ArticlePDF Available

The Role of Choice in Reader Engagement

Authors:

Abstract and Figures

Two experiments investigated the effect of choice on cognitive and affective engagement during reading. Both experiments compared college students who either selected what they read or were assigned the same story without being allowed to choose. Experiment 1 found that unrestricted choice heightened favorable affective perceptions of the reading experience compared with denied-choice and control groups but had no effect on cognitive measures of engagement. Experiment 2 replicated these findings when individuals within a single group were offered choice or were denied choice. The authors discuss the need for a more explicit theory of choice, which presently does not exist. (PsycINFO Database Record (c) 2012 APA, all rights reserved)
Content may be subject to copyright.
Journal of Educational Psychology
1998,
Vol. 90, No. 4, 705-714
Copyright 1998 by
the
American Psychological Association, Inc.
0022-0663/98/S3.00
The Role of Choice in Reader Engagement
Gregory Schraw, Terri Flowerday, and Marcy
F.
Reisetter
University of Nebraska—Lincoln
Two experiments investigated the effect of choice on cognitive and affective engagement
during reading. Both experiments compared college students who either selected what they
read or were assigned the same story without being allowed to choose. Experiment 1 found
that unrestricted choice heightened favorable affective perceptions of the reading experience
compared with denied-choice and control groups but had no effect on cognitive measures of
engagement. Experiment 2 replicated these findings when individuals within a single group
were offered choice or were denied choice. The authors discuss the need for a more explicit
theory of
choice,
which presently does not exist.
Most people feel that choice plays an important role in
their lives. Experts concur with this view, suggesting that
choice is an important determinant of interest, cognitive
processing, motivation, and even long-term health (Glasser,
1986;
Langer, 1989). For example, Kohn (1993) claimed
that choice among younger students positively affects activ-
ity level, enthusiasm, creativity, depth of comprehension,
self-regulation, and rate of learning. Many recent accounts
of effective instruction (Lepper, 1988; Stipek, 1997) and
classroom motivation (Pintrich & Schunk, 1996; Stipek,
1993) also support the positive effects of choice. It is
surprising then that so few educational studies have exam-
ined the role of choice empirically, or postulated theoretical
models that articulate the relationship between choice and
various aspects of cognitive and affective engagement.
Three current lines of research emphasize the role of
choice in learning and motivation: self-determination theory,
computerized testing theory, and reader response theory.
Self-determination theory states that choice has a positive
impact on cognitive and affective engagement because it
increases intrinsic motivation (Deci, 1992; Deci & Ryan,
1987;
Deci, Vallerand, Pelletier, & Ryan, 1991). Most
studies within this framework have examined the role that
controlling environments (e.g., teachers and structured class-
room settings) play in autonomy and learning (Flink,
Boggiano, & Barrett, 1990; Gromick & Ryan, 1987; Miser-
andino, 1996). These studies generally suggest that control-
ling environments reduce a sense of personal autonomy and
intrinsic motivation, and result in decreased learning and
poorer attitudes about school. A number of other studies
have examined the role that perceived control (i.e.,
self-
judgments of personal competence or autonomy) plays in
intrinsic motivation (Boggiano, Main, & Katz, 1988; Skin-
ner, Wellborn, & Cornell, 1990; Williams & Deci, 1996).
Gregory Schraw, Terri Flowerday, and Marcy F. Reisetter,
Department of Educational Psychology, University of Nebraska
Lincoln.
Correspondence concerning this article should be addressed to
Gregory Schraw, Department of Educational Psychology, 1313
Seaton Hall, University of Nebraska, Lincoln, Nebraska 68588.
Electronic mail may be sent to gschraw@unlinfo.unl.edu.
These studies indicate that greater perceived autonomy
results in higher levels of intrinsic motivation and enjoyment.
A review of the self-determination literature reveals only
three studies that specifically address the relationships
among choice, learning, and engagement. Zuckerman, Po-
rac,
Lathin, Smith, and Deci (1978) asked 80 college
students to select puzzles they would like to work on during
an experimental session. Forty yoked pairs were created
such that each individual in the choice condition selected
three of six puzzles to work on, then indicated how much
time he or she would allot to each puzzle. Individuals in the
yoked group were assigned the puzzles and the same time
allotments selected by individuals in the choice group.
Individuals who were allowed to choose reported a greater
feeling of control, indicated that they would be more willing
to return for another session of puzzle solving, and spent
significantly more time solving similar puzzles in a free-
choice period at the end of
the
experiment.
Cordova and Lepper (1996) examined the role of choice
when elementary school children used computer-aided learn-
ing environments to improve arithmetical and problem-
solving skills. Allowing children to make choices positively
affected several measures of affective engagement, includ-
ing perceived competence, a preference for greater task
difficulty, overall liking, and a greater willingness to stay
after class compared with students in a control group. In
contrast, there were few significant effects for choice with
respect to cognitive engagement variables. Specifically,
choice had no impact on use of hints, use of complex
problem-solving operations, or the amount of strategic play,
although those given choice performed significantly better
on a follow-up math test. A related study by Parker and
Lepper (1992) did not report differences in cognitive engage-
ment as a function of choice.
