untitled
Research Article
Test-Enhanced Learning
Taking Memory Tests Improves Long-Term Retention
Henry L. Roediger, III, and Jeffrey D. Karpicke
Washington University in St. Louis
ABSTRACT—Taking a memory test not only assesses what
one knows, but also enhances later retention, a phenome-
non known as the testing effect. We studied this effect with
educationally relevant materials and investigated whethe
testing facilitates learning only because tests offer an op-
portunity to restudymaterial. In two experiments, students
studied prose passages and took one or three immediate
free-recall tests, without feedback, or restudied the mate-
ial the same number of times as the students who received
tests. Students then took a final retention test 5 min, 2 days,
or 1 week later. When the final test was given after 5 min,
epeated studying improved recall relative to repeated
testing. However, on the delayed tests, prior testing pro-
duced substantially greater retention than studying, even
though repeated studying increased students’ confidence in
their ability to remember thematerial. Testing is a powerful
means of improving learning, not just assessing it.
In educational settings, tests are usually considered devices of
assessment. Students take tests in class to assess what they have
learned and take standardized tests like the SAT to assess thei
knowledge and aptitude. In many circumstances, such as uni-
versity lecture courses, tests are given infrequently (often just
two or three times a semester) and are generally perceived as a
other by faculty and students alike. We believe that the neglect
of testing in all levels of education is misguided. To state an
obvious point, if students know they will be tested regularly (say,
once a week, or even every class period), they will study more
and will space their studying throughout the semester rathe
than concentrating it just before exams (see Bangert-Drowns,
Kulik, & Kulik, 1991; Leeming, XXXXXXXXXXHowever, more impor-
tant for present purposes, testing has a powerful positive effect
on future retention. If students are tested on material and suc-
cessfully recall or recognize it, they will remember it better in
the future than if they had not been tested. This phenomenon,
called the testing effect, has been studied sporadically over a
long period of time (e.g., Gates, 1917), but is not well known
outside cognitive psychology.
Most experiments on the testing effect have been conducted in
the ve
al learning tradition using word lists (e.g., Hogan &
Kintsch, 1971; Izawa, 1967; McDaniel & Masson, 1985;
Thompson, Wenger, & Bartling, 1978; Tulving, 1967; Wheeler,
Ewers, &Buonanno, 2003) or picture lists (Wheeler &Roediger,
1992) as materials. There have been a few experiments using
materials found in educational contexts, beginning with Spitze
(1939; see too Glover, 1989, and McDaniel & Fisher, 1991).
However, the title of Glover’s article from 17 years ago still sums
up the cu
ent state of affairs: ‘‘The ‘testing’ phenomenon: Not
gone but nearly forgotten.’’
Our aim in the two experiments reported herewas to investigate
the testing effect under educationally relevant conditions, using
prose materials and free-recall tests without feedback (somewhat
akin to essay tests used in education). Most previous research has
used tests involving recognition (like multiple-choice tests) o
cued recall (like short-answer tests). A second purpose of ou
experiments was to determine whether testing facilitates learning
eyond the benefits of restudying the material. In some testing-
effect experiments, a study-test condition is compared with a
study-only condition on a delayed retention test. When the sub-
jects in the former condition outperform those in the latter on a
final test, one can wonder whether the testing effect is simply due
to study-test subjects being reexposed to the material during the
test. It is no surprise that students will learn more with two
presentations of material rather than one (although some of the
word-list experiments cited earlier overcame this problem; see too
Ca
ier & Pashler, 1992; Cull, XXXXXXXXXXTo evaluate this restudying
explanation of the testing effect, we had students in our control
conditions restudy the entire set of material—which should, if
anything, bias performance results in favor of this condition,
ecause students who take free-recall tests (without feedback)
can only reexperience whatever material they can recall.
Students in our experiments studied short prose passages
covering general scientific topics. In Experiment 1, they eithe
Address co
espondence to Henry L. Roediger, III, Department of
Psychology, Washington University, Campus Box 1125, One Brook-
ings Dr., St. Louis, MO 63130, e-mail: XXXXXXXXXX.
PSYCHOLOGICAL SCIENCE
Volume 17—Number 3 249Copyright r 2006 Association for Psychological Science
took a test on the material or studied it again before taking a final
etention test 5 min, 2 days, or 1 week later. In Experiment 2,
students studied a passage once and took three tests, studied
three times and took one test, or studied the passage four times.
They then took a final test 5 min or 1 week later. We predicted
that performance on immediate retention tests would increase
with the number of study opportunities, because massed prac-
tice typically produces short-term benefits (e.g., Balota,
Duchek, & Paullin, XXXXXXXXXXHowever, we predicted that taking
tests soon after studying would promote superior retention on
delayed tests relative to repeatedly studying the material. This
outcome would indicate that testing has positive effects on long-
term retention above and beyond any effect of re-presentation of
the material during the test.
