Retrieval Practice & Bloom's Taxonomy: Do Students Need Fact Knowledge Before Higher Order Learning?

Abstract

The development of students’ higher order learning is a critical component of education. For decades, educators and scientists have engaged in an ongoing debate about whether higher order learning can only be enhanced by building a base of factual knowledge (analogous to Bloom’s taxonomy) or whether higher order learning can be enhanced directly by engaging in complex questioning and materials. The relationship between fact learning and higher order learning is often speculated, but empirically unknown. In this study, middle school students and college students engaged in retrieval practice with fact questions, higher order questions, or a mix of question types to examine the optimal type of retrieval practice for enhancing higher order learning. In laboratory and K-12 settings, retrieval practice consistently increased delayed test performance, compared with rereading or no quizzes. Critically, higher order and mixed quizzes improved higher order test performance, but fact quizzes did not. Contrary to popular intuition about higher order learning and Bloom’s taxonomy, building a foundation of knowledge via fact-based retrieval practice may be less potent than engaging in higher order retrieval practice, a key finding for future research and classroom application.

Full text

Retrieval Practice & Bloom’s Taxonomy: Do Students Need Fact
Knowledge Before Higher Order Learning?
Pooja K. Agarwal
Washington University in St. Louis
The development of students’ higher order learning is a critical component of education. For decades,
educators and scientists have engaged in an ongoing debate about whether higher order learning can only
be enhanced by building a base of factual knowledge (analogous to Bloom’s taxonomy) or whether
higher order learning can be enhanced directly by engaging in complex questioning and materials. The
relationship between fact learning and higher order learning is often speculated, but empirically
unknown. In this study, middle school students and college students engaged in retrieval practice with
fact questions, higher order questions, or a mix of question types to examine the optimal type of retrieval
practice for enhancing higher order learning. In laboratory and K-12 settings, retrieval practice consis-
tently increased delayed test performance, compared with rereading or no quizzes. Critically, higher order
and mixed quizzes improved higher order test performance, but fact quizzes did not. Contrary to popular
intuition about higher order learning and Bloom’s taxonomy, building a foundation of knowledge via
fact-based retrieval practice may be less potent than engaging in higher order retrieval practice, a key
finding for future research and classroom application.
Educational Impact and Implications Statement
This study demonstrates that students’ higher order learning increases most from higher order
retrieval practice, or no-stakes quizzes with complex materials that engage students in bringing what
they know to mind. Although fact quizzes were beneficial for fact learning, they did not facilitate
higher order learning, contrary to popular intuition based on Bloom’s taxonomy.
Keywords: Bloom’s taxonomy, higher order learning, retrieval practice, testing effect, transfer
Supplemental materials: http://dx.doi.org/10.1037/edu0000282.supp
The development of students’ higher order learning is a critical
component of education. From both an educational perspective and
a scientific perspective, it is of practical interest to develop robust
strategies that increase higher order learning. For decades, educa-
tors and scientists have engaged in an ongoing debate about
instructional approaches: Should we build students’ foundation of
factual knowledge before engaging them in higher order learning,
or can higher order learning be enhanced directly by engaging
students in complex instructional techniques during the initial
learning process?
Regarding the first approach, many argue that to foster higher
order learning, we must focus on and reinforce students’ basic
knowledge. For instance, cognitive scientist Daniel Willingham
wrote, “Factual knowledge must precede skill” (2009, p. 19).
Diane Ravitch (2009), an education professor and historian, ar-
gued,
We have neglected to teach [teachers] that one cannot think critically
without quite a lot of knowledge to think about. Thinking critically
involves comparing and contrasting and synthesizing what one has
learned. And a great deal of knowledge is necessary before one can
begin to reflect on its meaning and look for alternative explanations.
This view is not new; there has been a long-standing call to
spend a substantial amount of instructional time on foundational
fact learning spanning more than 100 years (Bartlett, 1958;Bruner,
1977;Hirsch, 1996;James, 1900;Münsterberg, 1909).
On the other hand, many educators hold that less instructional
time should be spent on fact learning and more time should be
This article was published Online First June 7, 2018.
Pooja K. Agarwal, Department of Psychological and Brain Sciences,
Washington University in St. Louis.
I thank my Ph.D. advisor and dissertation committee chair, Henry L.
Roediger, III, for his dedicated mentorship and guidance. I also thank my
dissertation committee members for their significant input: David Balota,
Mark McDaniel, Michael Strube, Susan Fitzpatrick, and Keith Sawyer. I
am grateful to Columbia Middle School teacher Patrice Bain, principal
Roger Chamberlain, and the students for their participation. I also thank the
Roediger Memory Lab and the Balota Cognitive Psychology Lab, partic-
ularly Jason Finley and Geoffrey Maddox, for valuable discussions and
assistance with data analyses. This research was supported by the National
Science Foundation Graduate Research Fellowship Program, the Harry S.
Truman Scholarship Foundation, and the James S. McDonnell Foundation.
Correspondence concerning this article should be addressed to Pooja K.
Agarwal, who is now at the Department of Liberal Arts, Berklee College
of Music, Boston, MA 02115. E-mail: pooja@poojaagarwal.com
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.
Journal of Educational Psychology
© 2018 American Psychological Association 2019, Vol. 111, No. 2, 189–209
0022-0663/19/$12.00 http://dx.doi.org/10.1037/edu0000282
189

allotted for classroom activities that promote critical thinking,
analysis, and inquiry (Cuban, 1984;Dewey, 1916/1944;Kohn,
1999). For example, education professor Jal Mehta (2018) recently
argued,
What if, in science, we taught students the scientific method...by
having them write junior versions of scientific papers rather than
reading from textbooks?...[W]hy in school, do we think it has to be
dry basics first, and the interesting stuff only later?
According to the National Research Council (1987,p.8),
[T]he term “higher order” skills is probably itself fundamentally
misleading, for it suggests that another set of skills, presumably called
“lower order,” needs to come first. This assumption...justifies long
years of drill on the “basics” before thinking and problem solving are
demanded.
Higher Order Learning & Bloom’s Taxonomy
What, exactly, is “higher order” learning? Although there are
few agreed-upon definitions, higher order learning is frequently
classified using The Taxonomy of Educational Objectives by
Bloom, Engelhart, Furst, Hill, and Krathwohl (1956). The original
“Bloom’s taxonomy” included six categories of cognitive pro-
cesses, ranging from simple to complex: knowledge,comprehen-
sion,application,analysis,synthesis, and evaluation. Bloom et al.
explained that the taxonomy was designed as a step process: to
achieve a higher objective or category, one must first master
cognitive processes at a lower category. In other words, before
comprehension, application, or analysis can take place, a student
must first acquire knowledge.
In 2001, Anderson and coauthors of the original taxonomy
proposed a revised taxonomy (see Figure 1). The revised taxon-
omy highlights learning in verb tense: remember,understand
(previously called comprehension), apply,analyze,evaluate, and
create (previously called synthesis and reordered with evaluation).
Within the revised taxonomy, higher order learning is considered
to comprise the apply,analyze,evaluate, and create categories. On
the other hand, “lower order” learning requiring recognition, mem-
ory, and comprehension fall under the remember and understand
categories.
Bloom’s taxonomy has had a large impact on teacher prepara-
tion programs, classroom pedagogy, small- and large-scale assess-
ment programs, and educational research. In part because of its
simplicity, Bloom’s taxonomy has contributed to the collective
notion that foundational knowledge (literally the foundation or
base of the pyramid) precedes higher order learning (the categories
located higher in the pyramid). For example, educator Doug
Lemov (2017) explained,
Generally when teachers talk about “Bloom’s taxonomy,” they talk
with disdain about “lower level” questions. They believe, perhaps
because of the pyramid image which puts knowledge at the bottom,
that knowledge-based questions, especially via recall and retrieval
practice, are the least productive thing they could be doing in class.
No one wants to be the rube at the bottom of the pyramid.
Lemov’s comment perfectly illustrates the existing struggle to
identify whether knowledge at the base of Bloom’s taxonomy is a
prerequisite for higher order learning, or simply a nuisance in the
way of higher order learning.
In addition to Bloom’s taxonomy, how else might we define or
classify higher order learning? More recently, a taxonomy by
Barnett and Ceci (2002) is being used to classify students’ transfer
of knowledge across a number of content and context domains,
including physical location, type of task, and format. For example,
students’ “near transfer” along the content or knowledge domain
could involve learning how to calculate the Pythagorean theorem
and then applying this knowledge with a novel set of numbers. In
Figure 1. An illustration of the revised Bloom’s Taxonomy, based on Anderson et al. (2001).
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.
190 AGARWAL

contrast, “far transfer” along the physical domain could include a
medical student learning something from a textbook and then
applying it in clinical practice with patients (Pan & Agarwal, 2018;
Pan & Rickard, in press).
In terms of higher order learning, we are interested in far
transfer, which could occur across content and/or context domains.
In addition, Barnett and Ceci (2002) briefly offered a distinction
between horizontal and vertical transfer—where horizontal trans-
fer involves two tasks at the same level of complexity and vertical
transfer involves learning knowledge that is required for a wide
array of tasks that differ in complexity (p. 622, footnote 8). They
do not elaborate on the horizontal versus vertical distinction fur-
ther, unfortunately, but a third domain of “cognitive complexity”
may be crucial in identifying the type of higher order learning we
seek in educational settings.
Retrieval Practice and Transfer of Knowledge
Based on decades of scientific research, we have identified
robust strategies for enhancing students’ fact learning and transfer
of knowledge, including retrieval practice, spaced practice, and
interleaving (Dunlosky, Rawson, Marsh, Nathan, & Willingham,
2013;McDaniel, Roediger, & McDermott, 2007;Pashler et al.,
2007;Rohrer & Pashler, 2010). One strategy in particular—re-
trieval practice— dramatically improves long-term learning (Agar-
wal, Roediger, McDaniel, & McDermott, 2017;Roediger &
Karpicke, 2006b). In typical experiments, students study a set of
material (e.g., word pairs, foreign language vocabulary words,
prose passages), engage in retrieval practice (e.g., via free recall or
multiple-choice quizzes), and immediately or after a delay (e.g.,
ranging from hours, to days, to weeks) they complete a final test.
When students engage in retrieval activities that bring knowledge
to mind, learning is strengthened by the challenge (for recent
meta-analyses, see Adesope, Trevisan, & Sundararajan, 2017;
Rowland, 2014).
Specifically, recent research in classroom settings has demon-
strated that retrieval practice improves learning for diverse student
populations (e.g., middle school students to medical school stu-
dents), subject areas (e.g., introductory history to CPR skills), and
time delays (e.g., from a few days to 9 months; Kromann, Jensen,
& Ringsted, 2009;Lyle & Crawford, 2011;Roediger, Agarwal,
McDaniel, & McDermott, 2011). In addition, benefits from re-
trieval practice have been demonstrated in diverse educational
settings, including K-12 classrooms (Agarwal, Bain, & Chamber-
lain, 2012), undergraduate engineering courses (Butler, Marsh,
Slavinsky, & Baraniuk, 2014), and medical neurology courses
(Larsen, Butler, Lawson, & Roediger, 2013).
Retrieval practice also promotes students’ learning of complex
materials (Jensen, McDaniel, Woodard, & Kummer, 2014;
Karpicke & Aue, 2015;Pyc, Agarwal, & Roediger, 2014;Rawson,
2015; cf., van Gog & Sweller, 2015). In one study examining the
impact of retrieval practice on the learning of complex materials,
subjects studied bird exemplar picture-family name pairs (e.g., a
picture of a bird from the thrasher family) in either a repeated
study condition or a repeated quiz condition (Jacoby, Wahlheim, &
Coane, 2010). In the repeated study condition, subjects viewed
pictures of birds paired with their family names four times. For the
quiz condition, subjects studied picture-family name pairs once,
followed by three attempts to classify pictures for the eight bird
families. On a final test, recognition and classification perfor-
mance for picture exemplars was greater following repeated quiz-
zes than restudying. The materials in this study were more com-
plex than typical laboratory materials (e.g., word pairs, textbook
passages), whereas the cognitive processes comprised the lowest
two categories of Bloom’s taxonomy: remember and understand.
Moving to complex cognitive processes, recent research pro-
vides compelling evidence that retrieval practice is an effective
strategy to promote students’ transfer of learning (Butler, Black-
Maier, Raley, & Marsh, 2017;Carpenter, 2012;Karpicke & Blunt,
2011;Pan & Rickard, in press;Rohrer, Taylor, & Sholar, 2010).
Butler (2010) examined students’ transfer of knowledge across the
content domain described by Barnett and Ceci (2002), using pas-
sages drawn from Wikipedia and similar online sources. Subjects
read multiple passages and some were followed by repeated study-
ing, whereas others were followed by repeated quizzing. On final
tests, repeated quizzes led to greater fact and inferential learning,
compared with repeated studying. Importantly, when subjects were
given final test questions from a different knowledge domain (e.g.,
an initial quiz question about the wing structure of bats, in contrast
with a final test question about the wing structure of military
aircraft), repeated quizzes led to greater transfer performance than
repeated studying. In other words, retrieval practice improved
students’ transfer of content knowledge from bats to airplanes and
it also improved students’ inferential processing, moving closer
toward what we think of as higher order learning in terms of both
content and complexity.
In a real-world demonstration of the benefits of retrieval prac-
tice on transfer, McDaniel, Thomas, Agarwal, McDermott, and
Roediger (2013, Experiment 2) conducted an experiment in an 8th
grade science classroom in which retrieval practice included fact
and application questions (compared with no quizzing) and the
final tests also included fact and application questions. Of interest
was which type of retrieval practice would enhance final test
performance two weeks later, particularly on final application
questions. Retrieval practice significantly increased final test per-
formance regardless of format, and the application quiz condition
was most beneficial for application test performance. In other
words, when it comes to promoting complex learning, such as
students’ application of knowledge, engaging in complex retrieval
practice was more beneficial than starting with basic facts and
definitions. This finding supports the notion that higher order
learning can be enhanced directly by higher order retrieval prac-
tice, one of the first such demonstrations in an authentic classroom
setting. The study by McDaniel et al. (2013) did not examine
transfer across content or context domains; however, the materials
in the study extended students’ cognitive complexity from the
remember category to the apply category—the next step higher in
Bloom’s taxonomy (see Figure 1).
Critically, how far “up the pyramid” can we move student
learning, based on Bloom’s taxonomy? Should we first build
students’ foundation of knowledge or can we skip straight ahead to
complex retrieval practice during initial learning? The primary
purpose of this study was to examine the intuition that “factual
knowledge must precede skill” (Willingham, 2009). By using
complex passages, quizzes, and tests, I aimed to shed light on the
relationship between fact learning and higher order learning—a
critical distinction with practical and theoretical implications for
research and instruction.
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.
191
RETRIEVAL PRACTICE & BLOOM’S TAXONOMY

