between
interval-level individual difference variables and dependent variables
for the study. Potential covariates included two pre-measures (Goal
Orientation and Review Potential) that were administered prior to the
treatments as part of the CBI session, and aptitude data gathered for
subjects consenting to its collection. Since all research questions deal

with the effects of the three levels of humor treatment, the correlations
were conducted separately for each treatment group. The applicability of
a variable as a covariate was determined by the magnitude of the
correlation with a particular dependent variable, and in the homogeneity
of correlation with a particular dependent variable across treatments (the
same sign and roughly the same magnitude). Two individual difference
variables (Goal Orientation and Review Potential) were converted to
nominal-level for consideration as factors. The acceptability of a variable
as a factor was determined by cell sizes. The following sections describe
the results of the preliminary analyses related to individual difference
Goal Orientation. As described in Chapter 3, this variable was
included as a pre-measure in the hope that it would provide a means for
explaining part of the variability expected in the amount of reviewing
conducted by all subjects. An internal consistency reliability analysis was
conducted for the 9-item semantic differential Goal Orientation
instrument, resulting in a Cronbach's alpha coefficient of .79. Figure 4.1
is a scatterplot of the interval-level means from the instrument, with a
mean response of -5 indicating an extreme task orientation, a mean
response of 0 indicating an undecided position, and mean response of +5
indicating an extreme learning orientation. Consistent with the vertical
ordering scheme mentioned at the start of this chapter, treatments were
coded NH for No Humor, RH for Related Humor, and UH for Unrelated
Humor. These data indicated that most subjects identified (in varying

degrees) with a learning orientation according to their responses to the
verb pairs used by Ng and Bereiter (1991). Correlations by treatment for
interval-level Goal Orientation data with dependent variables did not
yield sufficient magnitudes or homogeneity to justify its use as a
covariate (see Appendix J).
Figure 4.1 Treatment vs. Mean Goal Orientation Response
The use of Goal Orientation data as a factor was investigated after
conversion to nominal-level. Goal Orientation means indicating a clear
preference (-.1 > mean > .1) were stratified into a "Task Orientation"
group and a "Learning Orientation group." One subject elected to skip
over the Goal Orientation instrument and was remove from the data set,
as were five subjects whose means did not indicate a clear preference.

Table 4.4 indicates very small cell sizes for the remaining 86 means that
precluded use of Goal Orientation as a factor.
No Humor
Task Learning
5 23
1 28
2 27
T otals:
Totals: 8 78 86
Table 4.4 Observed Frequency for I Goal Orientation I > .1
The result of the preliminary analysis of Goal Orientation for the
current study's population of 92 teacher education students is dissimilar
to the trend reported in the previous Ng and Bereiter (1991) study of 16
adults learning BASIC programming. In that study Goal Orientation was
distributed more evenly, with approximately one-third of the subjects
indicating a task orientation, one-third a learning orientation, and one-
third in-between. The difference in the two studies may be due to a
number of factors, including the small sample size used in the previous
study, or inter-dependence between the verb pairs used to identify Goal
Orientation and the content being taught. Possibly the writing of BASIC
programs is identified more with production of a working product than
is the writing of textual instructional objectives. Perhaps different verb
pairs should have been designed and piloted for the current study.
Nevertheless, the failure to pilot and possibly adjust the Goal

Orientation instrument resulted in data unusable as an interval-level
covariate (due to small and/or non-homogeneous correlations with the
dependent variables), or as a nominal-level factor (due to a lack of
variability and small cell sizes).
Review Potential. As discussed in Chapter 3, the pre-treatment
measure of a subject's proclivity toward reviewing was of interest as an
indication of his/her "learner control momentum." Review potential
was measured with a single semantic differential item; thus, reliability
could not be assessed. Figure 4.2 is a scatterplot of the interval-level
response for the item, with -5 indicating an extreme tendency to "press-
on" without reviewing, 0 indicating an undecided response, and +5
indicating an extreme predisposition toward "going back" to review. The
results indicated fairly good variability for the sample, with most subjects
indicating (in varying degrees) that they generally prefer to "press-on."
Correlations by treatment for interval-level Review Potential data with
dependent variables did not yield sufficient magnitudes or homogeneity
to justify its use as a covariate for the control of non-treatment
differences in review behavior (see Appendix J). The lack of reliability
data for this variable is a problem, and future studies of potential
individual differences in this area should include the development of a
better instrument.