Computerized testing theory provides a second line of
research that emphasizes the role of choice. These studies
typically compare computer-adapted tests (i.e., those in
which a computer selects test items for an examinee) with
self-adapted tests (i.e., those in which examinees select test
items from one of several preassigned difficulty levels).
These tests are becoming increasingly popular because they
are more efficient than paper-and-pencil versions of the
705
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
706
SCHRAW, FLOWERDAY, AND REISETTER
same test. Rocklin and O'Donnell (1987) and Rocklin,
O'Donnell, and Hoist (1995) found that college students
performed better on a self-adapted test than on a computer-
adapted version. Wise, Plake, Johnson, and Roos (1992)
showed that individuals taking the self-adapted version
reported significantly lower posttest anxiety. Wise, Roos,
Plake, and Nebelsick-Gullet (1994) found that the type of
test one completes has no effect on posttest anxiety provided
one is allowed to choose the test's format (i.e., California
Achievement Test vs. Scholastic Assessment Test). Thus,
choice is related to anxiety level. Students with high levels
of pretest anxiety generally chose the self-adapted version;
low-anxiety students generally chose the computer-adapted
version. In an extension of
this
finding, Wise, Roos, Leland,
Oats,
and McCrann (1996) reported that individuals with a
high desire for control were significantly more likely to
select a self-adapted rather than a computer-adapted test.
Thus,
being able to choose how one is tested appears to
reduce anxiety, regardless of the type of test one takes, and
especially if one has a high desire for
control.
Being allowed
to choose what items one completes, even if one cannot
choose the test format per se, also reduces anxiety. Overall,
this research suggests that choice reduces anxiety, and
reduced anxiety improves test performance.
Reader response theory provides a third line of research
that emphasizes the role of choice. Three assumptions
appear to be ubiquitous in this literature. One is that choice
promotes aesthetic reading, which is characterized by greater
empathy with characters and events within the story, as well
as a greater sense of pleasure and enjoyment when reading
(Rosenblatt, 1994; Zarillo & Cox, 1992). A second assump-
tion is that choice increases short-term interest in a topic
(Deci, 1992; Hynds, 1990). A third is that choice increases
deeper cognitive processing (Gambell, 1993; Gambrell &
Marinak, 1997), either in terms of generating a greater
number of text-relevant inferences or in constructing a
holistic interpretation of the text (van Dijk
&
Kintsch, 1983).
Indeed, Gambrell and Marinak (1997) claimed that "re-
search abounds that attests to the motivating quality of
choice in literacy learning" and that "research clearly
supports the notion that there is a strong correlation between
choice and the development of intrinsic motivation" (p. 214).
Unfortunately, we have been unable to identify any empiri-
cal studies that specifically examine the relationship be-
tween choice and reading engagement. One of the studies
cited prominently by Gambrell and Marinak (i.e., Turner,
1995) does not control for choice, even though it does
incorporate a variable generated from observations labeled
volitional control, which occurs "when students took ac-
tions to aid their concentration" making "efforts to prevent
distraction, like asking others to work quietly or asking for a
change of task" Turner, 1995, (p. 422).
Not all studies report positive effects for choice. For
example, Parker and Lepper (1992) did not report significant
effects on cognitive aspects of learning when grade-school
children were offered simple choices within a computer-
aided learning environment. Pollock and Sullivan (1990)
reported that self-controlled instruction was less effective
than other-controlled instruction. Cordova and Lepper (1996),
Hannafin and Sullivan (1996), and Morrison, Ross, and
Baldwin (1992) reported that choice had a positive effect on
some aspects of affective engagement but no effect on
cognitive engagement.
The research summarized above suggests three main
conclusions about the relationship between choice and
engagement. One conclusion is that choice is related posi-
tively to measures of affective engagement such as intrinsic
motivation, feelings of satisfaction, and reduced anxiety.
Virtually all of these measures are self-reported perceptions
of engagement. All of the empirical studies we reviewed
supported this claim. A second is that choice has less of an
influence, and perhaps none at all, on measures of cognitive
engagement such as strategy use, recalling main ideas, and
generating inferences. Most of these measures are objective
assessments of change in the amount or quality of perfor-
mance. Most studies reported no effects or effects for some
variables with small effect sizes. None reported robust
effects of the magnitude suggested by strong proponents of
choice (Kamii, 1991; Kohn,
1993).
A third is that the impact
of choice may be related to desire for control within a
particular task environment. Among those who seek choice,
choice appears to reduce anxiety and other affective distress
that interferes with performance. Among those who do not
seek personal control or choice in a testing situation, choice
may interfere with performance (Wise, 1994).
The purpose of the present research is to investigate the
effects of choice on different measures of cognitive and
affective engagement during
reading.