Theoretical Rationale
Three theoretical frameworks were utilized to explore the rela-
tionship between fact knowledge and complex higher order learn-
ing: the desirable difficulties framework, the transfer appropriate
processing framework, and what I refer to as the “foundation of
factual knowledge framework.” Importantly, these theories are not
mutually exclusive; rather, the three frameworks suggest both
similar and different results depending on the experimental con-
dition in the present study (see Table 1). By teasing apart these
theoretical similarities and differences in the current study, optimal
strategies to promote higher order learning can be developed and
implemented in authentic educational settings.
Desirable Difficulties Framework
According to the desirable difficulties framework, mental pro-
cesses that are challenging and effortful typically increase reten-
tion and application of knowledge (Bjork, 1994). Exposure to
“desirable difficulties,” particularly via retrieval activities, pro-
duces a struggle to bring information to mind that improves
learning and delays forgetting. In addition to retrieval practice, a
number of other strategies enhance learning to a greater extent than
their less effortful comparisons, such as spaced practice instead of
massed practice (Rohrer & Taylor, 2006) and interleaved practice
instead of blocked practice (Kornell & Bjork, 2008;Taylor &
Rohrer, 2010; see Dunlosky et al., 2013 for a review).
Furthermore, difficult retrieval practice may benefit delayed test
performance to a greater extent than easier retrieval practice
(Bjork, 1994;Gardiner, Craik, & Bleasdale, 1973;Pyc & Rawson,
2009). For example, Kang, McDermott, and Roediger (2007)
found that retrieval practice with short answer quizzes enhanced
final test performance to a greater extent than retrieval practice
with multiple-choice quizzes. Kang et al. concluded that short
answer quizzes may engage deeper recollective processing and
greater retrieval effort compared with multiple-choice quizzes,
thereby enhancing performance on both final short answer and
multiple-choice tests. These findings support the notion that re-
trieval practice with challenging questions (short answer or appli-
cation) can benefit performance on less and more demanding
criterial tests (both multiple-choice and short answer, or definition
and application; cf., McDermott, Agarwal, D’Antonio, Roediger,
& McDaniel, 2014).
Regarding the present study, the desirable difficulties frame-
work suggests that delayed test performance will be greater fol-
lowing retrieval practice compared with restudying or no quizzes,
consistent with previous research (e.g., Agarwal, Karpicke, Kang,
Roediger, & McDermott, 2008;Callender & McDaniel, 2009;
Carrier & Pashler, 1992). Initial higher order quizzes may serve as
a desirable difficulty, enhancing performance on both fact and
higher order tests (see Table 1). Meanwhile, mixed quizzes with
both fact and higher order questions (in Experiments 2 and 3) may
pose an additional challenge because the difficulty of questions
varies within quizzes, leading to even greater benefits on final test
performance than higher order quizzes.
Transfer Appropriate Processing
According to the transfer appropriate processing framework,
final performance is greatest when encoding processes engaged
during learning match retrieval processes engaged during testing
(McDaniel, Friedman, & Bourne, 1978;Morris, Bransford, &
Franks, 1977). This framework is often cited as another explana-
tion for why retrieval practice enhances long-term retention— by
engaging in retrieval practice, students match their initial process-
ing to the type of processing required at test (Roediger & Karpicke,
2006a).
In the classroom study described earlier, McDaniel et al. (2013)
demonstrated that challenging application quizzes promoted learn-
ing for both basic definition and complex application tests. They
Table 1
Theoretical Predictions and Obtained Results of Initial Retrieval Practice Condition on Final Test Performance, Compared With a
No Quiz Condition
Retrieval practice Final test Desirable difficulties
Transfer appropriate
processing
Foundation of factual
knowledge Obtained results
Fact quiz Fact test ⫹⫹ ⫹⫹ ⫹⫹ E1: ⫹⫹
E2: ⫹⫹
E3: N/A
Higher order quiz Higher order test ⫹⫹ ⫹⫹ 0 E1: ⫹⫹
E2: ⫹⫹
E3: ⫹⫹
Fact quiz Higher order test ⫹0⫹⫹ E1: 0
E2: 0
E3: N/A
Higher order quiz Fact test ⫹⫹ 0 0 E1: 0
E2: 0
E3: N/A
Mixed quiz Fact test ⫹⫹ ⫹ ⫹ E1: N/A
E2: ⫹⫹
E3: ⫹⫹
Mixed quiz Higher order test ⫹⫹ ⫹ ⫹ E1: N/A
E2: ⫹⫹
E3: ⫹⫹
Note. Facilitation: ⫹⫹; partial facilitation: ⫹; no prediction or effect: 0; not applicable: N/A.
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.
192 AGARWAL

also found results consistent with the transfer appropriate process-
ing framework: Across two experiments, test performance was
greatest when quiz questions matched the final test format, com-
pared with no quizzes or a quiz-test mismatch. For example,
student performance was greater on the application test following
application quizzes instead of definition quizzes.
Regarding the present study, the transfer appropriate processing
framework suggests that delayed test performance will be optimal
when initial quiz and final test formats match (see Table 1).
Specifically, retrieval practice with fact questions should benefit
delayed fact performance, and retrieval practice with higher order
questions should benefit delayed higher order performance (com-
pared with no quizzes or restudying). Further, a match between
quiz and test questions may promote performance to a greater
extent than a mismatch (fact quiz-higher order test or higher order
quiz-fact test).
Foundation of Factual Knowledge Framework
In accordance with the original Bloom’s taxonomy, the “foun-
dation of factual knowledge” framework suggests that we must
focus on and reinforce basic factual knowledge before we can
foster students’ higher order learning (Brown, Roediger, & Mc-
Daniel, 2014). For instance, Willingham (2009) argued that when
students practice facts until they are memorized, students can more
easily apply their deeper knowledge to higher order learning.
Retrieval practice of facts and procedures (e.g., applying the dis-
tributive property in algebra) may make recall automatic, thereby
requiring less effort and capacity from working memory. Subse-
quently, this process may enable students to use additional work-
ing memory capacity in more complex learning situations (Agar-
wal, Finley, Rose, & Roediger, 2017). Consistent with these
arguments, cognitive load theory posits that if cognitive demands
are reduced or diminished, subsequent learning may increase
(Plass, Moreno, & Brünken, 2010;Sweller, 2010;van Gog &
Sweller, 2015). As such, facilitation of fact learning via retrieval
practice may reduce cognitive demands and enhance final higher
order test performance, consistent with the foundation of factual
knowledge framework. Mental effort ratings (Paas, 1992) were
collected in the present study to elucidate this possibility.
Note that Willingham (2009) also distinguished between rote
knowledge and integrated or connected knowledge, the latter of
which may be required for higher order learning (see also Ausubel,
1961/1965;Ausubel, Novak, & Hanesian, 1978). For instance,
knowing the basic fact that George Washington was the first
president of the United States may not lead to a deeper under-
standing of United States civics or government. Willingham ar-
gued that, instead, the learning of connected knowledge (e.g.,
presidents are leaders who make important decisions) can facilitate
deep knowledge (e.g., if George Washington was the first presi-
dent, he must have made many important decisions) to construct a
rich understanding of a topic. In other words, simply learning
isolated facts without connecting them to a deeper understanding
of a topic may not benefit students’ higher order learning.
To directly examine the foundation of factual knowledge frame-
work, experiments included fact-based retrieval practice and de-
layed higher order tests. All fact questions in the present study
were developed to encompass key concepts or ideas from passages
(i.e., integrated knowledge, such as the goal of welfare programs),
rather than details such as names, dates, vocabulary words, defi-
nitions, and so forth (such as the year in which the Temporary
Assistance for Needy Families program was initiated). In keeping
with Willingham (2009) and the foundation of factual knowledge
framework, if an understanding of integrated concepts is required
for higher order learning, then delayed higher order test perfor-
mance should benefit from integrated fact quizzes (see Table 1).
Note that this prediction contradicts the transfer appropriate pro-
cessing framework, which suggests that the match between higher
order quizzes and a delayed higher order test will result in greater
test performance than the mismatch between fact quizzes and a
delayed higher order test.
Introduction to Experiments
The relationship between fact learning and higher order learning
is often speculated, but empirically unknown. To maximize learn-
ing while also experimentally manipulating cognitive complexity,
I used a retrieval practice paradigm with educational materials that
engaged students in lower order (e.g., remembering and under-
standing) and higher order (e.g., analyzing, evaluating, and creat-
ing) cognitive processes (see Figure 1). I also conducted a con-
ceptual replication in an authentic K-12 classroom in which
retrieval practice was embedded in students’ and teachers’ daily
activities, fluctuating schedules, and standard lessons (Experiment
3). Across three experiments, if initial fact learning increases
delayed higher order learning, findings will shed light on a long-
standing theoretical and practical debate— how best to achieve the
“holy grail” of the highest levels in Bloom’s taxonomy.
Experiment 1
In Experiment 1, college students participated in four retrieval
practice conditions: a study once condition, a study twice condi-
tion, study once followed by a fact quiz, and study once followed
by a higher order quiz. After two days, students completed fact and
higher order tests for each condition.
Method
Participants. Forty-eight college students (Mage ⫽20.58
years, 29 females) were recruited from the Department of Psychol-
ogy human subjects pool. Subjects received either credit toward
completion of a research participation requirement or cash pay-
ment ($25). Analyses were conducted only after data from 48
subjects were collected, a sample size determined at the outset of
the study using a power analysis with an assumed effect size of
d⫽0.5.
Design. A4⫻2 within-subject design was used, such that
four retrieval practice conditions (study once, study twice, fact
quiz, higher order quiz) were crossed with two delayed test types
(fact test, higher order test). Eight passages, two per retrieval
practice condition, were presented in the same order for all sub-
jects, but the order in which the conditions occurred was blocked
by retrieval practice condition and counterbalanced using a Latin
Square (see Appendix A). Retrieval practice conditions appeared
once in every ordinal position and were crossed with the two types
of final tests, creating eight counterbalancing orders. Six subjects
were randomly assigned to each of the eight orders. After a 2-day
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.
193
RETRIEVAL PRACTICE & BLOOM’S TAXONOMY

delay (i.e., 48 hr later), subjects completed one test type (a fact test
or a higher order test) per passage, with tests presented in the same
order in which passages were encountered during Session 1.
Materials. Eight passages were adapted from eight books
included in the “Taking Sides” McGraw-Hill Contemporary
Learning Series (http://www.mhcls.com;Daniel, 2006;Easton,
2006;Madaras & SoRelle, 1993;Moseley, 2007;Noll, 2001;Paul,
2002;Rourke, 1987). Each passage was approximately 1,000
words in length (M⫽1,006 words, range ⫽990 to 1016 words),
with half of each passage presenting one viewpoint of a contro-
versial topic, and the remaining half of each passage presenting the
opposite viewpoint (all passages are included in the online sup-
plementary material). For example, a passage entitled, “Does wel-
fare do more harm than good?” was adapted from Taking Sides:
Clashing Views on Controversial Social Issues (Finsterbusch &
McKenna, 1984), for which 500 words were drawn from the book
to describe a “yes” argument and approximately 500 words were
used to describe a “no” argument.
For Session 1, eight four-alternative multiple-choice fact ques-
tions and eight four-alternative multiple-choice higher order ques-
tions were developed for each passage (see Appendix B). For each
question type, approximately four questions pertained to the “yes”
argument and approximately four questions pertained to the “no”
argument. For Session 2, all question stems were rephrased and
multiple-choice alternatives were randomly reordered, but alterna-
tives were not rephrased. Across Sessions 1 and 2, regardless of
counterbalancing order, the correct multiple-choice alternative ap-
peared in every position (1, 2, 3, or 4) an equal number of times.
For fact questions, broad ideas stated in the passages were tested
to measure subjects’ overall understanding of the content. For
example, a fact question from the “Does welfare do more harm
than good?” passage included:
Which is the primary reason the “yes” author is against welfare
programs?
1. Welfare programs do not benefit recipients or taxpayers
2. Welfare programs create dependence for recipients
3. Welfare programs are too expensive for taxpayers
4. Welfare programs are not the government’s responsibility
The correct answer for this fact question is alternative #1, and
the answer was stated directly in the passage. Critically, all fact
questions in the present study were designed to encompass key
concepts or ideas from passages, rather than details such as names,
dates, vocabulary words, definitions, and so forth (e.g., the defi-
nition of welfare).
Higher order questions were developed in accordance with the
apply,analyze,evaluate, and create categories of Anderson et al.’s
(2001) revised Bloom’s taxonomy (see Figure 1). For apply ques-
tions, subjects were asked about a new situation or problem that
was related to a broad idea that was stated in the passage. For
example, an apply question from the welfare passage included:
What type of society would the “yes” author expect if there were no
welfare programs in the future?
1. A society in which all individuals are self-reliant and indepen-
dent
2. A society in which there would be no role for the government
3. A society in which no one would be required to pay taxes
4. A society in which all individuals are treated equally
The correct answer for this apply question is alternative #1 and
it could be inferred from the passage, but it was not stated explic-
itly.
For analyze questions, subjects were asked to differentiate the
authors’ arguments; they were presented with a statement and
asked which author (the “yes” author, the “no” author, both au-
thors, or neither author) would agree or disagree with the state-
ment. For example, an analyze question included:
Which author would agree with the following statement? “It is hon-
orable for the government to help society.”
1. The “yes” author
2. The “no” author
3. Both authors
4. Neither author
The correct answer for this analyze question is alternative #3.
For evaluate questions, subjects were asked to check or critique
an author’s argument by selecting a statement (which was not
presented in the passage) that most accurately summarized the
author’s argument. For example, an evaluate question included:
Which statement is an accurate evaluation or summary of the “yes”
author’s views?
1. Welfare programs can never work, because they are always too
expensive
2. Welfare programs are harmful, because they make bad situations
even worse
3. Welfare programs waste taxpayer money on people who do not
really need help
4. Welfare programs could work, but they rarely meet the needs of
the people
The correct answer for this evaluate question is alternative #4.
Lastly, for create questions, subjects were asked to plan or
predict an outcome for a novel situation that was not stated in the
passage; thus, the author’s potential reaction must be generated
based on information presented in the passage. For example, a
create question included:
How do you predict the “yes” author would react if he or she became
unemployed and needed welfare assistance?
1. The “yes” author might accept government assistance, but would
seek help from local organizations first
2. The “yes” author would not accept government assistance, but
would try to find a new job
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.
194 AGARWAL