NH K> ooo ogs> O 00 O <2> 00030
o <
ocsa> o goo o o o 6 o o ooo go o cso
Figure 4.2 Treatment vs. Review Potential Response
The use of the Review Potential data as a factor was investigated
via conversion to nominal-level. Review Potential item responses
indicating a clear preference (-.1 > response > .1) were stratified into a
"Press-on" group and a "Review" group. One subject elected to skip over
the item and was removed from the data set, as were 11 subjects whose
means did not indicate a clear preference. Table 4.5 provides the cell
sizes for the remaining 80 responses. These data indicated the potential
for use of Review Potential as a factor (with cautions noted due to the
lack of reliability data) in the analysis of review behaviors. However, as
discussed below, the factor did not impact the significance outcome of
any analyses.

Press on
No Humor 22 6
Related 15 11
Unrelated 19 7
Totals: 56 24
Table 4.5 Review Potential Observed Frequency
Aptitude. The consent form signed by all participants in the
experiment included a request for permission to use either Program for
Licensing Assessments for Colorado Educators (PLACE) Basic Skills
Test or California Aptitude Test (CAT) scores in the study. A total of
48 subjects (16 from each treatment group) granted permission for use of
their PLACE Basic Skills records, consisting of scores for three subscales:
Reading, Math, and Writing. As mentioned in Chapter 3, there were too
few subjects with CAT scores to use the data in any analyses.
Inspection of the tables in Appendix J indicated that none of the PLACE
aptitude variables showed promise as a covariate for any of the
dependent variables.
Enroute Data
As discussed in Chapter 3, a modification to the CBI was made at
the midpoint of the experiment (between the group composed of Classes
1 and 2 and the group composed of Classes 3 and 4). This modification,
which forced review for the first practice item, insured that all

treatments provided the subject with a chance to become familiar with
the nature of the review loop. After removal of 16 subjects who did not
review at all during sessions before the CBI adjustment, a review of
enroute data for the remaining 46 subjects in the first two classes
indicated that the number of Voluntary Review Choices ranged from 1
to 15. For the 46 subjects in the two classes experiencing the adjusted
method, the number of Voluntary Review Choices ranged from 0 to 15
(not counting the initial forced review). Figure 4.3 provides a scatterplot
of the number of Voluntary Review Choices by treatment. Table 4.6
shows that the mean number of Voluntary Review Choices for both the
Related Humor and Unrelated Humor treatments surpassed that for the
No Humor condition, with the Related Humor mean higher than that
for Unrelated Humor.
0 2 4 6 8 10 12 14 16
Voluntary Review Choices
Figure 4.3 Treatment vs. Voluntary Review Choices

Group: Count: 'dean: Std. Dev.: Std. Error:
No Humor 31 2.419 1.876 .337
Related 30 4.367 4.902 .895
Unrelated 31 3.581 4.081 .733
Table 4.6 Means for Voluntary Review Choices
Research Question 1
Is there a treatment main effect for number of Voluntary Review
Choices? Answering this question required evaluation of the methods
employed when the enroute data were gathered. Since all subjects in the
first two classes experiencing the CBI design before the modification
conducted at least one voluntary review in the first 20 percent of the
lesson, it was initially determined that an adjustment for number of
opportunities for voluntary review (16 before the CBI modification, 15
after the CBI modification) was probably not necessary. However, since
the normalization for the 6.25% difference in number of voluntary
review opportunities before and after the modification was not
complicated (performed simply by analyzing the fraction Voluntary
Review Choices/Opportunities), the result of the ANOVA for Voluntary
Review Choices is based on means for this ratio presented in Table 4.7.
The results of the oneway ANOVA shown in Table 4.8 indicate that
there was no significant difference between treatments for enroute
behavior in terms of the Voluntary Review Choices/Opportunities
(F = 2.04, p = .137).