This question is impor-
tant for at least two reasons. One is that many claims have
been made regarding the instrumental role of choice in
skilled reading (Gambell, 1993; Gambrell & Marinak, 1997;
Hynds, 1990; Rosenblatt, 1994), even though there is no
experimental research that we are aware of that specifically
addresses whether choosing a text positively affects either
cognitive or affective reading engagement. A second reason
is to clarify the potentially separable effects of choice on
cognitive and affective engagement. Existing research sug-
gests that choice positively affects affective engagement
(e.g., interest) without necessarily affecting cognitive engage-
ment (e.g., generating inferences).
We conducted two experiments by using seven measures
of cognitive and affective engagement (see the Method
section for further details). The three cognitive measures
included a 20-item multiple-choice test of the main ideas,
the number of thematic inferences made about the text, and
the number of critical responses made about the text. We
also included a 13-item desire for control scale as a
preexperimental covariate. The four affective measures
included interest ratings, the type of personal responses
reported after reading, a holistic score indicating each
reader's overall emotional reaction to the passage, and a
12-item survey completed after the experiment that mea-
sured personal responses such as enthusiasm and sense of
control while reading.
We tested the extent to which choice of what one reads
affects the seven types of cognitive and affective engage-
ment described above. We expected choice to positively
affect affective engagement but to have little or no impact on
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
READER CHOICE
707
cognitive engagement. Henceforth, we refer to this as the no
cognitive engagement hypothesis. This outcome would be
consistent with previous findings but would be contrary to
claims from a variety of researchers that choice uniformly
increases affective and cognitive engagement (Gambrell &
Marinak, 1997; Kamii, 1991; Kohn, 1993; Langer, 1989;
Lepper, 1988). Empirical data are needed to resolve that
apparent conflict between anecdotal reports and available
research. We believe there is a strong folk-psychological
sentiment that choice necessarily increases cognitive engage-
ment during reading (Gambell, 1993; Gambrell & Marinak,
1997) and other activities (Kohn, 1993). This perception
may be erroneous. Presently, it is not clear from the literature
whether choice increases cognitive engagement at all, and if
it does, it is not clear under what conditions it increases
engagement or what the magnitude of that increase may be.
The present research represents a preliminary test of the no
cognitive engagement hypothesis.
Experiment 1
Method
Participants. Seventy-eight college undergraduates (47 women,
31 men) participated in the experiment as part of their course
requirement in educational psychology.
Design. We used a between-subjects, one-way analysis of
variance design with three groups, which included an unrestricted-
choice, a denied-choice, and a control group.
The unrestricted-choice group chose from among three texts.
The denied-choice group was assigned a text under the assumption
that other participants who had been given a choice elected not to
read it. The control group received no special instructions.
The control group was included to assess the relative impact that
choice had on reading engagement. One possibility is that the
unrestricted-choice and control groups do not differ from one
another, yet differ from the denied-choice group. This would
suggest that choice per se does not affect engagement, whereas
denied choice negatively affects engagement. Another possibility is
that the unrestricted-choice group differs from both the denied-
choice and control groups. This would suggest that choosing what
one reads positively affects engagement, whereas denying choice
has no detrimental effects per se. A third possibility is that each of
the three groups differ significantly from each other. This would
suggest that
choice,
and the denial of choice, each affect engagement.
Materials. The materials consisted of a 13-item desire for
control scale, three different short texts, a 10-item interest question-
naire, a 20-item multiple-choice test booklet for each of the three
texts,
an essay booklet in which participants described their
reactions to the text, and a 12-item attitudes checklist.
The desire for control scale includes 13 items developed by Wise
et al. (1996).
1
This scale measures the degree to which individuals
want control in a testing situation. Ratings were made using a
5-point Likert scale that ranged from never applies to me (1) to
always applies to me (5). In three separate experiments, Wise et al.
reported a single factor solution with a coefficient alpha of
approximately .80. We included this scale as a possible covariate,
given that differences among the three groups with respect to desire
for control might confound our results. Coefficient alpha in
Experiment
1
was .78.
The three experimental texts consisted of a 900-word fictional
narrative entitled The Book of
Sand,
a
1,100-word
expository text
entitled Winter Depression: A Case of Being SAD, and a
1,200-
word mixed text entitled The Burning of Kuwait. Each text was
written at a lOth-grade readability level. The Book of Sand was
adapted from a work by Jorge Luis Borges (1977); Winter
Depression was adapted from Science News (Bower, 1992); The
Burning of Kuwait was adapted from Time (Elmer-DeWitt, 1991).
Each text booklet included a cover sheet with a one-sentence
description of the story. These descriptors were "A fictional story
about a very mysterious book and its owners," "A popular science
report on the latest research findings and treatments of winter
depression," and "A factual article about the long term effects of
the Persian Gulf war." Descriptors were written to be of identical
length and specificity.