3. The “yes” author might accept government assistance, but would
try to find a new job first
4. The “yes” author would not accept government assistance, but
would seek help from local organizations
The correct answer for this create question is alternative #2. All
reading passages and questions developed are included in the
online supplementary materials (see Appendix B for sample ques-
tions).
Procedure. Subjects were tested in small groups (up to five
people) using E-Prime 2.0 software (Schneider, Eschman, & Zuc-
colotto, 2007). At the beginning of Session 1, subjects were
instructed that they would be reading passages and taking
multiple-choice tests. Subjects were presented with a sample pas-
sage about the Nicaraguan Contras, drawn from the same book
series as experimental materials, for 20 seconds. They were
instructed to familiarize themselves with the computer program’s
scrolling feature (viewing the entire passage using the up and
down keys on the keyboard) without worrying about reading the
passage. Next, subjects were presented with a practice test of two
4-alternative multiple-choice questions (self-paced) that asked
subjects whether they turned off their cell phone and whether they
could return in two days; in other words, subjects did not receive
sample test questions related to the sample passage. After respond-
ing to the two practice questions, subjects were asked to make a
mental effort rating on a nine-point scale (adapted from Paas,
1992), which was followed by 10 seconds of feedback, to accli-
mate subjects to the experimental procedure.
After the instruction phase during Session 1, subjects completed
two blocks: First, subjects read all eight passages, presented in the
same order for all subjects. Second, subjects completed four
within-subject conditions (two per passage), blocked by retrieval
practice condition (see Appendix A). In other words during the
second block, subjects did not reencounter two of the passages (in
the study once condition), they read two of the passages for a
second time (in the study twice condition), they completed quizzes
with fact questions for two of the passages, and they completed
quizzes with higher order questions for the remaining two pas-
sages.
During 6-min study periods, each passage was presented in its
entirety on the computer screen and subjects were able to scroll up
and down to read the complete text at their own pace. Subjects
were asked to study the passage during the time allotted and after
six minutes, the computer moved on to the next passage (during
the first reading block) or to the appropriate condition (during the
second condition block following the restudy condition). Subjects’
keyboard presses were recorded during study periods to ensure that
all subjects scrolled appropriately through the passages from the
top of the passage to the bottom.
During self-paced initial quiz periods, multiple-choice questions
(blocked by passage) were presented one at a time, in a different
random order for each subject. Subjects were asked to type a
number (1, 2, 3, or 4) corresponding to the multiple-choice alter-
native (forced choice). As soon as subjects responded to each
question, the computer moved on (i.e., subjects were not allowed
to change their answer) and subjects were asked to estimate the
mental effort required (“How much mental effort did you invest in
this question”) on a 9-point scale (adapted from Paas, 1992)by
typing a number corresponding to the rating. After subjects rated
their mental effort, the computer presented immediate feedback for
10 seconds by displaying the word “CORRECT” or “INCORRECT”
corresponding to subjects’ response, while also displaying the
original question and the correct answer (without incorrect
multiple-choice lures). After 10 seconds, the computer moved on
to the next question. In other words, multiple-choice question
responses were self-paced, whereas feedback was experimenter-
controlled and presented for 10 seconds per item. In summary,
subjects completed a question, provided a mental effort rating, and
then viewed feedback, followed by the next question. After each
passage and quiz (regardless of condition), subjects received a 15
second break, during which the computer screen displayed,
“Please clear your mind and wait for the computer to move on.”
Then, the computer moved on to the next passage or condition,
according to subjects’ counterbalancing order.
After two days, subjects returned for Session 2 and completed
multiple-choice fact tests for four of the passages and multiple-
choice higher order tests for the other four passages. Testing proce-
dures outlined above for Session 1 were followed during Session 2,
except subjects did not receive feedback during Session 2.
In sum, subjects participated in four within-subject retrieval
practice conditions, crossed with two delayed test types. Depen-
dent variables measured included accuracy on test questions, re-
sponse times for test questions, mental effort ratings for test
questions, and response times for mental effort ratings. The entire
procedure lasted approximately two and a half hours across the
two sessions. At the end of the experiment, subjects were debriefed
and thanked for their time.
Results
All results in the current study were significant at an alpha level
of .05. A Greenhouse-Geisser correction was applied to analyses
of variance (ANOVAs) when the sphericity assumption was vio-
lated (Greenhouse & Geisser, 1959) and a Bonferroni correction
for multiple comparisons was applied to pvalues from ttests by
multiplying the pvalue by the number of comparisons (Rice,
1989). Effect sizes reported include partial eta-squared (␩p
2) for
ANOVAs (Pearson, 1911;Pierce, Block, & Aguinis, 2004) and
Cohen’s dfor ttests (Cohen, 1988). Error bars in figures represent
95% confidence intervals, specifically calculated for within-
subject designs according to methods described by Cousineau
(2005) and Morey (2008). Data from response times and mental
effort ratings did not contribute to the overall findings from the
present study, as discussed in the General Discussion. Thus, these
data are not reported and are available upon request.
Initial quiz performance. Initial performance on the fact quiz
and the higher order quiz is displayed in Table 2. As predicted,
initial performance was greater on the fact quiz (59%) compared
with performance on the higher order quiz (47%), likely because of
item difficulty, confirmed by a one-way ANOVA on initial per-
formance, F(1, 47) ⫽12.62, p⫽.001, ␩p
2⫽.21.
Final test performance. Final test performance on rephrased
questions for the four retrieval practice conditions is displayed in
Table 2 and Figure 2. Reliability (Cronbach’s alpha) was .501 for
final fact test performance and .467 for final higher order perfor-
mance. When collapsed over final test type, delayed test perfor-
mance was lower for the study once (49%) and study twice (51%)
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.
195
RETRIEVAL PRACTICE & BLOOM’S TAXONOMY

conditions, and greater for the fact quiz (62%) and higher order
quiz (62%) conditions. When collapsed over initial learning con-
dition, overall performance on the final fact test was greater (60%)
than performance on the final higher order test (53%), likely due
to item difficulty. A 4 (retrieval practice condition: study once,
study twice, fact quiz, higher order quiz) ⫻2 (delayed test type:
fact, higher order) repeated measures ANOVA on delayed test
performance indicated a main effect of retrieval practice condition,
F(3, 141) ⫽13.47, p⬍.001, ␩p
2⫽.22, a main effect of delayed
test type, F(1, 47) ⫽12.47, p⫽.001, ␩p
2⫽.21, and a significant
interaction between retrieval practice condition and test type, F(3,
141) ⫽27.39, p⬍.001, ␩p
2⫽.37.
Regarding delayed performance on the fact test, post hoc ttests
confirmed a significant effect of retrieval practice, such that final
fact test performance for the fact quiz condition (78%) was sig-
nificantly greater than final fact test performance for the study
once (54%) and study twice (54%) conditions, t(47) ⫽5.96, p⬍
.001, d⫽1.23 and t(47) ⫽6.63, p⬍.001, d⫽1.24, respectively.
However, final fact test performance was similar for the higher
order quiz condition (53%) compared with the study once and
study twice conditions, ts⬍1, indicating that retrieval practice
with higher order questions did not benefit final fact test perfor-
mance. In other words, an initial fact quiz improved final fact test
performance (78%) to a much greater degree than an initial higher
order quiz (53%), t(47) ⫽6.93, p⬍.001, d⫽1.29. There was no
effect of restudying when comparing final fact test performance
for the study once and study twice conditions, t⬍1.
Regarding delayed performance on the higher order test, post
hoc ttests also confirmed a significant effect of retrieval practice,
such that final higher order test performance for the higher order
quiz condition (72%) was significantly greater than final higher
order test performance for the study once (44%) and study twice
(49%) conditions, t(47) ⫽8.17, p⬍.001, d⫽1.39 and t(47) ⫽
5.31, p⬍.001, d⫽1.12, respectively. However, final higher order
test performance was similar for the fact quiz condition (46%)
compared with the study once and study twice conditions, ts⬍1,
indicating that retrieval practice with fact questions did not benefit
final higher order test performance. In other words, an initial
higher order quiz improved final higher order test performance
(72%) to a much greater degree than an initial fact quiz (46%),
t(47) ⫽6.73, p⬍.001, d⫽1.21. Again, there was no effect of
restudying on final higher order test performance when comparing
the study once and study twice conditions, t(47) ⫽1.40, p⬎.05.
In sum, initial retrieval practice enhanced final test performance,
but only when the initial quiz type (fact or higher order) matched
the final test type (fact or higher order, respectively). For these
congruent conditions, performance was marginally greater for the
fact quiz-fact test condition (78%) than for the higher order quiz-
higher order test condition (72%), t(47) ⫽1.94, p⫽.059, d⫽
0.32, though this difference may be due to relative difficulty
between fact versus higher order test questions.
Discussion
In Experiment 1, retrieval practice with higher order questions
greatly improved delayed higher order test performance by 23–
28% (compared with studying once or twice, Figure 2). Consistent
with prior research, retrieval practice with fact questions also
improved delayed fact performance by 24%. When the type of
initial quizzes matched the type of final test, even when final test
questions were rephrased, retrieval practice yielded comparable
benefits on performance for both fact and higher order learning.
Thus, results from Experiment 1 are consistent with predictions
based on the transfer appropriate processing framework (see
Table 1).
Critically, retrieval practice with fact questions did not enhance
delayed higher order test performance, contrary to the foundation
of factual knowledge framework. In addition, retrieval practice
with higher order questions did not enhance delayed fact test
performance, contrary to the desirable difficulty framework. There
were no benefits from restudying on delayed test performance,
even when the first and second study periods were spaced over
time, replicating previous findings (Agarwal et al., 2008;Callender
& McDaniel, 2009;Carrier & Pashler, 1992;Karpicke & Roedi-
ger, 2007;Roediger & Karpicke, 2006b;Wheeler, Ewers, & Buo-
nanno, 2003).
In sum, there were no benefits of initial fact quizzes on delayed
higher order test performance, no benefits of initial higher order
quizzes on delayed fact test performance, and no benefit of re-
studying on delayed test performance, regardless of test type. Of
Figure 2. Delayed test performance (proportion correct after two days) as
a function of retrieval practice condition from Experiment 1. Errors bars
represent 95% confidence intervals.
Table 2
Initial Quiz and Delayed Test Performance (Proportion Correct) as
a Function of Retrieval Practice Condition From Experiment 1
Condition Initial quiz
Final fact
test
Final higher
order test
Delayed
average
Study once .54 (.21) .44 (.18) .49
Study twice .54 (.21) .49 (.19) .51
Fact quiz .59 (.17) .78 (.19) .46 (.22) .62
Higher order quiz .47 (.15) .53 (.21) .72 (.21) .62
Average .53 .60 .53
Note. Standard deviations are displayed in parentheses.
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.
196 AGARWAL