Group: Count: 'dean: Std. Dev.: Std. Error:
No Humor 31 .156 .122 .022
Related 30 .287 .328 .06
Unrelated 31 .233 .271 .049
Table 4.7 Means for Voluntary Review Choices/Opportunities
Source: DF: Sum Squares: 'dean Square: :-test:
Between qroups 2 .264 .132 2.035
Within qroups 89 5.774 .065 p = .1367
Total 91 6.038
Table 4.8 Oneway ANOVA for Voluntary Review
Choices / Opportunities
Review Potential as a Factor
As mentioned in the section on preliminary analyses, the
categorization of the interval level results from the Review Potential
item resulted in fairly good variability and cell sizes for the sample. In
an attempt to improve power for the analysis of enroute data, the means
of Table 4.9 were tested via a factorial 3 (Treatment) x 2 (Review
Potential) ANOVA for Voluntary Review Choices/Opportunities. Table
4.10 shows that analysis did not yield significant main effects or
interactions. Thus, the addition of this factor did not change the
significance outcome for the above oneway ANOVA that addressed
Research Question 1.

Review-Press on Press on Review Totals:
No Humor 22 6 28
C .159 .193 .166
£ 15 11 26
1- Unrelated 19 7 .334 26
.231 .259
Totals: 56 24 80
.237 .254 .243
Table 4.9 Voluntary Review Choices/Opportunities Means
vs. Treatment and Review Potential
Source: df: Sum of Squares: dean Square: F-test: P value:
Treatment (A) 2 .182 .091 1.273 .286
Review-Press on (B) 1 3.237E-4 3.237E-4 .005 .9465
AB 2 .158 .079 1.106 .3362
Error 74 5.288 .071
NOTE: No missing cells found. 12 cases deleted with missing values.
Table 4.10 Factorial (Treatment x Review Potential) ANOVA for
Voluntary Review Choices/Opportunities
Achievement Data
Figure 4.4 shows that test scores for all subjects (number correct
out a possible 16) ranged from 3 to 14 for the Posttest administered at the
end of the CBI lesson; and from 3 to 13 for the paper and pencil Delayed
Posttest (with items identical to the immediate posttest) administered
two weeks after the CBI session. All 92 subjects in the sample completed
the Posttest; however, only 81 subjects were available at the two-week
mark for the Delayed Posttest (make-up sessions were not conducted).

O Posttest
Delayed Posttest


-i----1----1----1----1----1----1----1----1----1-----1---1 ---1
2 4 6 8 10 12 14
Number Correct for Posttest and Delayed Posttest
Figure 4.4 Treatment vs. Posttest and Delayed Posttest Achievement
Table 4.11 indicates that the performance for subjects in all
treatments was similar, with a mean score for the sample of 9.522 correct
out of 16 possible (59.5%) on the Posttest, and 7.827 correct out of 16
possible (48.9%) on the Delayed Posttest. Thus, mastery of the subject
matter was not achieved. In fact, using a traditional grading scale, mean
scores for the sample represent failing grades for both measures (< 60%).
Measure Posttest Delayed
Treatment No Humor 31 9.161 29 7.931
Related 30 9.667 25 7.72
Unrelated 31 9.742 27 7.815
Totals: 92 9.522 81 7.827
Table 4.11 Summary for Posttest and Delayed Posttest Achievement

Research hypotheses pertaining to the effects of review screen
treatment on achievement were addressed with a repeated measures
(two within, one between factor) ANOVA conducted for the means of
the sample (N = 81) shown in Table 4.12.
Effect: Treatment
Dependent: Achievement Tests Count Mean Std. Dev. Std. Error
No Humor 58 8.552 2.767 .363
Related 50 8.760 2.512 .355
Unrelated 54 8.815 2.713 .369
Effect: Achievement Tests Treatment Dependent: Achievement Tests Count Mean Std. Dev. Std. Error
Posttest, No Humor 29 9.172 2.778 .516
Posttest, Related 25 9.800 2.062 .412
Posttest, Unrelated 27 9.815 2.543 .489
Delayed Posttest, No Humor 29 7.931 2.658 .494
Delayed Posttest, Related 25 7.720 2.525 .505
Delayed Posttest, Unrelated 27 7.815 2.543 .489
Table 4.12 Means for Achievement Repeated Measures ANOVA
Research Question 2
Is there a treatment main effect for combined immediate and
delayed achievement? The results of a repeated measures ANOVA
shown in Table 4.13 indicate that there was no significant treatment
effect (F = .111, p = .8948) for combined immediate and delayed posttest
scores among groups exposed to the three different review screen