Only data from the Winter Depression story were used in
statistical analyses in Experiment 1 (i.e., 56 participants). We used
this story because participants selected it more frequently (« = 17)
than either of the other stories. Individuals in the unrestricted-
choice group who selected the The Book of Sand (n = 13) and The
Burning of Kuwait (n = 9) were dropped from the data analysis in
order to eliminate comparisons among different texts. In the final
analyses, individuals in the unrestricted-choice group consisted
exclusively of those who selected Winter Depression. Individuals
in the denied-choice and control groups were assigned this story.
The interest questionnaire included 10 items adapted from the
perceived interest questionnaire used by Schraw, Bruning, and
Svoboda
(1995).
Ratings were made with a 5-point Likert scale that
ranged from strongly disagree (1) to strongly agree (5). Schraw et
al.
(1995) found mat all items loaded on a single factor with a
coefficient alpha of .83. Coefficient alpha based on the 56
participants in Experiment
1
was .90.
Each text had an accompanying 18-item multiple-choice test. All
items on the Winter Depression test assessed understanding of main
ideas included in the story. A typical item is shown below;
Based on the article, one can conclude that light-related
depression occurs with certainty in:
(a) humans
(b) rats and humans
(c) all mammals
(d) all animals
Overall performance on the test among the 56 participants was
approximately 66% correct completion of test items. Coefficient
alpha was .88.
The essay booklet asked participants to provide a two-page
written reaction to the
Winter
Depression story that addressed two
aspects of the text: (a) what they think the story means and (b) what
kind of personal thoughts and feelings the story evoked. Essays
were written immediately after the multiple-choice test was
finished. Responses were grouped into three main categories:
thematic, critical, and personal responses (see Appendix A).
Twelve subcategories (four under thematic, three under critical,
and five under personal responses) were identified in part on the
basis of a content analysis of randomly selected protocols (Weber,
1985) and similar categories proposed by Many and Wiseman
(1992).
The thematic responses category included retellings, elabora-
tions,
interpretations, and the evaluation of evidence information.
The critical responses category included statements about new
learning, difficulty understanding the text, and critical analyses of
the text
ideas.
The personal responses category included statements
pertaining to reader engagement, cognitive reactions, affective
reactions, empathy with events and characters, and relating experi-
ences described in the text to one's own life.
J
A11 materials and scales not included in the appendixes are
available from Gregory Schraw.
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
708 SCHRAW, FLOWERDAY, AND REISETTER
Each written protocol was scored
by
Gregory Schraw and Teiri
Flowerday,
who
worked together. Each protocol
was
read
in its
entirety, followed by
a
sentence-by-sentence analysis of responses.
Most statements were scored within
1 of the 12
subcategories.
Some statements were excluded because they were irrelevant
or
unrelated
to the
essay's main objectives. Both judges discussed
differences until a consensus was reached; thus, there was complete
agreement
for
each
of
the
56
protocols. Each essay
was
given
an
overall rating
as
well using
a
5-point
scale that ranged from very
negative reaction (1)
to
very
positive reaction (5).
These protocols served
two
related purposes.
The
first
was to
compare the total number of responses among the three experimen-
tal groups.
A
second purpose was
to
compare responses
in
each
of
the three categories. The thematic and critical response categories
were used
as
indexes
of
cognitive engagement, whereas
the
personal response category
was
used
as an
index
of
affective
engagement.
The attitudes checklist assessed
12
different aspects
of
partici-
pants'
affective engagement, including enjoyment, satisfaction,
effort, deeper processing, motivation, fairness,
and
sense
of
self-
control (see Appendix
B).
These
12
dimensions were selected
on
the basis
of
claims appearing
in the
choice literature
and the
dimensions identified
by
Kohn (1993). Ratings were made using
a
5-point
Likert scale that ranged from strongly disagree
(1) to
strongly agree (5). Coefficient alpha for the full scale was .86.
Procedure. Individuals were randomly assigned to one of three
groups. Participants
in the
unrestricted-choice group were given
three different texts
in
sealed packets and asked
to
choose one text
to read
on the
basis
of a
one-sentence descriptor
on the
packet's
cover. Descriptors consisted
of
informationally equivalent summa-
ries of the story. The two remaining packets were returned before
the chosen packet was opened. Thirty-nine individuals participated
in the unrestricted-choice group.
Participants
in the
denied-choice
and
control groups were
assigned
the
Winter Depression story. Individuals
in the
denied-
choice group were told (fictitiously) that other groups
had
been
asked
to
choose among three texts
and
that participants
in
those
groups selected either The Burning of Kuwait or The Book of
Sand.
As
a
consequence, individuals were needed
to
read
the
Winter
Depression text
to
balance
the
research design. Individuals
in the
control group received no special instructions.
Procedures were identical
for the
three groups following
the
initial phase. Individuals completed
the
13-item desire
for
control
scale, then were given
as
much time
as
needed
to
read the story
as
carefully
as
possible. Average completion time during
the
reading
phase was
8
min. Individuals next completed the interest question-
naire, followed
by the
two-page essay. Participants were given
as
much time
as
needed
to
complete
the
essay phase, although
everyone finished within
20 min. The
attitudes checklist
was
completed last
and
required approximately
3 min. The
entire
experiment required approximately 50 min to complete.