the three theoretical frameworks, results from Experiment 1 most
closely align with the transfer appropriate processing framework.
Experiment 2
Experiment 2 was designed to examine whether a mix of both
fact and higher order question types used during retrieval practice
would be more beneficial for enhancing delayed test performance
than using one type of question during initial retrieval practice.
Considering the results from Experiment 1 where fact quizzes did
not enhance delayed higher order test performance and higher
order quizzes did not enhance delayed fact test performance,
retrieval practice with both types of questions during initial learn-
ing may benefit delayed test performance, regardless of test type.
When receiving a fact quiz and a higher order quiz in close
succession during the first session (i.e., in a mixed condition),
students may be more likely to activate elaborative information
(Carpenter, 2009), create more robust mediators or connections
between concepts (Pyc & Rawson, 2010), and subsequently trans-
fer knowledge from initial learning to a delayed higher order test.
In addition, providing students with prompts or hints to transfer
their knowledge to novel situations enhances their subsequent
transfer performance (Gick & Holyoak, 1980), thus a mixed quiz
condition may prompt students to transfer their knowledge during
initial learning and also on delayed tests (see also Butler et al.,
2017;Pan & Rickard, in press). From an instructional standpoint,
recent depictions of Bloom’s taxonomy in the shape of a wheel or
web eliminate the typical hierarchical structure in favor of an
integrated approach (see Knapp, 2016, for examples). Thus, mixed
complexity during initial learning may serve as an effective alter-
native to the prevalent “factual knowledge first” standpoint.
In Experiment 2, subjects participated in four retrieval practice
conditions, each after studying a passage once: they completed one
higher order quiz, two higher order quizzes, two fact quizzes, and
two mixed quizzes. After two days, subjects completed fact
and higher order tests for each condition. In Experiment 1, no-quiz
and restudy conditions did not improve final test performance.
Thus, these conditions were not included in Experiment 2. Instead,
the comparisons of interest in Experiment 2 were the optimal
combination of quizzes for improving delayed fact and higher
order learning—namely, two fact quizzes, two higher order quiz-
zes, or a mix of quizzes.
Given the results from Experiment 1, it was expected that the
higher order quiz (once or 1X) and higher order quiz (twice or 2X)
conditions would benefit delayed higher order test performance,
and that the fact quiz (twice or 2X) condition would benefit
delayed fact test performance. Regarding one quiz versus two
quizzes, it was predicted that two higher order quizzes would
provide an additional benefit to delayed higher order test perfor-
mance compared with one higher order quiz. On the other hand,
this additional benefit may be due to reexposure to the same item
twice; that is, question stems were only rephrased between the
initial and final sessions, not between the first and second initial
quizzes.
Regarding the mixed quiz condition (2X, with one fact and one
higher order quiz), it may be the case that including both question
types provides students with “the best of both worlds”—the mixed
quiz condition could enhance delayed performance on both types
of tests compared with the nonmixed retrieval practice conditions
(see Table 1). In line with the transfer appropriate processing
framework, engaging in both types of processing during initial
learning may have the greatest overlap in processing required for
the two final test types, enhancing delayed performance. At the
same time, one quiz of each format (in the mixed quiz condition)
may not prove as potent as two quizzes of the same format;
therefore, it was unclear whether the mixed quiz (2X) condition
would provide a smaller or larger benefit to delayed test perfor-
mance compared with the fact (2X) and higher order (2X) quiz
conditions.
Method
Participants. Forty-eight college students (Mage ⫽20.04
years, 31 females) were recruited from the Department of Psychol-
ogy human subjects pool. Subjects received either credit toward
completion of a research participation requirement or cash pay-
ment ($25). Subjects who participated in Experiment 2 did not
participate in Experiment 1. Analyses were conducted only after
data from 48 subjects were collected, a sample size determined at
the outset of the study using a power analysis with an assumed
effect size of d⫽0.5.
Design. A4⫻2 within-subject design was used, such that
four retrieval practice conditions [higher order quiz (1X), higher
order quizzes (2X), fact quizzes (2X), mixed quizzes (2X)] were
crossed with two delayed test types (fact test, higher order test).
Eight passages, two per retrieval practice condition, were pre-
sented in the same order for all subjects. The order in which the
conditions occurred was blocked by retrieval practice condition
and counterbalanced using a Latin Square (see Appendix A).
Retrieval practice conditions appeared once in every ordinal posi-
tion and were crossed with the two types of final tests, creating
eight counterbalancing orders. Six subjects were randomly as-
signed to each of the eight orders. Specifically for the mixed quiz
(2X) condition, subjects completed a fact quiz followed by a
higher order quiz, or they completed a higher order quiz followed
by a fact quiz. Order of quizzes in the mixed quiz condition was
counterbalanced equally across subjects (see Appendix A).
After a 2-day delay (i.e., 48 hr later), subjects completed one test
type (a fact test or a higher order test) per passage. Tests were
presented in the same order in which passages were encountered
during Session 1.
Materials. The same materials from Experiment 1 were used
in Experiment 2 (see Appendix B for sample questions).
Procedure. The same procedures used in Experiment 1 were
used in Experiment 2, except that subjects completed three blocks
during Session 1: First, subjects read all eight passages, presented
in the same order for all subjects. Second, subjects completed the
first quiz block with eight quizzes (one quiz per passage, presented
in the same order as passages during the reading block). Third,
subjects completed a second quiz block with six quizzes [one quiz
per passage, except for passages in the higher order quiz (1X)
condition, again presented in the same order]. After two days,
subjects returned for Session 2 and completed multiple-choice fact
tests for four of the passages and multiple-choice higher order tests
for the other four passages.
In sum, subjects participated in four within-subject retrieval
practice conditions, crossed with two delayed test types. Depen-
dent variables measured included accuracy on test questions, re-
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.
197
RETRIEVAL PRACTICE & BLOOM’S TAXONOMY

sponse times for test questions, mental effort ratings for test
questions, and response times for mental effort ratings. The entire
procedure lasted approximately two and a half hours across the
two sessions. At the end of the experiment, subjects were debriefed
and thanked for their time.
Results
Data from response times and mental effort ratings did not
contribute to the overall findings from the present study, as dis-
cussed in the General Discussion. Thus, these data are not reported
and are available upon request.
Initial quiz performance. Initial performance during the first
and second quiz blocks is displayed in Table 3. There was no effect
of counterbalancing order on initial quiz performance for the
mixed quiz (2X) condition (fact-higher order or higher order-fact),
F⬍1, therefore means were collapsed over initial order for
subsequent analyses (see Table 3 for the complete set of means).
For the first quiz block, initial performance was greatest for the
fact quiz (2X) condition (57%), followed by initial performance
for the mixed quiz (2X, 52%, collapsed over quiz order), higher
order quiz (2X, 49%), and higher order quiz (1X, 47%) conditions,
respectively. For the second quiz block, initial performance was
again greatest for the fact quiz (2X) condition (91%), followed by
the higher order quiz (2X, 83%) and mixed quiz (2X, 53%,
collapsed over quiz order) conditions.
A 3 [retrieval practice condition: higher order quiz (2X), fact
quiz (2X), mixed quiz (2X)] ⫻2 (quiz block: first, second)
repeated measures ANOVA on initial performance revealed a
significant main effect of retrieval practice condition, F(2, 94) ⫽
64.27, p⬍.001, ␩p
2⫽.58, a significant main effect of quiz block,
F(1, 47) ⫽356.69, p⬍.001, ␩p
2⫽.88, and a significant interac-
tion between retrieval practice condition and quiz block, F(2,
94) ⫽42.77, p⬍.001, ␩p
2⫽.48. As displayed in Table 3, the
higher order quiz (2X) and fact quiz (2X) conditions resulted in a
similar increase in performance from the first quiz block to the
second quiz block (34% for each condition).
Performance in the mixed quiz (2X) condition, on the other
hand, remained relatively constant across quiz blocks (see Table
3). Note that performance for each quiz block includes subjects’
performance on both types of quizzes (fact and higher order). This
finding suggests a replication of Experiment 1, namely that re-
trieval practice on one quiz format did not benefit performance on
a second quiz of a different format, even in close succession during
the first session—performance on the second quiz in the mixed
condition was similar to performance on the first quiz of the same
type in the fact quiz (2X) and higher order quiz (2X) conditions.
In general, the fact quiz (2X) performance resulted in substan-
tially greater performance during both the first and second quiz
blocks compared with the other initial learning conditions, likely
because of differences in item difficulty between fact and higher
order questions. Post hoc comparisons confirmed that the fact quiz
(2X) condition resulted in greater performance than the higher
order quiz (1X) and higher order quiz (2X) conditions on the first
quiz block, t(47) ⫽4.00, p⬍.001, d⫽0.71 and t(47) ⫽2.66, p⫽
.011, d⫽0.56, respectively, but fact quiz (2X) performance was
not significantly greater than mixed quiz (2X) performance on the
first quiz block, t(47) ⫽1.91, p⬎.05, likely because the mixed
quiz condition includes subjects whose first quiz was also a fact
quiz. On the second quiz block, the fact quiz (2X) condition
resulted in greater performance than the higher order quiz (2X) and
mixed quiz (2X) conditions, t(47) ⫽4.29, p⬍.001, d⫽0.77 and
t(47) ⫽15.66, p⬍.001, d⫽3.13, respectively.
Final test performance. Final test performance for the four
retrieval practice conditions is displayed in Table 3 and Figure 3.
Reliability (Cronbach’s alpha) was .462 for final fact test perfor-
mance and .254 for final higher order performance. There was no
effect of counterbalancing order on final test performance for the
mixed quiz (2X) condition (fact-higher order or higher order-fact),
F⬍1, therefore means were collapsed over counterbalance order
for subsequent analyses (see Table 3 for the complete set of
means).
As seen on the far right side of Table 3, delayed test perfor-
mance was greatest for the mixed quiz (2X) condition (75%),
compared with the fact quiz (2X, 69%), higher order quiz (2X,
69%), and higher order quiz (1X, 65%) conditions, respectively.
Overall performance for the two test types was similar: 69%
correct on the final fact test and 70% correct on the final higher
order test. A 4 [retrieval practice condition: higher order quiz (1X),
higher order quizzes (2X), fact quizzes (2X), mixed quizzes
(2X)] ⫻2 (delayed test type: fact, higher order) repeated measures
ANOVA on delayed test performance revealed a main effect of
retrieval practice condition, F(3, 141) ⫽4.85, p⫽.003, ␩p
2⫽.09,
and a significant interaction between retrieval practice condition
and delayed test type, F(3, 141) ⫽86.23, p⬍.001, ␩p
2⫽.65.
Regarding delayed performance on the fact test, post hoc ttests
confirmed that the fact quiz (2X) condition (90%) and the mixed
Table 3
Initial Quiz and Delayed Test Performance (Proportion Correct) as a Function of Retrieval
Practice Condition From Experiment 2
Condition First quiz Second quiz Final fact test
Final higher
order test
Delayed
average
Higher order quiz (1X) .47 (.11) .54 (.23) .77 (.17) .65
Higher order quizzes (2X) .49 (.14) .83 (.12) .53 (.22) .85 (.13) .69
Fact quizzes (2X) .57 (.17) .91 (.08) .90 (.13) .48 (.19) .69
Mixed quizzes (2X) .52 (.19) .53 (.15) .78 (.18) .71 (.18) .75
Mixed: Fact-higher .58 (.22) .47 (.11) .81 (.17) .71 (.18) .76
Mixed: Higher-fact .45 (.13) .60 (.15) .76 (.18) .71 (.19) .73
Average .53 .76 .69 .70
Note. Standard deviations are displayed in parentheses.
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.
198 AGARWAL

quiz (2X) condition (78%) resulted in greater delayed test perfor-
mance compared with the higher order quiz (1X, 54%) and the
higher order quiz (2X, 53%) conditions, ts⬎6.10, ps⬍.001, ds⬎
1.21. The difference in delayed fact test performance between the
fact quiz (2X) and mixed quiz (2X) conditions was also significant,
t(47) ⫽3.72, p⫽.006, d⫽0.77.
Regarding delayed performance on the higher order test, post
hoc ttests confirmed that the higher order quiz (2X, 85%), higher
order quiz (1X, 77%), and mixed quiz (2X, 71%) conditions
resulted in greater delayed test performance compared with the
fact quiz (2X) condition (48%), ts⬎5.80, ps⬍.001, ds⬎1.24.
The difference between the higher order quiz (2X) and the mixed
quiz (2X) conditions was also significant, t(47) ⫽4.52, p⬍.001,
d⫽0.84; however, neither of these two conditions differed sig-
nificantly from the higher order quiz (1X) condition, ps⬎.05.
Lastly, the difference in delayed performance between the con-
gruent conditions, namely delayed fact test performance for the
fact quiz (2X) condition (90%) and delayed higher order test
performance for the higher order quiz (2X) condition (85%), was
not significant, t(47) ⫽2.01, p⬎.05, and performance was close
to ceiling levels. The difference between the mixed quiz (2X)
condition on the delayed fact test (78%) versus the delayed higher
order test (71%) was marginally significant, t(47) ⫽2.08, p⫽
.088, d⫽0.39.
Consistent with Experiment 1, the congruent conditions (fact
quizzes-fact test, higher order quizzes-higher order test) resulted in
the greatest delayed test performance compared with the mixed
quiz (2X) condition, suggesting that two quizzes of the same
format are more potent for long-term learning than one quiz of
each format. Interestingly, the fact quiz (2X) condition still did not
benefit delayed higher order performance, even when compared
with only one initial higher order quiz, providing further evidence
that a boost in fact learning does not necessarily improve delayed
higher order performance.
Discussion
In Experiment 2, retrieval practice with two higher order quizzes
improved delayed higher order test performance by an additional
8% compared with only one higher order quiz (see Figure 3).
When the type of initial quizzes matched the type of final test,
retrieval practice yielded benefits for both fact and higher order
learning to a greater extent than one quiz of each format (in the
mixed quiz condition). Replicating Experiment 1, retrieval practice
with fact questions did not enhance delayed higher order test
performance, inconsistent with the foundation of factual knowl-
edge framework. In addition, retrieval practice with higher order
questions did not enhanced delayed fact test performance. The
findings from Experiment 2 provide further evidence that retrieval
practice is the most powerful when questions encountered during
initial quizzes are similar to questions on a final test, consistent
with the transfer appropriate processing framework.
Experiment 3
Experiment 3 was designed to investigate whether results from
Experiments 1 and 2 would replicate in an applied setting with a
different population (6th grade students) and a different content
domain (world history). In previous research with middle school
students, retrieval practice enhanced delayed test performance
compared with no quizzes, although information to be learned was
mainly fact-based (Agarwal et al., 2012;McDaniel, Agarwal,
Huelser, McDermott, & Roediger, 2011;McDaniel et al., 2013;
McDermott et al., 2014;Roediger et al., 2011). Do younger stu-
dents receive the same benefit from mixed quizzing as do college
students? It may be the case that, particularly for younger students,
the close succession of both question types within one quiz facil-
itates students’ building of connections across the two types of
processing, while also providing a prompt or hint to transfer
knowledge from fact questions to the immediate higher order
questions (Gick & Holyoak, 1980;Pan & Rickard, in press).
In Experiment 3, students completed mixed quizzes or higher
order-only quizzes, and learning was measured on final fact and
higher order tests. In accordance with prior research, retrieval
practice (regardless of quiz condition) was expected to enhance
both delayed fact and higher order test performance, compared
with delayed test performance on nonquizzed items. Based on
findings from Experiments 1 and 2, higher order quizzes were
expected to enhance delayed higher order test performance, but not
delayed fact test performance. Regarding mixed quizzes, the trans-
fer appropriate processing framework suggests a partial improve-
ment for both types of delayed tests, based on a partial overlap of
similar question types (see Table 1).
Method
Participants. One hundred forty-two 6th grade students (M⫽
24 students in each of six classroom sections; 71 males, 71
females) from a Midwestern suburban middle school participated
in the current experiment. Assent from each student was obtained
in accordance with guidelines from the Human Research Protec-
tion Office. Twelve students declined to include their data in the
study (although they participated in all classroom activities), and
data from eight special education students were excluded from
Figure 3. Delayed test performance (proportion correct after two days) as
a function of retrieval practice condition from Experiment 2. Errors bars
represent 95% confidence intervals.
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.
199
RETRIEVAL PRACTICE & BLOOM’S TAXONOMY