EXPERIMENT 1
Method
Subjects
One hundred twenty Washington University undergraduates,
ages 18 to 24, participated in partial fulfillment of course re-
quirements.
Materials
Two prose passages were selected from the reading compre-
hension section of a test-preparation book for the Test of English
as a Foreign Language (TOEFL; Rogers, XXXXXXXXXXEach passage
covered a single topic (‘‘The Sun’’ and ‘‘Sea Otters’’), and each
was divided into 30 idea units for scoring purposes. The pas-
sages were 256 and 275 words in length, respectively.
Design
A 2 � 3 mixed-factorial design was used. Learning condition
(restudy vs. test) was manipulated within subjects, and delay of
the final test (5 min, 2 days, or 1 week) was manipulated between
subjects. The order of learning conditions (restudy or test) and
the order of passages (‘‘The Sun’’ or ‘‘Sea Otters’’) were coun-
te
alanced across subjects.
Procedure
Subjects were tested during two sessions, in small groups (4 o
fewer). They were told that Phase 1 consisted of four 7-min
periods and that during any given period they would be asked to
study one passage for the first time, restudy one of the passages,
or take a recall test over one of the passages. During each study
period, subjects read one passage for 7 min. During the test
period, subjects were given a test sheet with the title of the to-be-
ecalled passage printed at the top and were asked to write down
as much of the material from the passage as they could re-
member, without concern for exact wording or co
ect order.
Subjects solved multiplication problems for 2 min between
periods and for 5 min after the final period in Phase 1.
Phase 2 occu
ed after a 5-min, 2-day, or 1-week retention
interval. In Phase 2, subjects were asked to recall the passages
that they had learned in Phase 1. The recall instructions were
identical to those given in Phase 1. Each retention test lasted 10
min, and subjects were instructed by the experimenter to draw a
line on their test sheets to mark their place after each 1-min
interval during the recall periods (Roediger & Thorpe, XXXXXXXXXXAt
the end of the experiment, subjects were de
iefed and thanked
for their participation.
Results and Discussion
Scoring
Subjects’ recall responses were scored by giving 1 point for each
co
ectly recalled idea unit (out of 30). Initially, 40 recall tests
were scored by two raters, and the Pearson product-moment
co
elation (r) between their scores was .95. Given the high
inte
ater reliability, the remaining recall tests were scored by
one rater.
Initial Test
On the initial 7-min test, subjects recalled on average 20.9 idea
units, or approximately 70% of the passage. No differences were
observed for the two passages or for the different counte
al-
ancing orders.
Final Test
Themean proportion of idea units recalled on the final tests afte
the three retention intervals is shown in Figure 1. The cumu-
lative recall data showed that subjects had exhausted thei
knowledge by the end of the retention interval and are not re-
ported here. After 5 min, subjects who had studied the passage
Fig. 1. Mean proportion of idea units recalled on the final test after a 5-
min, 2-day, or 1-week retention interval as a function of learning condition
(additional studying vs. initial testing) in Experiment 1. E
or bars rep-
esent standard e
ors of the means.
250 Volume 17—Number 3
Test-Enhanced Learning
twice recalled more than subjects who had studied once and
taken a recall test. However, this pattern of results was reversed
on the delayed tests 2 days and 1 week later. On these tests of
long-term retention, subjects who had taken an initial test re-
called more than subjects who had only studied the passages.
The results were submitted to a 2 � 3 analysis of variance
(ANOVA), with learning condition (restudying or testing) and
etention interval (5 min, 2 days, or 1 week) as independent
variables. This analysis revealed a main effect of testing versus
estudying, F(1, XXXXXXXXXX,Zp
2 5 .24, which indicated that,
overall, initial testing produced better final recall than addi-
tional studying. Also, the analysis revealed a main effect of re-
tention interval, F(2, XXXXXXXXXX, Zp
2 5 .46, which indicated
that forgetting occu
ed as the retention interval grew longer.
However, these main effects were qualified by a significant
Learning Condition � Retention Interval interaction, F(2, 117)
5 32.10, Zp
2 5 .35, indicating that restudying produced bette
performance on the 5-min test, but testing produced bette
performance on the 2-day and 1-week tests.
Post hoc analyses confirmed that on the 5-min retention tests,
estudying produced better recall than testing (81% vs. 75%),
t XXXXXXXXXX, d XXXXXXXXXXHowever, the opposite pattern of results
was observed on the delayed retention tests. After 2 days, the
initially tested group recalled more than the additional-study
group (68% vs. 54%), t XXXXXXXXXX, d XXXXXXXXXXThe benefits of
initial testing were also observed after 1 week: The tested group
ecalled 56%of thematerial, whereas the restudy group recalled
only 42%, t XXXXXXXXXX, d XXXXXXXXXXFigure 1 depicts anothe
interesting finding: The initially tested group recalled as much
on the 1-week retention test as the additional-study group did
after only 2 days (the initially tested group actually recalled
slightly more)