Results
Separate analyses were conducted for the desire for
control, interest questionnaire, multiple-choice test, essay
responses, and the attitudes checklist. All statistical tests
were conducted at thep < .05 level unless otherwise noted.
Table 1 shows the means and standard deviations for the
desire for control, interest, and multiple-choice tests. The 13
scores on the 5-point desire for control scale were combined
to create a single composite score that could range from 0 to
65.
Mean scores among the three groups did not differ on
this dimension, F(2, 53) = 1.77, MSE = 38.07, p = .18.
Table 1
Means and Standard Deviations for Desire for Control,
Interest, and Multiple-Choice Test in Experiment 1
Outcome measure
Desire
for
control
Interest
Multiple-choice test
Unrestricted-
choice
M
SD
48.64
7.16
32.17
5.98
12.00
2.34
Group
Denied-
choice
M
SD
52.04
5.45
30.95
7.00
11.67
2.17
Control
M
49.05
29.27
12.22
SD
5.94
6.64
2.04
Note.
For
unrestricted-choice group,
n =
17;
for
denied-choice
group,
n
21;
for
control group,
n
18.
Scores
on the
interest questionnaire were combined
to
create
a single composite score that could range from
0
to 50. Mean
scores
did not
differ among
the
three groups,
F(2
f
53) =
0.85,
MSE =
43.45,
p > .40.
Mean scores
on the
18-item
multiple-choice test
did not
differ either,
F(2, 53) = 0.32,
MSE =
4.18,
p> .70.
Table 2 includes means and standard deviations for the 12
subcategories derived from essay responses. A test of total
scores (i.e., the sum of all responses among the 12 subcatego-
ries) was not significant, F(2, 53) = 2.47, MSE =
10.28,
p =
.09.
We also conducted individual F tests for each of the 12
subcategories by using a protected alpha level of p < .005.
When we used this criterion, none of the subcategories
differed significantly.
Table 3 shows the means and standard deviations for the
12 items on the attitudes checklist. An analysis, using a
one-way multivariate analysis of variance, of all 12 scores
reached significance, approximate F(24, 84) = 2.19. We
also tested each of the 12 items by using individual F tests
with a protected alpha level ofp < .005. With this criterion,
Item 6 (I liked what I was asked to do in this study), F(2,
53) = 7.14, MSE = 0.60, and Item 12 (I felt I had a great
Table 2
Means and Standard Deviations for Personal Response
Essay Scores in Experiment 1
Type
of
response
Retellings
Elaborations
Interpretations
Evidence
New learning
Difficulty understanding
Critical analyses
Engagement
Cognitive reactions
Affective reactions
Empathy
Personal experiences
Unrestricted-
choice
M
0.64
0.64
1.00
0.58
0.58
0.11
0.59
1.52
1.41
0.82
0.12
2.24
SD
1.49
1.11
1.22
1.00
0.50
0.49
1.00
0.80
1.76
1.18
0.49
1.20
Group
Denied-
choice
M
0.59
1.04
1.09
0.68
0.23
0.05
0.55
1.13
1.01
0.95
0.09
2.55
SD
0.90
1.09
1.26
1.75
0.43
0.21
1.01
1.12
1.63
1.25
0.29
2.08
Control
M
1.88
1.41
1.82
0.94
0.53
0.00
1.00
1.29
1.00
0.41
0.23
1.64
SD
1.49
1.22
1.74
1.08
0.80
0.00
1.46
1.10
1.27
0.87
0.56
1.53
Note.
For
unrestricted-choice group,
n =
17;
for
denied-choice
group,
n
21;
for
control group,
n
18.
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
READER CHOICE
709
Table 3
Means and Standard Deviations for
the
Attitudes
Checklist in Experiment 1
Item
1
2
3
4
5
6
7
8
9
10
11
12
Unrestricted-choice
M
3.88
4.11
3.94
3.94
2.94
3.65
3.71
3.88
3.53
3.06
4.65
3.94
SD
0.60
0.78
0.96
0.66
1.08
0.70
1.10
1.11
0.94
0.83
0.61
1.08
Group
Denied-choice
M
3.43
3.52
3.38
3.62
2.86
3.29
3.61
3.62
3.00
3.04
4.71
3.80
SD
0.68
0.98
0.68
0.74
0.96
0.78
0.92
0.97
0.95
1.02
0.46
1.07
Control
M
3.28
3.22
3.22
3.78
2.83
2.66
3.33
3.U
3.22
3.22
4.44
2.77
SD
0.96
0.94
1.11
0.55
0.92
0.84
1.13
0.96
0.55
1.00
0.98
1.39
Note. For unrestricted-choice group, n
17; for denied-choice
group, n =
21;
for control group, n = 18.
deal of control in this study), F(2, 53) = 5.57, MSE = 1.32,
reached significance. Post hoc comparisons conducted using
Fisher's procedure (Levin, Serlin, & Seaman, 1994) indi-
cated that both the unrestricted-choice and denied-choice
groups were more likely than was the control group to state
that they liked what they were asked to do in the study, even
though neither of these groups differed from each other. An
identical pattern was found on Item 12. Both the unre-
stricted- and denied-choice groups reported feeling more
control than did the control group.