analyses because they received accommodations (e.g., additional
study and quiz opportunities outside of class).
Design. A3⫻2 within-subjects design was used, such that
three retrieval practice conditions (higher order quizzes, mixed
quizzes, nonquizzed) were crossed with two delayed test types
(fact test, higher order test). Conditions were manipulated across
two chapters of Social Studies material, with chapters presented in
the same order for all students (as determined by the classroom
teacher). Six classroom sections were split into two sets of three
class sections each; in other words, sections 1, 3, and 6 constituted
Set A, and sections 2, 5, and 7 constituted Set B. For the first
chapter, Set A students completed three quizzes with higher order
questions, whereas Set B students completed three quizzes with a
mix of question types. For the second chapter, the retrieval practice
conditions switched. At the end of each chapter unit (approxi-
mately 7– 8 school days in length; 48 hr after the third quiz),
students completed a final test comprised of both question types
(fact and higher order), with all questions presented in a different
random order for each of the six classroom sections. To maximize
power using the largest number of items per condition as possible,
while reducing classroom time required for the manipulation, a
restudy condition was not included in this experiment (prior re-
search demonstrated that retrieval practice enhanced delayed test
performance compared with a restudy exposure control in the same
school; McDermott et al., 2014;Roediger et al., 2011).
Materials. Two social studies textbook chapters (Russian
Revolution and World War II; Banks et al., 1997), assigned by the
classroom teacher, were used in this experiment. Each chapter was
approximately 2,350 words in length (e.g., 2,335 words for the
Russian Revolution chapter and 2,407 words for the World War II
chapter).
Twelve four-alternative multiple-choice fact questions and 12
four-alternative multiple-choice higher order questions were de-
veloped for each textbook chapter (see Appendix B for sample
questions). The classroom teacher approved all questions and
multiple-choice alternatives. Across all initial quizzes and delayed
tests, each classroom section received a different set of quizzed and
nonquizzed items, and every item was quizzed or not quizzed at least
once. In addition, for every initial quiz and final test, each classroom
section received a different random order of questions and the
multiple-choice alternatives were randomly reordered. The correct
multiple-choice alternative appeared in every position (A, B, C, or D)
an equal number of times across quizzes and tests.
Initial quizzes comprised either eight higher order questions or
a mix of fact and higher order questions (four of each type). For
mixed quizzes, question type (fact or higher order) was blocked
and order was counterbalanced across classroom sections and
quizzes, with questions presented in random order within question
type block. Questions that were not on initial quizzes (a non-
quizzed control condition) were covered in the textbook chapter
and also during the teacher’s lessons.
Final chapter tests comprised all multiple-choice fact and higher
order questions (12 fact and 12 higher order questions per chapter).
Final chapter test questions and alternatives were the same as those
from initial quizzes (i.e., questions were not rephrased) because of
the teacher’s concerns regarding floor effects and all 24 items were
presented in random order (not blocked by question type) on final
chapter tests. All reading passages and questions developed are
included in the online supplementary material.
For fact questions, broad concepts stated in the chapters were
tested to measure students’ overall understanding of the content.
For example, a fact question from the “Russian Revolution” text-
book chapter included:
Why was Nicholas II forced to give up his role as tsar?
A. Because the Duma elected a new tsar
B. Because Stalin took over the government
C. Because his wife and children moved to Moscow
D. Because of angry protestors, soldiers, and railroad workers
The correct answer for this fact question is alternative D and this
answer was stated directly in the textbook chapter. In contrast to
typical laboratory experiments, all fact questions in the present
study were drawn from authentic classroom material and designed
to encompass key concepts or ideas from the textbook chapters,
rather than details such as names, dates, vocabulary words, defi-
nitions, and so forth (e.g., the year in which the Russian Revolu-
tion began).
Higher order questions were developed based on categories
from a revised Bloom’s taxonomy (apply,analyze, and evaluate;
Anderson et al., 2001; see Figure 1). For example, an analyze
question from the “Russian Revolution” chapter included:
Which person would agree with the following statement? “Revolu-
tions are hard to prevent.”
A. Alexander II
B. Lenin
C. Nicholas II
D. Stalin
The correct answer for this analyze question is alternative C.
Higher order questions from the taxonomic create category were
not included in this experiment, due to the teacher’s concerns that
6th grade students may have difficulty extending textbook facts to
completely novel situations (i.e., floor effects).
Procedure. Students completed initial quizzes individually
via a clicker response system (Ward, 2007) in the classroom using
a computer, projector, and projection screen at the front of the
classroom. At the beginning of the study, students were instructed
that they would be taking quizzes (via clickers, with which stu-
dents were already familiar) and tests as part of a research study,
and that their scores may or may not count for a grade. In actuality,
students’ scores were not factored into their individual grades;
instead, the average score for each of the six classroom sections
counted toward a pizza party held at the end of the school year.
The classroom section with the highest score on each initial quiz or
final test received five points toward the pizza party. The class-
room section with the second highest score on each quiz or test
received four points toward the pizza party. Additional classroom
assignments and exams also factored into students’ pizza party
point totals, as determined by the classroom teacher.
In general, each chapter unit lasted approximately one week. For
each unit, students read a chapter from their social studies text-
book, listened to seven or eight lessons, participated in quizzes (the
experimental manipulation), and completed standard assignments
developed by the teacher.
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.
200 AGARWAL

Before the first lesson, students completed a prequiz via clickers
without having read the textbook chapter. Immediately after the
prequiz (i.e., on the same day), students began reading the chapter
and participated in the teacher’s corresponding lesson. After 2–3
school days, during which students completed the chapter reading
and the teacher covered all chapter content, students completed a
postquiz via clickers. Two days after the postquiz, the classroom
teacher reviewed all chapter material, which was followed by a
review quiz via clickers.
During all clicker quizzes (pre-, post-, and review quizzes),
multiple-choice questions were displayed on a projection screen at
the front of the classroom one at a time, in a different random order
for each classroom section. I read the question stem and four
multiple-choice alternatives aloud. After I was finished reading the
question and alternatives, I made the software’s response option
available and students were asked to press a letter (A, B, C, or D)
on their clicker remote corresponding to the multiple-choice alter-
native (forced choice). Once all students in the classroom re-
sponded (after approximately one minute), I closed the response
option and the clicker software displayed a green checkmark next
to the correct alternative (i.e., immediate feedback was adminis-
tered during quizzes). I read aloud the question stem and the
correct answer, and then moved on to the next question. Each
clicker quiz comprised eight questions, which took each class
section approximately seven to nine minutes to complete. Mental
effort ratings were not collected.
Two days after the review quiz, students completed a final test
comprised of all 24 multiple-choice questions for the chapter. Final
chapter tests were administered online (via Google Docs, http://
docs.google.com), whereas students sat individually at PC com-
puters. The chapter test was self-paced and students viewed each
multiple-choice question one at a time. Once students selected a
multiple-choice alternative, they moved on to the next question;
however, the online chapter test also allowed students to return to
earlier questions if they wanted to review or change their answers.
Once students responded to all 24 test questions, students were no
longer able to return to the test to change their answers. No
feedback was provided during the final chapter test.
In sum, students participated in three within-subject retrieval
practice conditions, crossed with two final test types. The depen-
dent variable of interest was accuracy on final test questions. The
entire procedure was followed for two textbook chapters over the
course of two weeks. At the end of the study, students were
debriefed and thanked for their time.
Results
Thirty-four students were absent during at least one initial quiz
or the final chapter test and their data were excluded from the
reported analyses to ensure the integrity of the experimental ma-
nipulation. Thus, data reported are from 88 students (Mage ⫽
11.58 years, 48 females); similar patterns of results were found
when data from all students who assented to participate were
included (i.e., n⫽122 absent and present students, excluding
special education students).
Note that middle school students were assigned to class sections
before the current study began and this assignment was nonran-
dom. Thus, data from students are nested within class section, and
again nested within set. Experiment 3 was carried out completely
within-subjects, all materials and conditions were completely
counterbalanced, and there was no significant difference in final
test performance between the two sets (p⫽.082). Even so, nested
individuals tend to be more alike than individuals selected at
random (Raudenbush & Bryk, 2002). Because the number of
levels within nests was low, the use of a multilevel model to
determine the influence of nonrandom assignment on performance
was not possible and means have not been adjusted to account for
this nesting.
Initial quiz performance. Initial quiz performance for the
first (prequiz), second (postquiz), and third (review) quizzes is
displayed in Table 4. In general, initial performance increased
from the prequiz (38%) to the postquiz (71%) and also to the
review quiz (84%), as a result of textbook reading, classroom
lessons, and immediate feedback received during the clicker quiz-
zes. Across the initial quizzes, performance was slightly greater in
the mixed quiz condition (66%) compared with performance in the
higher order quiz condition (62%), likely because of the inclusion
of fact questions in the mixed quiz condition.
A 2 (retrieval practice condition: higher order quizzes, mixed
quizzes) ⫻3 (quiz type: pre, post, review) repeated measures
ANOVA on initial quiz performance revealed a marginal main
effect of retrieval practice condition, F(1, 87) ⫽3.55, p⫽.063,
␩p
2⫽.039, and a significant main effect of quiz type, F(2, 174) ⫽
442.05, p⬍.001, ␩p
2⫽.84; however, the interaction was not
significant, F(2, 174) ⫽2.03, p⬎.05. In other words, students’
initial quiz performance increased across the three quizzes, and did
so similarly for the mixed quiz and the higher order quiz condi-
tions.
Final test performance. Performance on the final tests, ad-
ministered two days after the review quizzes, is displayed in Table
4and Figure 4. Reliability (Cronbach’s alpha) was .709 for final
fact test performance and .686 for final higher order performance.
A univariate ANOVA with set as a between-subjects factor (two
sets of three classroom sections each; see Appendix A) revealed no
significant difference on overall final test performance, F(1, 87) ⫽
3.12, p⫽.082, therefore means were collapsed over set for
subsequent analyses.
Table 4
Initial Quiz and Delayed Test Performance (Proportion Correct) as a Function of Retrieval Practice Condition From Experiment 3
Condition Pre-quiz Post-quiz Review quiz Final fact test
Final higher
order test Delayed average
Non-quizzed .64 (.18) .56 (.18) .60
Higher order quizzes .38 (.16) .68 (.21) .82 (.17) .64 (.20) .75 (.21) .70
Mixed quizzes .38 (.18) .73 (.19) .87 (.15) .91 (.17) .82 (.21) .86
Average .38 .71 .84 .73 .71
Note. Standard deviations are displayed in parentheses.
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.
201
RETRIEVAL PRACTICE & BLOOM’S TAXONOMY