Research Question 3
Is there a time (i.e., time-of-testing) main effect? Table 4.13 shows
that there was a significant difference (F = 39.374, p < .0001) between
combined immediate posttest scores and combined delayed posttest
scores. The means presented in Table 4.12 indicate the direction of the
time main effect, with test scores after a two week period (Delayed
Posttest) significantly less that the immediate (Posttest) scores.
Research Question 4
Is there a treatment by time interaction? Table 4.13 indicates that
there was no significant difference (F = .942, p = .3942) in the change in
test scores over time among groups exposed to the three different review
screen treatments.
Source: df: Sum of Squares: 'dean Square: :-test: 3 value:
Treatment (A) 2 2.165 1.082 .111 .8948
subjects w. groups 78 758.613 9.726
Repeated Measure (B) 1 124.469 124.469 39.374 .0001
AB 2 5.956 2.978 .942 .3942
B x subjects w. groups 78 246.575 3.161
NOTE: No missing cells found. 11 cases deleted with missing values.
Table 4.13 Achievement Repeated Measures ANOVA
Attitude Data
Immediately after the end of the CBI instructional sequence, and
before the posttest, attitudes were measured with two semantic
differential scales (each consisting of two subscales), and an open

response item that solicited comments from subjects in the Related
Humor and Unrelated Humor treatments. This section presents the
results of quantitative and qualitative analyses of these attitude data.
Attitude Subscales
As discussed in Chapter 3, Attitude Scale A was constructed to
assess subject attitudes regarding the CBI lesson format. Internal
consistency reliability analysis conducted for the two 7-item semantic
differential subscales for Attitude Scale A compared favorably with data
from the pilot study. Cronbach's alpha coefficients of .87 and .81 were
obtained for the Fun and Design subscales, respectively. Attitude Scale B
was constructed to assess subject attitudes regarding the lesson content
(Instructional Objectives). Internal consistency reliability analysis
conducted for the two 5-item semantic differential subscales for Attitude
Scale B also compared favorably with pilot data. Cronbach's alpha
coefficients of .83 and .94 were obtained for the Interest and Relevancy
subscales, respectively.
Figures 4.5, 4.6, 4.7, and 4.8 present scatterplots of the interval-level
means for the Fun, Design, Interest, and Relevancy attitude subscale
instruments, respectively. For all four subscale measures a mean
response of -5 indicated an extreme negative attitude, a mean response
of 0 corresponded to an undecided stance, and a mean response of +5
indicated an extreme positive attitude.

Treatment Treatment
aa> ooo o
2 3 4 5
Figure 4.5 Treatment vs. Mean Fun Attitude Subscale Response
Design Subscale
Figure 4.6 Treatment vs. Mean Design Attitude Subscale Response

Interest Subscale
Figure 4.7 Treatment vs. Mean Interest Attitude Subscale Response
oo ooo ocoooocdooo<-
I | I | I | I | I | I | I | I | I | I
-5 -4 -3 -2 -1 012 3 4 5
Relevancy Subscale
Figure 4.8 Treatment vs. Mean Relevancy Attitude Subscale Response
Analysis of the attitude instrument involved the rejection of two
cases because of missing data. A subject in the Unrelated Humor

treatment elected to skip over Attitude Scale A (Fun and Design
subscales) measure, and a subject in the Related Humor treatment chose
to skip Attitude Scale B (Interest and Relevancy subscales). An analysis
of mean responses (see Table 4.14) from the four attitude subscales for the
remaining 90 subjects was analyzed using a oneway MANOVA.
Fun Subscale
Count Mean Std. Dev. Std.Error
No Humor 31 .810 1.168 .210
Related 29 1.859 1.123 .208
Unrelated 30 .815 1.355 .247
Count Design Subscale Mean Std. Dev. Std. Error
No Humor 31 1.293 1.479 .266
Related 29 1.782 1.262 .234
Unrelated 30 1.437 1.112 .203
Count Interest Subscale Mean Std. Dev. Std.Error
No Humor 31 1.460 1.856 .333
Related 29 1.963 1.241 .230
Unrelated 30 1.609 1.256 .229
Count Relevancy Subscale Mean Std. Dev. Std. Error
No Humor 31 2.436 1.603 .288
Related 29 2.889 1.267 .235
Unrelated 30 2.671 1.655 .302
Table 4.14 Means Tables for Treatment vs. Attitude Subscales