Discussion
None of the groups differed with respect to desire for
control, interest, multiple-choice test performance, or essay
responses. In addition, the three groups differed on only 2 of
the 12 items on the attitudes checklist. In this case, the
control group reported feeling less control than either the
unrestricted- or denied-choice groups, which did not differ
on any of the 12 items.
These findings clearly support the no cognitive engage-
ment hypothesis, given that none of the cognitive measures,
and only one of the affective measures (i.e., the attitudes
checklist), differed among the three groups. Surprisingly,
with respect to the attitudes checklist, denying choice may
actually improve attitudes compared with the control group.
One possible explanation is that participants believed they
were helping the researchers by reading an otherwise
undesirable text.
Experiment 2
Experiment 1 revealed few differences among the three
groups. We conducted a second experiment to examine two
aspects of this outcome in more detail. The first aspect
pertained to the experience of denied choice. In Experiment
1,
individuals were told they would be reading a text that the
other groups chose not to read. Some individuals in the
denied-choice group may have concluded that they were
providing a useful service to us (the researchers) even
though they were denied choice. This possibility could mask
potentially negative aspects of denied choice. Experiment 2
made the denial of choice more salient by randomly dividing
a single group of participants into two subgroups. One group
was allowed to choose one of two different texts; the other
group was assigned a text. We reasoned that this manipula-
tion would give those in the denied-choice subgroup a clear
reason for feeling disgruntled. Failure to identify differences
in cognitive engagement between the two groups under
these conditions would provide a stronger test of the no
cognitive engagement hypothesis.
Experiment 2 also sought to conduct a more powerful
statistical test. One argument that could be made against
Experiment 1 is that it failed to detect subtle differences
among groups because of low statistical power. Although the
observed data did not support this claim, we nevertheless
thought it useful to control for this possibility. Experiment 2
included two groups with approximately 60 participants in
each. This scenario permitted us to detect subtle differences
between groups.
Method
Participants. One hundred sixty-four college undergraduates
(91 women, 73 men) participated in the experiment as part of their
course requirement in educational psychology. Of these partici-
pants,
43 selected The Burning of Kuwait text and therefore were
dropped from the statistical analysis. All of the remaining 121
participants read the
Winter
Depression text.
Design. We used a two independent-groups design, which
included a choice and a denied-choice group.
Materials. The materials were identical to those used in
Experiment 1, with two minor changes. First, individuals were
asked to choose from two rather than three texts. Second, partici-
pants in Experiment 2 wrote two essays. The first essay was the
same as in Experiment 1. In Essay 2, participants described their
personal reactions to participating in the experiment. This essay
took approximately 5 min to complete. Response categories were
identified on the basis of a content analysis of randomly selected
protocols (Weber, 1985). Statements were scored as positive or
negative within one of five general categories. The first category
included statements pertaining to research participation (e.g., I
liked this more than others I have done). The experimental
materials category included statements concerning any aspect of
the materials (e.g., The story was interesting). The writing essay
category included statements specifically directed at writing either
Essay 1 or Essay 2 (e.g., I think the essay does a better job of
showing you what I learned from the story). The choice format
category included statements related to choice (e.g., I liked being
able to choose, but felt sorry for the other people). The miscella-
neous category included statements that were relevant to the essay
instructions but did not fit the other categories (e.g., I don't like to
have to wait for others before I can go on to the next part of the
study).
Procedure. Individuals participated in one of three large sec-
tions of
57,51,
and
53
individuals. Individuals in each session were
randomly selected from a sign-in roster and divided into two
subgroups. Approximately
40%
of participants were assigned to the
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
710 SCHRAW, FLOWERDAY, AND REISETTER
denied-choice condition. The remaining 60% were assigned to the
choice condition. Those in the denied-choice condition were given
a copy of
the
Winter Depression story in a sealed materials packet.
Those in the choice condition were given two sealed materials
packets containing the Winter Depression and The Burning of
Kuwait stories. Participants were asked to select one of these
stories to read on the basis of the one-sentence descriptors printed
on the cover. The unselected packet was then returned. Thereafter,
procedures were identical to Experiment
1
with the inclusion of the
additional experimental participation essay.
Results
Analyses paralleled those reported in Experiment 1.
Coefficient alpha for each of the instruments used in
Experiment 2 was as follows: desire for control, .88; interest,
.92;
multiple-choice test, .87; and attitudes checklist, .91.
Table 4 shows the means and standard deviations for the
desire for control, interest, and multiple-choice tests. Mean
scores on the desire for control scale did not differ between
the unrestricted and denied-choice groups, r(l, 119) = 1.77,
p > .05 or for the multiple-choice test of main ideas, /(I,
119) = 0.24, p < .50. In contrast, scores on the interest
questionnaire differed significantly, *(1, 119) = 3.15, p <
.05.