Delayed test performance (collapsed over test type) was greatest
for the mixed quiz condition (86%), followed by the higher order
quiz (70%) and nonquizzed (60%) conditions. In addition, delayed
performance was similar for the final fact (73%) and final higher
order (71%) tests. A 3 (retrieval practice condition: higher order
quizzes, mixed quizzes, nonquizzed) ⫻2 (final test type: fact,
higher order) repeated measures ANOVA on final test perfor-
mance revealed a significant main effect of retrieval practice
condition, F(2, 174) ⫽128.98, p⬍.001, ␩p
2⫽.60, a marginal
main effect of final test type, F(1, 87) ⫽3.19, p⫽.078, ␩p
2⫽.04,
and a significant interaction between retrieval practice condition
and final test type, F(2, 174) ⫽28.30, p⬍.001, ␩p
2⫽.25.
For the final fact test, the mixed quiz condition resulted in far
greater performance (91%) than the higher order quiz and non-
quizzed conditions (64% each), t(47) ⫽12.24, p⬍.001, d⫽1.44
and t(47) ⫽13.63, p⬍.001, d⫽1.55, respectively. For the final
higher order test, the mixed quiz condition again resulted in the
greatest performance (82%) compared with the higher order (75%)
and nonquizzed (56%) conditions, t(87) ⫽2.27, p⫽.078 (p⫽
.026 without Bonferroni correction), d⫽0.34 and t(87) ⫽12.24,
p⬍.001, d⫽1.37, respectively. Consistent with Experiments 1
and 2, higher order retrieval practice led to significantly greater
higher order test performance compared with the nonquizzed con-
dition, t(87) ⫽7.87, p⬍.001, d⫽0.99.
Overall, mixed retrieval practice produced the greatest level of
performance on both fact and higher order final tests, while pro-
viding a marginal benefit above and beyond the benefit from
higher order retrieval practice on delayed higher order test perfor-
mance.
Discussion
Retrieval practice dramatically increased learning for middle
school students compared with no quizzes, contributing to a grow-
ing body of applied research (McDaniel et al., 2011,2013;Mc-
Dermott et al., 2014;Roediger et al., 2011). Remarkably, mixed
quizzes increased final fact test performance from the letter grade
equivalent ofaDtoanA⫺(a difference of 27%, Figure 4). Mixed
quizzes also produced a slight improvement over and above higher
order quizzes on delayed higher order test performance (7%),
although this difference was marginally significant. Replicating
Experiments 1 and 2, findings were consistent with the transfer
appropriate processing framework (see Table 1): benefits from
higher order quizzes were limited to the higher order test, and
benefits from the mixed quizzes extended to both types of final
tests.
General Discussion
Contrary to popular intuition, building a foundation of factual
knowledge via retrieval practice did not enhance students’ higher
order learning. Instead, students’ final fact test and higher order
test performance was greatest following retrieval practice that
matched in cognitive complexity based on Bloom’s taxonomy: fact
quizzes enhanced final fact test performance and higher order
quizzes enhanced final higher order test performance. Retrieval
practice increased learning by 20 –30% under laboratory condi-
tions with college students and also in an authentic K-12 class-
room.
Fact Quizzes Do Not Enhance Higher Order Learning
Why didn’t fact quizzes improve higher order learning in the
present study, as many cognitive scientists and educators contend?
First, students may have been unaware that information on fact
quizzes was related to final higher order tests, thus they did not
transfer their knowledge without explicit instructions to do so.
Chan, McDermott, and Roediger (2006, Experiment 3) found a
benefit of retrieval practice on novel items when subjects were
instructed to adopt a “broad retrieval strategy” during study,
whereas subjects who were told to adopt a “narrow retrieval
strategy” did not demonstrate a benefit of retrieval practice on
related novel items. Butler (2010, Experiment 3) also found a
benefit of retrieval practice on far transfer to novel items when
subjects were explicitly told that the final test was related to
information learned during the initial session (see also Chan,
2009). Furthermore, a classic study by Gick and Holyoak (1980)
demonstrated that students’ conceptual knowledge remains “inert”
when not explicitly told to use previously learned information on
novel items (see also Bransford, Sherwood, Vye, & Rieser, 1986;
Pan & Rickard, in press). Thus, students in the current study may
have transferred their factual knowledge to the higher order test
questions had they been given explicit instructions, prompts, or
hints.
Second, quiz and final test questions were multiple-choice, a
retrieval practice format that improves learning in laboratory and
classroom settings (Bjork, Little, & Storm, 2014;McDermott et
al., 2014;Roediger, Agarwal, Kang, & Marsh, 2010). Some edu-
cators argue that higher order learning cannot be facilitated or
measured using multiple-choice quizzes or tests. Instead, educators
often advocate for paper assignments, essay tests, open-book tests,
and ongoing portfolio evaluations to determine higher order learn-
ing (Ausubel et al., 1978;Hart, 1994;Kohn, 1999;Martinez,
1999). It is possible that by using multiple-choice items, an ele-
ment of genuine higher order learning may have been lost in the
Figure 4. Delayed test performance (proportion correct after two days) as
a function of retrieval practice condition from Experiment 3. Errors bars
represent 95% confidence intervals.
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.
202 AGARWAL

present study. Alternatively, multiple-choice quizzes may have
provided an added benefit from short answer quizzes because
well-constructed multiple-choice alternatives can facilitate re-
trieval of information pertaining to correct and incorrect alterna-
tives, as well as enhance transfer to novel information (Gierl,
Bulut, Guo, & Zhang, 2017;Little & Bjork, 2015;Little, Bjork,
Bjork, & Angello, 2012;Marsh & Cantor, 2014).
In the present study, multiple-choice alternatives were devel-
oped to require students to make fine-grained distinctions between
similar concepts. For instance, in Experiments 1 and 2, college
students were asked to evaluate an author’s views on welfare and
whether the government’s primary role is to advance morality,
security, equality, or liberty (see Appendix B). Each of these
multiple-choice alternatives related to the reading passage, the
target question, and current policy on welfare programs. As such,
the multiple-choice alternatives may have increased, rather than
decreased, potential transfer from fact to higher order questions.
Further examination of higher order learning with a variety of
question formats—including multiple-choice—may shed light on
whether students’ transfer of fact knowledge to higher order learn-
ing is dependent upon the type of questions used during retrieval
practice.
It is important to note that all multiple-choice alternatives re-
mained identical from initial retrieval practice to final tests for all
three experiments. In addition, Experiments 1 and 2 included
slightly rephrased final test questions, whereas Experiment 3 in-
cluded final test questions that were identical to initial quiz ques-
tions. Recent research has found that rephrased quiz or test ques-
tions do not facilitate transfer to a greater extent than identical
questions (Butler, 2010;Pan & Rickard, in press); thus, although
rephrased questions would be ideal for examining the flexibility of
learning, it is uncertain whether materials in the current study
diminished or eliminated transfer of fact knowledge. Further, if
memorization of quiz and test items were a concern, then one
would expect ceiling performance greater than 90% on final tests,
which was not found in the present experiments.
Third, cognitive load theory suggests that fact quizzes should
facilitate higher order learning by reducing cognitive demands
required during the final test (Plass et al., 2010;Sweller, 2010;van
Gog & Sweller, 2015). It was expected that mental effort ratings
on higher order tests would be lower when preceded by fact
quizzes compared with higher order quizzes (rated on a 9-point
scale, adapted from Paas, 1992; data available upon request).
Unfortunately, mental effort ratings did not shed light on this
puzzle. Across experiments and conditions, mental effort ratings
during higher order tests were lower when students first completed
higher order quizzes. In other words, students’ mental effort rat-
ings did not reveal sensitivity to a foundation of factual knowledge
on higher order learning.
The key finding that fact quizzes did not enhance higher order
learning directly contradicts the long-held belief that “factual
knowledge must precede skill” (Willingham, 2009). Instead, a
match between initial quiz and final test questions produced the
greatest learning in the current study, consistent with the transfer
appropriate processing framework (Morris et al., 1977; see Table
1, rows 1 and 2). In addition, discrepant conditions in the present
study (fact quiz-higher order test and higher order quiz-fact test;
see Table 1, rows 3 and 4) did not promote learning, contrary to
recent findings (Hinze & Wiley, 2011;Jensen et al., 2014;Mc-
Dermott et al., 2014). Whether a foundation of factual knowledge
promotes higher order learning—and under what conditions—
remains to be seen. Exploration of prompts to transfer and question
format may yield fact-based retrieval strategies that push student
learning higher on Bloom’s taxonomy and to greater levels of
complexity.
Mixed Quizzes Enhance Higher Order Learning
Mixed quizzes, comprising both fact and higher order questions,
increased higher order test performance more than fact quizzes (in
Experiment 2) and slightly more than higher order quizzes (in
Experiment 3). The robust benefits of mixed quizzes on higher order
learning is consistent with predictions from the three theoretical
frameworks of interest: the desirable difficulty framework suggests
that mixed quizzes pose a challenge for students by switching be-
tween question complexity during retrieval (see also Butler et al.,
2017); the transfer appropriate processing framework suggests a ben-
efit of mixed quizzes because of an overlap in processing with the
final test; and the foundation of factual knowledge framework pre-
dicts a benefit from mixed quizzes when students are given the
opportunity to engage in fact questions during initial learning (see
Table 1, rows 5 and 6).
Compared with higher order quizzes, why were mixed quizzes
more potent for higher order learning in Experiment 3 (middle school
students) than in Experiment 2 (college students)? Of course, students
of different ages may benefit differentially from mixed retrieval
practice. Varied complexity may be advantageous by providing a
scaffold between lower order and higher order questions, particularly
for students who have limited experience with complex materials (i.e.,
6th grade students in Experiment 3). Meanwhile, scaffolding from
varied complexity may be unnecessary for students who already have
experience extracting factual information from complex materials
(i.e., college students in Experiment 2); thus, older students may reap
more benefits from higher order retrieval practice. Although pure
speculation, the current study is the first to implement a retrieval
practice paradigm under similar conditions for two distinct student
populations. More research is necessary to ascertain whether factual
knowledge improves higher order learning, but also to ascertain
whether it differs for different age groups. Support for the foundation
of factual knowledge framework is frequently articulated from a K-12
perspective, an indication that the relationship between fact learning
and higher order learning might be unique for younger students.
Another major difference between Experiment 2 and Exper-
iment 3 was the procedure followed for mixed quizzes. In
Experiment 2, mixed quizzes were counterbalanced across nu-
merous within-subject conditions (e.g., students completed a
fact quiz on the welfare passage, next completed quizzes on
other passages, and then completed a higher order quiz on the
welfare passage; see Appendix A). In contrast, mixed quizzes in
Experiment 3 comprised fact and higher order questions con-
currently. Furthermore, within-subject conditions were admin-
istered one after another during a single session in Experiment
2, whereas middle school students in Experiment 3 spent one
week with one type of retrieval practice (higher order or mixed)
and then spent another week with the other type of retrieval
practice.
These procedural differences may have influenced students’
motivation to pay attention and learn the material, resulting in
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.
203
RETRIEVAL PRACTICE & BLOOM’S TAXONOMY

different benefits from mixed quizzing in the two experiments.
Prior research indicates that students prefer learning activities
that vary in terms of item difficulty, particularly in mathematics
(Skinner, Fletcher, Wildmon, & Belfiore, 1996;Wildmon,
Skinner, & McDade, 1998). Recent research also indicates that
students adjust their learning strategies based on expectancy for
different types of tests (Agarwal, D’Antonio, Roediger, McDer-
mott, & McDaniel, 2014;Agarwal & Roediger, 2011;Jensen et
al., 2014). Broadly, it would have been challenging for college
students to modulate their motivation specifically for the mixed
condition because of the extensive counterbalancing of pas-
sages and quizzes. It is possible that middle school students
exerted more motivational effort during the week in which they
had mixed quizzes compared with the week of higher order
quizzes, even though students were provided with an incentive
(a pizza party) based on quiz and test performance during both
weeks.
It is also unlikely that students in Experiments 2 and 3 would
have had greater motivation in the mixed quiz condition due to
the novelty of quiz items. To recognize which conditions were
mixed or not mixed, college students would have needed to
keep track of multiple-choice items across seven intervening
quizzes on unrelated passages, and middle school students
would have needed to ascertain the distinction between a
change in chapter material versus a change in the complexity of
quiz questions. Considering the amount of attention required to
do this in either experiment, the likelihood that the mixed quiz
condition increased student motivation as a result of item nov-
elty is doubtful. (I thank an anonymous reviewer for bringing
this possibility to my attention.)
When it comes to instruction, what type of retrieval practice
will help students achieve the highest levels of Bloom’s taxon-
omy? Surprisingly, fact-based retrieval practice only increased
fact learning, whereas higher order and mixed retrieval practice
increased higher order learning. If we want to reach the top of
Bloom’s taxonomy, building a foundation of knowledge via
fact-based retrieval practice may be less potent than engaging in
higher order retrieval practice at the outset, a key finding for
future research and classroom application.
References
Adesope, O. O., Trevisan, D. A., & Sundararajan, N. (2017). Rethinking
the use of tests: A meta-analysis of practice testing. Review of Educa-
tional Research, 87, 659 –701. http://dx.doi.org/10.3102/00346543
16689306
Agarwal, P. K., Bain, P. M., & Chamberlain, R. W. (2012). The value of
applied research: Retrieval practice improves classroom learning and
recommendations from a teacher, a principal, and a scientist. Educa-
tional Psychology Review, 24, 437– 448. http://dx.doi.org/10.1007/
s10648-012-9210-2
Agarwal, P. K., D’Antonio, L., Roediger, H. L., III, McDermott, K. B., &
McDaniel, M. A. (2014). Classroom-based programs of retrieval prac-
tice reduce middle school and high school students’ test anxiety. Journal
of Applied Research in Memory & Cognition, 3, 131–139. http://dx.doi
.org/10.1016/j.jarmac.2014.07.002
Agarwal, P. K., Finley, J. R., Rose, N. S., & Roediger, H. L., III. (2017).
Benefits from retrieval practice are greater for students with lower
working memory capacity. Memory, 25, 764 –771. http://dx.doi.org/10
.1080/09658211.2016.1220579
Agarwal, P. K., Karpicke, J. D., Kang, S. H. K., Roediger, H. L., III, &
McDermott, K. B. (2008). Examining the testing effect with open- and
closed-book tests. Applied Cognitive Psychology, 22, 861– 876. http://
dx.doi.org/10.1002/acp.1391
Agarwal, P. K., & Roediger, H. L., III. (2011). Expectancy of an open-
book test decreases performance on a delayed closed-book test. Memory,
19, 836 – 852. http://dx.doi.org/10.1080/09658211.2011.613840
Agarwal, P. K., Roediger, H. L., McDaniel, M. A., & McDermott, K. B.
(2017). How to use retrieval practice to improve learning. St. Louis,
MO: Washington University in St. Louis. Retrieved from http://www.
retrievalpractice.org
Anderson, L. W., Krathwohl, D. R., Airasian, P. W., Cruikshank, K. A.,
Mayer, R. E., Pintrich, P. R.,...Wittrock, M. C. (2001). A taxonomy for
learning, teaching, and assessing: A revision of Bloom’s taxonomy of
educational objectives (abridged ed.). New York, NY: Addison Wesley
Longman.
Ausubel, D. P. (1965). In defense of verbal learning. In R. Anderson & D.
Ausubel (Eds.), Readings in the psychology of cognition (pp. 87–102).
New York, NY: Holt, Rinehart, & Winston. (Original work published
1961)
Ausubel, D. P., Novak, J. D., & Hanesian, H. (1978). Educational psy-
chology: A cognitive view (2nd ed.). New York, NY: Holt, Rinehart, and
Winston.
Banks, J. A., Beyer, B. K., Contreras, G., Craven, J., Ladson-Billings, G.,
McFarland, M. A., & Parker, W. C. (1997). World: Adventures in time
and place. New York, NY: Macmillan/McGraw-Hill.
Barnett, S. M., & Ceci, S. J. (2002). When and where do we apply what we
learn? A taxonomy for far transfer. Psychological Bulletin, 128, 612–
637. http://dx.doi.org/10.1037/0033-2909.128.4.612
Bartlett, F. C. (1958). Thinking: An experimental and social study. West-
port, CT: Greenwood Press.
Bjork, E. L., Little, J. L., & Storm, B. C. (2014). Multiple-choice testing as
a desirable difficulty in the classroom. Journal of Applied Research in
Memory & Cognition, 3, 165–170. http://dx.doi.org/10.1016/j.jarmac
.2014.03.002
Bjork, R. A. (1994). Memory and metamemory considerations in the
training of human beings. In J. Metcalfe & A. Shimamura (Eds.),
Metacognition: Knowing about knowing (pp. 185–205). Cambridge,
MA: MIT Press.
Bloom, B. S. (Ed.), Engelhart, M. D., Furst, E. J., Hill, W. H., & Krath-
wohl, D. R. (1956). The taxonomy of educational objectives: The clas-
sification of educational goals (Handbook 1: Cognitive domain). New
York, NY: David McKay Company.
Bransford, J. D., Sherwood, R., Vye, N., & Rieser, J. (1986). Teaching
thinking and problem solving: Research foundations. American Psychol-
ogist, 41, 1078 –1089. http://dx.doi.org/10.1037/0003-066X.41.10.1078
Brown, P. C., Roediger, H. L., & McDaniel, M. A. (2014). Make it stick:
The science of successful learning. Cambridge, MA: Harvard University
Press. http://dx.doi.org/10.4159/9780674419377
Bruner, J. S. (1977). The process of education. Cambridge, MA: Harvard
University Press.
Butler, A. C. (2010). Repeated testing produces superior transfer of learn-
ing relative to repeated studying. Journal of Experimental Psychology:
Learning, Memory, and Cognition, 36, 1118 –1133. http://dx.doi.org/10
.1037/a0019902
Butler, A. C., Black-Maier, A. C., Raley, N. D., & Marsh, E. J. (2017).
Retrieving and applying knowledge to different examples promotes
transfer of learning. Journal of Experimental Psychology: Applied, 23,
433– 446. http://dx.doi.org/10.1037/xap0000142
Butler, A. C., Marsh, E. J., Slavinsky, J. P., & Baraniuk, R. G. (2014).
Integrating cognitive science and technology improves learning in a
STEM classroom. Educational Psychology Review, 26, 331–340. http://
dx.doi.org/10.1007/s10648-014-9256-4
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.
204 AGARWAL