Research Question 5
Is there a review screen treatment effect for attitudes regarding the
CBI lesson and the content? Table 4.15 indicates that the oneway
MANOVA resulted in overall significance for the treatment effect
(Wilks' F = 2.13, p < .05). Also shown in Table 4.15 are the results of
follow-up univariate analyses that identify the Fun subscale as the main
contributor to the multivariate F statistic with a significant treatment
effect (F = 7.24, p < .001). Univariate results for the Design, Interest, and
Relevancy subscales failed to yield any significant effects for treatment.
EFFECT .. Treatment
Multivariate Tests of Significance (S = 2, M = 1/2, N = 41 )
Test Name Value Approx. F Hypoth. DF Error DF Sig. of F
Pillais .17684 2.06114 8.00 170.00 .042
Hotel lings .21184 2.19788 8.00 166.00 .030
Wilks .82427 2.13044 8.00 168.00 .036
Roys .17032
Notes: 2 cases rejected because of missing data,
F statistic for Wilks' Lambda is exact.
EFFECT .. Treatment (Cont.) Univariate F-tests with (2,87) D. F.
Variable Hypoth. SS Error SS Hypoth. MS ErrorMS F Sig. of F
Fun 21.55435 129.48743 10.77718 1.48836 7.24097 .001
Design 3.75257 146.09915 1.87628 1.67930 1.11730 .332
Interest 3.97440 192.15964 1.98720 2.20873 .89970 .410
Relevancy 3.07783 201.52355 1.53892 2.31636 .66437 .517
Table 4.15 Treatment vs. Attitude Subscales MANOVA

Post-hoc pairwise comparisons were conducted for the Fun
subscale, using the Tukey HSD procedure. The test yielded significant
differences between the Related Humor treatment (mean = 1.86) and
both the No Humor treatment (mean = .81) and the Unrelated Humor
treatment (mean = .82).
Open Response Analyses
Perhaps the most valuable data gathered during the experiment
were the verbatim open-ended responses from the subjects exposed to
review screen humor. Related Humor and Unrelated Humor
treatments included the open response screen exhibited as Appendix H.
The following sections present the results of quantitative and qualitative
analyses for the responses to this measure.
Chi-square analyses. Of particular interest were the opinions of
subjects in the Related Humor and Unrelated Humor treatment groups
regarding the effects of humor on their review behaviors. The
relationship between Related and Unrelated Humor treatments for this
question were analyzed using a treatment by Review Code contingency
matrix. Comments that related to reviewing were categorized and coded,
with a lack of comment or an equivocal remark also noted. One Review
Code was assigned to each subject according to the following scheme:
None No opinion with respect to humor affecting review
Con Negative opinion with respect to humor affecting review
Pro Positive opinion with respect to humor affecting review
Neutral Neutral stance with respect to humor affecting review.

The bar chart in Figure 4.9 dearly indicated that the number of
subjects expressing positive opinions on this topic in both humor
treatments surpassed the number of subjects expressing negative
R = Related
30- U = Unrelated
I20' ________________
10- R u
*_________________ " R '
_______R________ u U
oil M ---------------------,---------------,---------------,------Li
The primary interest, however, is in the comparison of the
number of negative (Con) versus the number of positive (Pro)
comments across treatments. Inspection of Table 4.16 resulted in the
immediate conclusion that these data resulted in no significant
relationship between type of humor treatment and the Pro and Con
coded comments relating to review behaviors (X2 = .057, p = .81 for the
Pro and Con cells).
R = Related
U = Unrelated
None Con Pro
Humor Effect with Respect to Reviewing
Figure 4.9 Treatment vs. Open Response Review Codes