The unrestricted-choice group reported more interest in
the text than the denied-choice group.
Table 5 includes means and standard deviations for the 12
subcategories derived from essay responses. A test of total
scores (i.e., the sum of all responses among the 12 subcatego-
ries) for the choice (M = 8.37, SD = 2.25) and denied-
choice groups (M = 9.08, SD = 2.25) was not significant,
t(l, 119) = -1.71,p = .09. We also conducted individual t
tests for each of the 12 subcategories using/? < .005 level of
significance as our critical value. None of the 12 categories
differed at this level.
Table 6 shows the means and standard deviations for the
experimental participation essay. A test of total scores (i.e.,
the sum of all responses among the 10 subcategories) for the
choice (M - 1.93, SD = 1.07) and denied-choice groups
(M = 1.92, SD = 1.11) was not significant, t(l
t
119) - 0.05,
p =
.95.
We also conducted individual t tests for each of the
10 subcategories by using p < .005 level of significance as
our critical value. Individual t tests revealed significant
differences between the two groups on 4 of the 10 subcatego-
ries: positive comments about research participation, nega-
tive comments about experimental materials, and positive
Table 4
Means and Standard Deviations for Desire for Control,
Interest, and Multiple-Choice
Test
in Experiment 2
Outcome measure
Desire for control
Interest
Multiple-choice test
Group
Unrestricted-choice
M
51.01
30.70
11.67
SD
6.83
6.74
2.21
Denied-choice
M
49.93
26.03
11.57
SD
7.13
9.24
2.29
Table 5
Means and Standard Deviations for Personal Response
Essay Scores in Experiment 2
Type of
response
Retellings
Elaborations
Interpretations
Evidence
New learning
Difficulty understanding
Critical analyses
Engagement
Cognitive
Affective
Empathy
Personal experiences
Overall rating
Group
Unrestricted-choice
M
0.91
1.25
0.51
0.14
0.36
0.05
0.26
0.98
2.15
0.36
0.16
1.22
2.96
SD
1.24
1.03
0.73
0.34
0.61
0.22
0.48
0.96
1.37
0.69
0.41
1.07
0.95
Denied-choice
M
0.82
1.38
0.52
0.16
0.27
0.06
0.52
0.89
2.31
0.57
0.04
1.51
2.98
SD
1.32
1.45
0.84
0.44
0.45
0.25
1.07
0.81
1.44
0.93
0.21
1.66
0.94
Note. For unrestricted-choice group, n = 58; for denied-choice
group, n = 63.
and negative comments about the choice format. Table 6
reveals the direction of these differences. Generally, the
choice group was significantly more positive about research
participation and choice format, whereas the denied-choice
group was significantly more negative about the experimen-
tal materials and choice format.
Table 7 shows the means and standard deviations for the
12 items on the attitudes checklist. An analysis of all 12
scores using Hotelling's T
2
reached significance, T
2
(12,
108) = 81.39, p < .05. We also tested each of the 12 items
using individual t tests with a protected alpha level of p <
.005.
When we used this criterion, Item 2 (I felt good about
the choice of text),
*(1,
119) = 4.13; Item 6 (I liked what I
was asked to do in this study), f(l, 119) = 3.15; Item 8 (I
appreciate the choices I got to make in this study), r(l,
Table 6
Means and Standard Deviations for Experimental
Participation Essay Scores in Experiment 2
Group
Unrestricted-choice Denied-choice
Responses M SD
M SD
Positive
Research participation
Experimental materials
Essay writing
Choice format
Miscellaneous
Negative
Research participation
Experimental materials
Essay writing
Choice format
Miscellaneous
0.74
0.41
0.08
0.24
0.08
0.07
0.10
0.12
0.01
0.05
0.51
0.50
0.28
0.43
0.28
0.26
0.31
0.33
0.13
0.22
0.47
0.30
0.05
0.03
0.11
0.13
0.37
0.19
0.19
0.07
0.53
0.46
0.21
0.17
0.31
0.36
0.55
0.38
0.40
0.33
Note. For unrestricted-choice group, n 58; for denied-choice
group, n = 63.
Note. For unrestricted-choice group, n = 58; for denied-choice
group, n = 63.
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
READER CHOICE 711
Table
7
Means
and
Standard Deviations for the Attitudes
Checklist
in
Experiment
2
Item
1
2
3
4
5
6
7
8
9
10
11
12
Group
Unrestricted-choice
M
3.67
3.79
3.37
3.65
2.81
3.55
3.27
4.10
3.15
3.12
4.68
3.98
SD
0.99
1.12
1.12
0.82
0.96
0.84
1.05
0.74
1.02
0.86
0.56
0.96
Denied-choice
M
3.31
2.88
2.74
3.44
2.36
3.00
3.01
2.76
2.92
2.76
4.03
2.63
SD
1.18
1.27
1.28
1.04
1.15
1.06
1.43
1.21
1.14
1.21
1.35
1.38
Note. For unrestricted-choice group, n = 58; for denied-choice
group, n = 63.