Callender, A. A., & McDaniel, M. A. (2009). The limited benefits of
rereading educational texts. Contemporary Educational Psychology, 34,
30 – 41. http://dx.doi.org/10.1016/j.cedpsych.2008.07.001
Carpenter, S. K. (2009). Cue strength as a moderator of the testing effect:
The benefits of elaborative retrieval. Journal of Experimental Psychol-
ogy: Learning, Memory, and Cognition, 35, 1563–1569. http://dx.doi
.org/10.1037/a0017021
Carpenter, S. K. (2012). Testing enhances the transfer of learning. Current
Directions in Psychological Science, 21, 279 –283. http://dx.doi.org/10
.1177/0963721412452728
Carrier, M., & Pashler, H. (1992). The influence of retrieval on retention.
Memory & Cognition, 20, 633– 642. http://dx.doi.org/10.3758/
BF03202713
Chan, J. C. K. (2009). When does retrieval induce forgetting and when
does it induce facilitation? Implications for retrieval inhibition, testing
effect, and text processing. Journal of Memory and Language, 61,
153–170. http://dx.doi.org/10.1016/j.jml.2009.04.004
Chan, J. C. K., McDermott, K. B., & Roediger, H. L., III. (2006). Retrieval-
induced facilitation: Initially nontested material can benefit from prior
testing of related material. Journal of Experimental Psychology: Gen-
eral, 135, 553–571. http://dx.doi.org/10.1037/0096-3445.135.4.553
Cohen, J. (1988). Statistical power analysis for the behavioral sciences
(2nd ed.). Hillsdale, NJ: Erlbaum.
Cousineau, D. (2005). Confidence intervals in within-subject designs: A
simpler solution to Loftus and Masson’s method. Tutorials in Quanti-
tative Methods for Psychology, 1, 42– 45. http://dx.doi.org/10.20982/
tqmp.01.1.p042
Cuban, L. (1984). Policy and research dilemmas in the teaching of reason-
ing: Unplanned designs. Review of Educational Research, 54, 655– 681.
http://dx.doi.org/10.3102/00346543054004655
Daniel, E. L. (Ed.), (2006). Taking sides: Clashing views in health and
society (7th ed.). Dubuque, IA: McGraw-Hill Companies, Inc.
Dewey, J. (1944). Democracy and education: An introduction to the
philosophy of education. New York, NY: The Free Press. (Original work
published 1916)
Dunlosky, J., Rawson, K. A., Marsh, E. J., Nathan, M. J., & Willingham,
D. T. (2013). Improving students’ learning with effective learning tech-
niques: Promising directions from cognitive and educational psychol-
ogy. Psychological Science in the Public Interest, 14, 4 –58. http://dx
.doi.org/10.1177/1529100612453266
Easton, T. A. (Ed.), (2006). Taking sides: Clashing views on environmental
issues (11th ed.). Dubuque, IA: McGraw-Hill Companies, Inc.
Finsterbusch, K., & McKenna, G. (Eds.). (1984). Taking sides: Clashing
views on controversial social issues (3rd ed.). Guilford, CT: Dushkin
Publishing Group.
Gardiner, F. M., Craik, F. I. M., & Bleasdale, F. A. (1973). Retrieval
difficulty and subsequent recall. Memory & Cognition, 1, 213–216.
http://dx.doi.org/10.3758/BF03198098
Gick, M. L., & Holyoak, K. J. (1980). Analogical problem solving.
Cognitive Psychology, 12, 306 –355. http://dx.doi.org/10.1016/0010-
0285(80)90013-4
Gierl, M. J., Bulut, O., Guo, Q., & Zhang, X. (2017). Developing, analyz-
ing, and using distractors for multiple-choice tests in education: A
comprehensive review. Review of Educational Research, 87, 1082–
1116. http://dx.doi.org/10.3102/0034654317726529
Greenhouse, S. W., & Geisser, S. (1959). On methods in the analysis of
profile data. Psychometrika, 24, 95–112. http://dx.doi.org/10.1007/
BF02289823
Hart, D. (1994). Authentic assessment: A handbook for educators. Menlo
Park, CA: Addison Wesley Publishing Company.
Hinze, S. R., & Wiley, J. (2011). Testing the limits of testing effects using
completion tests. Memory, 19, 290 –304. http://dx.doi.org/10.1080/
09658211.2011.560121
Hirsch, E. D. (1996). The schools we need and why we don’t have them.
New York, NY: Doubleday.
Jacoby, L. L., Wahlheim, C. N., & Coane, J. H. (2010). Test-enhanced
learning of natural concepts: Effects on recognition memory, classifica-
tion, and metacognition. Journal of Experimental Psychology: Learning,
Memory, and Cognition, 36, 1441–1451. http://dx.doi.org/10.1037/
a0020636
James, W. (1900). Talks to teachers on psychology: And to students on
some of life’s ideals. New York, NY: Henry Holt and Company.
Jensen, J. L., McDaniel, M. A., Woodard, S. M., & Kummer, T. A. (2014).
Teaching to the test...ortesting to teach: Exams requiring higher order
thinking skills encourage greater conceptual understanding. Educational
Psychology Review, 26, 307–329. http://dx.doi.org/10.1007/s10648-013-
9248-9
Kang, S. H. K., McDermott, K. B., & Roediger, H. L., III. (2007). Test
format and corrective feedback modify the effect of testing on long-term
retention. European Journal of Cognitive Psychology, 19, 528 –558.
http://dx.doi.org/10.1080/09541440601056620
Karpicke, J. D., & Aue, W. R. (2015). The testing effect is alive and well
with complex materials. Educational Psychology Review, 27, 317–326.
http://dx.doi.org/10.1007/s10648-015-9309-3
Karpicke, J. D., & Blunt, J. R. (2011). Retrieval practice produces more
learning than elaborative studying with concept mapping. Science, 331,
772–775. http://dx.doi.org/10.1126/science.1199327
Karpicke, J. D., & Roediger, H. L., III. (2007). Repeated retrieval during
learning is the key to long-term retention. Journal of Memory and
Language, 57, 151–162. http://dx.doi.org/10.1016/j.jml.2006.09.004
Knapp, M. (2016, October 11). 5 gorgeous depictions of Bloom’s taxonomy
[Blog post]. Retrieved from https://news.nnlm.gov/nto/2016/10/11/5-
gorgeous-depictions-of-blooms-taxonomy/
Kohn, A. (1999). The schools our children deserve: Moving beyond tra-
ditional classrooms and “tougher standards.” Boston, MA: Houghton
Mifflin Company.
Kornell, N., & Bjork, R. A. (2008). Learning concepts and categories: Is
spacing the “enemy of induction”? Psychological Science, 19, 585–592.
http://dx.doi.org/10.1111/j.1467-9280.2008.02127.x
Kromann, C. B., Jensen, M. L., & Ringsted, C. (2009). The effect of testing
on skills learning. Medical Education, 43, 21–27. http://dx.doi.org/10
.1111/j.1365-2923.2008.03245.x
Larsen, D. P., Butler, A. C., Lawson, A. L., & Roediger, H. L., III. (2013).
The importance of seeing the patient: Test-enhanced learning with
standardized patients and written tests improves clinical application of
knowledge. Advances in Health Sciences Education, 18, 409 – 425.
http://dx.doi.org/10.1007/s10459-012-9379-7
Lemov, D. (2017, April 3). Bloom’s taxonomy: That pyramid is a problem
[Blog post]. Retrieved from http://teachlikeachampion.com/blog/
blooms-taxonomy-pyramid-problem/
Little, J. L., & Bjork, E. L. (2015). Optimizing multiple-choice tests as
tools for learning. Memory & Cognition, 43, 14 –26. http://dx.doi.org/
10.3758/s13421-014-0452-8
Little, J. L., Bjork, E. L., Bjork, R. A., & Angello, G. (2012). Multiple-
choice tests exonerated, at least of some charges: Fostering test-induced
learning and avoiding test-induced forgetting. Psychological Science,
23, 1337–1344. http://dx.doi.org/10.1177/0956797612443370
Lyle, K. B., & Crawford, N. A. (2011). Retrieving essential material at the
end of lectures improves performance on statistics exams. Teaching of
Psychology, 38, 94 –97. http://dx.doi.org/10.1177/0098628311401587
Madaras, L., & SoRelle, J. M. (Eds.). (1993). Taking sides: Clashing views
on controversial issues in American history (5th ed., Vol. 1). Guilford,
CT: Dushkin Publishing Group.
Marsh, E. J., & Cantor, A. D. (2014). Learning from the test: Dos and
don’ts for using multiple-choice tests. In M. A. McDaniel, R. F. Frey,
S. M. Fitzpatrick, & H. L. Roediger (Eds.), Integrating cognitive science
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.
205
RETRIEVAL PRACTICE & BLOOM’S TAXONOMY