119) = 7.14; Item 11 (I was treated fairly in this study), r(l,
119) = 3.42; and Item 12 (I felt I had a great deal of control
in this study), *(1,119) = 6.16, reached significance. Ratings
were higher in all cases for the unrestricted- compared with
the denied-choice group.
Discussion
Results revealed several differences between
the
choice
and denied-choice groups.
One was
that
the
choice group
reported more interest
in the
story compared with
the
denied-choice group.
A
second
was
that individuals
in the
choice group reported more positive comments about their
participation
in the
experimental reaction essay,
as
well
as
more favorable comments about the choice format. Those
in
the denied-choice group reported more negative reactions.
Third, individuals
in the
choice group reported more favor-
able reactions
on
five questions from
the
attitudes checklist
compared with
the
denied-choice group.
These findings indicate that offering some individuals
choice, whereas others within
the
same group do
not
receive
choice, positively affects
the
choice-receiving individuals'
perceptions
of
their engagement.
In
contrast, choice
had no
impact
on any of the
cognitive engagement variables (i.e.,
multiple-choice test, thematic responses,
and
critical
re-
sponses)
or
personal responses from
the
first essay. Collec-
tively, these results support
the no
cognitive engagement
hypothesis without reservation.
General Discussion
The purpose
of
this research
was to
investigate
the
effect
of choice
on
cognitive
and
affective engagement during
reading. Most previous studies have reported
a
positive
relationship between choice
and
affective measures such
as
interest
and
intrinsic motivation.
In
contrast,
few
have
reported increased cognitive processing, despite many anec-
dotal claims
to the
contrary. We conducted
two
experiments
to assess
the no
cognitive engagement hypothesis, which
states that choice increases self-report measures
of
affective
engagement
but
has no effect
on
cognitive engagement.
Experiment
1
compared unrestricted-choice, denied-
choice,
and
control groups
on
seven measures
of
cognitive
and affective engagement. None
of the
groups differed
on
any
of
the measures with
the
exception
of
two items
on the
end-of-experiment attitudes checklist.
In
this case, individu-
als
in
the unrestricted-choice group reported more favorable
attitudes regarding their participation
and
greater perceived
control compared with
the
experimental control group.
The
unrestricted-choice group
did not
differ from
the
denied-
choice group
on any of the
measures. Experiment
2 com-
pared choice
and
denied-choice groups
in a
single physical
setting. Neither group differed
on any of the
measures
of
cognitive engagement, whereas the unrestricted-choice group
reported more interest,
a
more favorable response
to
experi-
mental materials,
and
greater liking
for
the choice format
on
the self-report attitudes checklist. These experiments
are the
first that
we
know
of to
provide
a
controlled experimental
test
of
the role
of
choice
in
reader engagement.
Our findings were consistent with most previous research
studies, including Hannafin
and
Sullivan (1996)
and
Morri-
son
et al.
(1992),
who
reported that choice
had no
effect
on
cognitive engagement,
as
well
as
Pollock
and
Sullivan
(1990),
who
reported that self-controlled instruction
was
less effective than other-controlled instruction.
Our
findings
were consistent
as
well with
a
number
of
studies reporting
that choice increases affective engagement such
as
situ-
ational interest (Deci, 1992), self-reported feelings
of
satis-
faction (Langer, 1989),
and
greater personal control (Kohn,
1993).
In contrast,
our
findings were
at
odds with many
of the
claims that
one
encounters
in the
motivational (Deci,
1992;
Langer,
1989) and
reading response literatures (Gambell,
1993;
Gambrell
&
Marinak, 1997; Hynds, 1990) that choice
improves cognitive engagement.
For
example, Kohn (1993)
argued that choice increases
a
variety
of
measures
of
cognitive engagement variables such
as
deeper processing
and creativity.
On the
basis
of our
findings
and the
mixed
findings encountered
in the
existing research (Cordova
&
Lepper,
1996;
Parker
&
Lepper, 1992),
we
conclude that
strong claims about
the
relationship between choice
and
cognitive engagement
are
inflated,
at
least with regard
to
adult readers. Many
of
these claims,
in our
opinion, appear
to
be
anecdotal
in
nature
and
based
on a
strong folk-
psychological belief that choice invariably enhances
all
manner of performance. Although these assumptions may
be
true,
there
is no
compelling evidence
to
support them
at the
present time.
Given
our
failure
to
reject
the no
cognitive engagement
hypothesis,
we
believe
it is
incumbent upon
us to
scrutinize
our findings
in
greater detail.
We
believe there
are at
least
three potential counterexplanations that could
be
raised
to
dispute
our
findings.
One is
that
our
experiments lacked
methodological
and
statistical precision.
We
argue that
a
detailed inspection
of our
experimental materials, method,
and design will rule
out
this claim. Individuals were
randomly selected
and
assigned
to
groups that received
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
</