with innovative teaching in STEM disciplines (pp. 37–52). St. Louis,
MO: Washington University Libraries.
Martinez, M. E. (1999). Cognition and the question of test item format.
Educational Psychologist, 34, 207–218. http://dx.doi.org/10.1207/
s15326985ep3404_2
McDaniel, M. A., Agarwal, P. K., Huelser, B. J., McDermott, K. B., &
Roediger, H. L. (2011). Test-enhanced learning in a middle school
science classroom: The effects of quiz frequency and placement. Journal
of Educational Psychology, 103, 399 – 414. http://dx.doi.org/10.1037/
a0021782
McDaniel, M. A., Friedman, A., & Bourne, L. E. (1978). Remembering the
levels of information in words. Memory & Cognition, 6, 156 –164.
http://dx.doi.org/10.3758/BF03197441
McDaniel, M. A., Roediger, H. L., III, & McDermott, K. B. (2007).
Generalizing test-enhanced learning from the laboratory to the class-
room. Psychonomic Bulletin & Review, 14, 200 –206. http://dx.doi.org/
10.3758/BF03194052
McDaniel, M. A., Thomas, R. C., Agarwal, P. K., McDermott, K. B., &
Roediger, H. L. (2013). Quizzing in middle-school science: Successful
transfer performance on classroom exams. Applied Cognitive Psychol-
ogy, 27, 360 –372. http://dx.doi.org/10.1002/acp.2914
McDermott, K. B., Agarwal, P. K., D’Antonio, L., Roediger, H. L., III, &
McDaniel, M. A. (2014). Both multiple-choice and short-answer quizzes
enhance later exam performance in middle and high school classes.
Journal of Experimental Psychology: Applied, 20, 3–21. http://dx.doi
.org/10.1037/xap0000004
Mehta, J. (2018, January 4). A pernicious myth: Basics before deeper
learning [Blog post]. Retrieved from http://blogs.edweek.org/edweek/
learning_deeply/2018/01/a_pernicious_myth_basics_before_deeper_
learning.html
Morey, R. D. (2008). Confidence intervals from normalized data: A cor-
rection to Cousineau (2005). Tutorials in Quantitative Methods for
Psychology, 4, 61– 64. http://dx.doi.org/10.20982/tqmp.04.2.p061
Morris, C. D., Bransford, J. D., & Franks, J. J. (1977). Levels of processing
versus transfer-appropriate processing. Journal of Verbal Learning &
Verbal Behavior, 16, 519 –533. http://dx.doi.org/10.1016/S0022-
5371(77)80016-9
Moseley, W. G. (Ed.), (2007). Taking sides: Clashing views on African
issues (2nd ed.). Dubuque, IA: McGraw-Hill Companies, Inc.
Münsterberg, H. (1909). Psychology and the teacher. New York, NY: D.
Appleton and Company.
National Research Council. (1987). Education and learning to think.
Washington, DC: The National Academies Press.
Noll, J. W. (Ed.), (2001). Taking sides: Clashing views on controversial
educational issues (11th ed.). Guilford, CT: Dushkin/McGraw-Hill.
Paas, F. G. W. C. (1992). Training strategies for attaining transfer of
problem-solving skill in statistics: A cognitive-load approach. Journal of
Educational Psychology, 84, 429 – 434. http://dx.doi.org/10.1037/0022-
0663.84.4.429
Pan, S. C., & Agarwal, P. K. (2018). Retrieval practice and transfer of
learning: Fostering students’ application of knowledge. San Diego, CA:
University of California at San Diego. Retrieved from http://www
.retrievalpractice.org
Pan, S. C., & Rickard, T. C. (in press). Transfer of test-enhanced learning:
Meta-analytic review and synthesis. Psychological Bulletin.
Pashler, H., Bain, P. M., Bottge, B. A., Graesser, A., Koedinger, K.,
McDaniel, M., & Metcalfe, J. (2007). Organizing instruction and study
to improve student learning. Washington, DC: National Center for
Education Research, Institute of Education Sciences, U.S. Department of
Education. Retrieved from http://ncer.ed.gov.http://dx.doi.org/10.1037/
e607972011-001
Paul, E. L. (Ed.), (2002). Taking sides: Clashing views on controversial
issues in sex and gender (2nd ed.). Guilford, CT: McGraw-Hill/Dushkin.
Pearson, K. (1911). On a correction needful in the case of the correlation
ratio. Biometrika, 8, 254 –256. http://dx.doi.org/10.2307/2331454
Pierce, C. A., Block, R. A., & Aguinis, H. (2004). Cautionary note on
reporting eta-squared values from multifactor ANOVA designs. Educa-
tional and Psychological Measurement, 64, 916 –924. http://dx.doi.org/
10.1177/0013164404264848
Plass, J. L., Moreno, R., & Brünken, R. (Eds.). (2010). Cognitive load
theory. New York, NY: Cambridge University Press. http://dx.doi.org/
10.1017/CBO9780511844744
Pyc, M. A., Agarwal, P. K., & Roediger, H. L. (2014). Test-enhanced learning. In
V. A. Benassi, C. E. Overson, & C. M. Hakala (Eds.), Applying science
of learning in education: Infusing psychological science into the cur-
riculum. Washington, DC: APA Society for the Teaching of Psychology.
Pyc, M. A., & Rawson, K. A. (2009). Testing the retrieval effort hypoth-
esis: Does greater difficulty correctly recalling information lead to
higher levels of memory? Journal of Memory and Language, 60, 437–
447. http://dx.doi.org/10.1016/j.jml.2009.01.004
Pyc, M. A., & Rawson, K. A. (2010). Why testing improves memory:
Mediator effectiveness hypothesis. Science, 330, 335. http://dx.doi.org/
10.1126/science.1191465
Raudenbush, S. W., & Bryk, A. S. (2002). Hierarchical linear models:
Applications and data analysis methods (2nd ed.). Thousand Oaks, CA:
Sage.
Ravitch, D. (2009, September 15). Critical thinking? You need knowl-
edge. The Boston Globe. Retrieved from http://archive.boston.com/
bostonglobe/editorial_opinion/oped/articles/2009/09/15/critical_
thinking_you_need_knowledge/
Rawson, K. A. (2015). The status of the testing effect for complex mate-
rials: Still a winner. Educational Psychology Review, 27, 327–331.
http://dx.doi.org/10.1007/s10648-015-9308-4
Rice, W. R. (1989). Analyzing tables of statistical tests. Evolution; Inter-
national Journal of Organic Evolution, 43, 223–225. http://dx.doi.org/
10.1111/j.1558-5646.1989.tb04220.x
Roediger, H. L., Agarwal, P. K., Kang, S. H. K., & Marsh, E. J. (2010).
Benefits of testing memory: Best practices and boundary conditions. In
G. M. Davies & D. B. Wright (Eds.), New frontiers in applied memory
(pp. 13– 49). Brighton, UK: Psychology Press.
Roediger, H. L., Agarwal, P. K., McDaniel, M. A., & McDermott, K. B.
(2011). Test-enhanced learning in the classroom: Long-term improve-
ments from quizzing. Journal of Experimental Psychology: Applied, 17,
382–395. http://dx.doi.org/10.1037/a0026252
Roediger, H. L., III, & Karpicke, J. D. (2006a). The power of testing
memory: Basic research and implications for educational practice. Per-
spectives on Psychological Science, 1, 181–210. http://dx.doi.org/10
.1111/j.1745-6916.2006.00012.x
Roediger, H. L., III, & Karpicke, J. D. (2006b). Test-enhanced learning:
Taking memory tests improves long-term retention. Psychological Sci-
ence, 17, 249 –255. http://dx.doi.org/10.1111/j.1467-9280.2006.01693.x
Rohrer, D., & Pashler, H. (2010). Recent research on human learning
challenges conventional instructional strategies. Educational Re-
searcher, 39, 406 – 412. http://dx.doi.org/10.3102/0013189X103
74770
Rohrer, D., & Taylor, K. (2006). The effects of overlearning and distrib-
uted practise on the retention of mathematics knowledge. Applied Cog-
nitive Psychology, 20, 1209 –1224. http://dx.doi.org/10.1002/acp.1266
Rohrer, D., Taylor, K., & Sholar, B. (2010). Tests enhance the transfer of
learning. Journal of Experimental Psychology: Learning, Memory, and
Cognition, 36, 233–239. http://dx.doi.org/10.1037/a0017678
Rourke, J. T. (Ed.), (1987). Taking sides: Clashing views on controversial
issues in world politics (1st ed.). Guilford, CT: Dushkin Publishing
Group.
Rowland, C. A. (2014). The effect of testing versus restudy on retention: A
meta-analytic review of the testing effect. Psychological Bulletin, 140,
1432–1463. http://dx.doi.org/10.1037/a0037559
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.
206 AGARWAL

Schneider, W., Eschman, A., & Zuccolotto, A. (2007). E-prime 2 user’s
guide. Pittsburgh, PA: Psychology Software Tools.
Skinner, C. H., Fletcher, P. A., Wildmon, M., & Belfiore, P. J. (1996).
Improving assignment preference through interspersing additional prob-
lems: Brief versus easy problems. Journal of Behavioral Education, 6,
427– 436. http://dx.doi.org/10.1007/BF02110515
Sweller, J. (2010). Cognitive load theory: Recent theoretical advances. In
J. Plass, R. Moreno, & R. Brünken (Eds.), Cognitive load theory (pp.
29 –47). New York, NY: Cambridge University Press. http://dx.doi.org/
10.1017/CBO9780511844744.004
Taylor, K., & Rohrer, D. (2010). The effects of interleaved practice. Applied
Cognitive Psychology, 24, 837– 848. http://dx.doi.org/10.1002/acp.1598
van Gog, T., & Sweller, J. (2015). Not new, but nearly forgotten: The
testing effect decreases or even disappears as the complexity of
learning materials increases. Educational Psychology Review, 27,
247–264. http://dx.doi.org/10.1007/s10648-015-9310-x
Ward, D. (2007). eInstruction: Classroom performance system [Computer
software]. Denton, Texas: EInstruction Corporation.
Wheeler, M. A., Ewers, M., & Buonanno, J. F. (2003). Different rates of
forgetting following study versus test trials. Memory, 11, 571–580.
http://dx.doi.org/10.1080/09658210244000414
Wildmon, M. E., Skinner, C. H., & McDade, A. (1998). Interspersing
additional brief, easy problems to increase assignment preference on
mathematics reading problems. Journal of Behavioral Education, 8,
337–346. http://dx.doi.org/10.1023/A:1022823314635
Willingham, D. T. (2009). Why don’t students like school: A cognitive
scientist answers questions about how the mind works and what it means
for the classroom. San Francisco, CA: Jossey-Bass.
Appendix A
Counterbalancing Orders
Experiment 1
Welfare Vaccines Multicul Biotech SexDiff Lincoln Superfund WWII
1, 2 Study Twice Study Twice Fact Fact Study Once Study Once Higher Order Higher Order
3, 4 Study Once Study Once Higher Order Higher Order Fact Fact Study Twice Study Twice
5, 6 Higher Order Higher Order Study Once Study Once Study Twice Study Twice Fact Fact
7, 8 Fact Fact Study Twice Study Twice Higher Order Higher Order Study Once Study Once
Note. Odd counterbalancing orders received final fact tests first, alternating with final higher order tests. Even orders received final higher order tests first,
alternating with final fact tests.
Experiment 2
Welfare Vaccines Multicul Biotech SexDiff Lincoln Superfund WWII
1, 2 Fact 2X Fact 2X Mixed (H-F) Mixed (H-F) Higher 1X Higher 1X Higher 2X Higher 2X
3, 4 Mixed (F-H) Mixed (F-H) Fact 2X Fact 2X Higher 2X Higher 2X Higher 1X Higher 1X
5, 6 Higher 1X Higher 1X Higher 2X Higher 2X Mixed (H-F) Mixed (H-F) Fact 2X Fact 2X
7, 8 Higher 2X Higher 2X Higher 1X Higher 1X Fact 2X Fact 2X Mixed (F-H) Mixed (F-H)
Note. Odd counterbalancing orders received final fact tests first, alternating with final higher order tests. Even orders received final higher order tests first,
alternating with final fact tests.
Experiment 3
Russian Revolution World War II
Set A (three class sections) Higher Order Only Fact ⫹Higher Order Mix
Set B (three class sections) Fact ⫹Higher Order Mix Higher Order Only
(Appendices continue)
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.
207
RETRIEVAL PRACTICE & BLOOM’S TAXONOMY

Appendix B
Sample Questions
(Appendices continue)
Experiments 1 and 2
Session Fact question Higher order question
Retrieval
practice
Which is the primary reason the “yes” author is against
welfare programs?
Which statement is an accurate evaluation or summary of the “yes”
author’s views?
1. Welfare programs don’t benefit recipients or taxpayers 1. Welfare programs can never work, because they are always
too expensive
2. Welfare programs create dependence for recipients 2. Welfare programs are harmful, because they make bad
situations even worse
3. Welfare programs are too expensive for taxpayers 3. Welfare programs could work, but they rarely meet the needs
of the people
4. Welfare programs are not the government’s responsibility 4. Welfare programs waste taxpayer money on people who don’t
really need help
Final test The “yes” author is against welfare programs, largely because The “yes” author would agree with which statement?
1. Welfare programs are too expensive for taxpayers 1. Welfare programs could work, but they rarely meet the needs
of the people
2. Welfare programs don’t benefit recipients or taxpayers 2. Welfare programs waste taxpayer money on people who don’t
really need help
3. Welfare programs are not the government’s responsibility 3. Welfare programs can never work, because they are always
too expensive
4. Welfare programs create dependence for recipients 4. Welfare programs are harmful, because they make bad
situations even worse
Retrieval
practice
The “no” author argues that vaccines may always carry some
amount of risk, but that this risk
Which author would agree with the following statement? “The
ends justify the means.”
1. Is a possibility with any medical procedure 1. The “yes” author
2. Is too small to be of concern to the community 2. The “no” author
3. Should be of concern to scientists, not parents 3. Both authors
4. Is less than the likelihood of a disease epidemic 4. Neither author
Final test The “no” author argues that vaccine risk Which author would agree with the following statement? “The
achieved outcome is more important than the process along
the way.”
1. Should be of concern to scientists, not parents 1. The “no” author
2. Is a possibility with any medical procedure 2. Neither author
3. Is less than the likelihood of a disease epidemic 3. Both authors
4. Is too small to be of concern to the community 4. The “yes” author
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.
208 AGARWAL

Received June 14, 2017
Revision received March 7, 2018
Accepted March 7, 2018 䡲
Experiment 3
Session Fact question Higher order question
Retrieval practice
and final test
Why were Nicholas II and the Duma in constant
conflict?
Based on what you know about Nicholas II, how would he treat poor
people?
A. Because the Duma wanted to help the poor A. He would share some power with the poor
B. Because Nicholas II wanted to help the poor B. He would help the poor
C. Because the Duma wanted to support communism C. He would take money away from the poor
D. Because Nicholas II wanted control of all of
Russia’s power
D. He would ignore the poor
Retrieval practice
and final test
Under Stalin, how would you describe everyday life for
the Russian people?
Which person would agree with the following statement? “People are
the most productive when they are told what to do by one person,
instead of listening to many people or doing what they want.”
A. Stalin controlled all aspects of people’s lives A. Nicholas II
B. People were free to do whatever they wanted B. Lenin
C. Stalin forced all people to go to church C. Stalin
D. People were allowed to choose their careers D. Alexander II
Retrieval practice
and final test
What did Franklin Roosevelt do during World War II? Based on what you know about Franklin Roosevelt, what would he do if
Spain attacked the U.S.?
A. He dropped an atomic bomb on Japan A. He would surrender to Spain
B. He killed Adolf Hitler B. He would negotiate with Spain
C. He joined the Axis war effort C. He would attack Spain in return
D. He declared war on Japan D. He would drop an atomic bomb on Spain
Retrieval practice
and final test
Why did Hitler join forces with Japan? Which statement is an accurate summary of Hitler’s views?
A. So they could work together to expand their
empires
A. By invading the Soviet Union, Germany can increase food
production
B. So they could both take over the United States B. By invading the Soviet Union, Germany can create a master race
C. So Germany could build an army base in Japan C. By invading the Soviet Union, Germany can expand its empire
D. So Japan wouldn’t join the Allied Forces D. By invading the Soviet Union, Germany can strengthen its military
Note. Multiple-choice alternatives remained identical across initial quizzes and final tests for all experiments. The correct answers for each question are
italicized. All materials are included in the online supplementary material.
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.
209
RETRIEVAL PRACTICE & BLOOM’S TAXONOMY