THE JOURNAL OF

EDUCATIONAL PSYCHOLOGY
Volume XXX December, 1939 Number 9

STUDIES IN RETENTION
HERBERT F. SPITZER
State University of Iowa*

The present investigation was planned with special reference to

(1) the effect of recall on retention, (2) the relationship between the
rate of forgetting and the ability of the subjects, and (3) the effect of
item difficulty on the form of retention curves.
PROBLEM

The importance of retention is shown by the fact that growth or

improvement in skills, knowledges, and attitudes is dependent upon
the learner's retention of the effects of previous experience. The use
of recall as an aid to retention has been emphasized by theorists on
methods of study and by investigators in the field of memory. The
lack of experimental evidence on the effect of recall on retention where
conditions approach those of actual schoolroom practice prompted this
investigation. The primary purpose of the investigation was to deter-
mine the effect of recall on retention of facts which children acquire
through reading when the materials and the methods of study are
similar to those used in classroom situations. Two subsidiary prob-
lems of the investigation relate to the effect of item difficulty upon the
form of retention curves and to the relationship between the learning
ability of the students and the rate of forgetting.
Related Research.—A number of experimental studies have dealt
with various aspects of the problem of this investigation. Myers8
found that immediate recall in the form of written reproduction was
beneficial to later reproduction of a list of unrelated words. An indi-
rect or incidental method of learning the words had been employed.
* This article reports, in part, an investigation conducted by the writer as a
partial fulfillment of the requirements for the Ph.D. degree in the College of
Education of the State University of Iowa. The assistance of Dr. Ernest Horn,
under whose direction the study was conducted, is gratefully acknowledged.
641
642 The Journal of Educational Psychology

Gates2 found that some recall in the form of recitation was an aid in
memorizing biographical prose. This was true when tests were given
immediately and also after four hours. Gates also reports finding a
high positive correlation between immediate and delayed recall.
Yoakam7 concluded that an immediate recall in the form of a test was
of more value to retention than was a single reading. Jones3 found
that recall tests aided the retention of information acquired from
lectures by college students Other studies which are related to some
phases of this study are reported by Dietze,1 Keys,4 Raffel,6 and
Young.8
SUBJECTS AND MATERIAL

The subjects for this experiment were thirty-six hundred five

sixth-grade pupils of nine Iowa cities. This was the entire sixth-grade
population of ninety-one elementary schools. These schools were
arbitrarily divided into ten groups. Groups I, II, III, IV, V, VI, VII,
and VIII, each comprising approximately four hundred children, were
used to obtain the data for the main part of the experiment. Group IX
was used to obtain data on the effect of immediate repetition of tests,
and Group X was used to obtain information on previous knowledge.
Thus Groups IX and X were, in a sense, used as control groups.
The reading materials used in this study were two articles, six
hundred four and five hundred seventy-seven words in length, printed
in a six-page folder. The first of these articles entitled "Peanuts,"
designated A, and the accompanying test, were used as a sectioning
device; while the second article, entitled "Bamboos," designated B,
and the accompanying test, were used to obtain the data on retention.
An attempt was made to write articles with content that would be
relatively new to the children, highly factual, authentic, of the proper
difficulty, and similar in type to the material that children read in their
regular school work. Both of the topics dealt with are treated briefly
if at all in geography books used by the children. The first paragraph
of Article B and five of the items of Test B, which are wholly or partially
based on the content of this paragraph, are given below.
Near Savannah, Georgia, is the Plant Introduction Garden of the United
States Department of Agriculture. In one section of this garden are the
bamboos. These plants are members of the grass family. They resemble
their relatives, corn and wheat, in structure of stem which is rounded, divided
into joints, and more or less hollow. Bamboos also resemble pines and
spruces by having tall, straight trunks and cone shaped heads or crowns, and
 Studies in Retention 643

by being evergreens. Although they are evergreens, a new set of leaves

appears each spring. On rare occasions bamboos flower and produce seeds.
After flowering, the plant usually dies.
1. Who maintains an experimental garden near Savannah, Georgia?
( ) The Bamboo Growers Association ( ) The U. S. Government ( )
The U. S. Custom Office ( ) The florists of Savannah ( ) The State of
Georgia.
2. To which family of plants do bamboos belong? ( ) trees ( ) ferns
( ) grasses ( ) mosses ( ) fungi.
4. Which two trees do the bamboos resemble most? ( ) royal and date
palms ( ) willow and tamarack ( ) white oak and burch ( ) walnut and
hickory ( ) pine and spruce.
18. How often do bamboos produce seed? ( ) every spring ( ) never
( ) on rare occasions ( ) every other year ( ) every third year.
19. What usually happens to a bamboo plant after the flowering period?
( ) It dies ( ) It begins a new growth ( ) It sends up new plants from
the roots ( ) It begins to branch out ( ) It begins to grow a rough bark.
A twenty-five item test (see example above) on each article was
used to measure retention. Through an item analysis of each test, the
difficulty and discriminating power (tetrachoric " r " ) of each item
were obtained. Items too difficult or of poor discriminating power
were eliminated after the preliminary trials, leaving a total of twenty-
five items in each test. The reliability of the two tests determined by
equal halves technique was .77 and .80, respectively. The correlation
between Test A and Test B based on scores of six hundred twenty-two
pupils who took both tests immediately after reading the articles
was .76.
PROCEDURE
Both the reading material and the tests were presented to the
pupils by their regular teachers according to directions supplied by
the investigator. In the printed directions given at the beginning of
the experiment, the children were told that they were taking part in a
learning experiment. They were also told that they were to try to
remember the information given in the articles. Similar directions
were printed at the top of each article.
On the first day of the experiment, all the children in Groups I-VIII
read Article A and took Test A. They also read Article B, but only
Groups I and II took Test Bi.* The other six groups took a test
* Thefirsttime pupils took Test B, it was labeled Bi\ the second time they took
Test B, it was labeled Bi\ and the third time, it was labeled B».
644 The Journal of Educational Psychology
TABLE I.—DIAGRAM OP PROCEDUBB

Time in days 0 1 7 14 21 28 63

Groups
I B, B, B,
II B, B, B,
III B, B\
IV Bi B2
V B. B,
VI B, B,
VII B,
VIII • • •• B,

which had little relation to the content of Article B. This test was
given in order to keep them from expecting a later test. The real
Test B was given to these six groups (III-VIII) at varying time
intervals after the start of the experiment. A diagram of the testing
procedure followed is given in Table I. This table shows that the
pupils of Group I took Test Bi immediately after reading Article B,
repeated the test (Test 2?2) after one day, and again repeated the same
test (Test B3) after twenty-one days. The pupils of Group IV took
Test JBI seven days after reading Article B and repeated the test
(Test B^ twenty-one days after reading the article. The procedure
followed by the other six groups is shown in the table.
The reading materials were not referred to after the initial study
period, nor did the pupils know that there were to be delayed tests.
The teachers were instructed not to discuss the articles or the tests
with the pupils. Pupils were given eight minutes to read the articles
and ten minutes for each of the tests.
The pupils of Group IX read both articles and took both tests on
the first day of the experiment. They took Test B a second time
immediately after completing the first attempt. For this second test
they were instructed to try to improve their first score. This pro-
cedure was followed for the purpose of obtaining information on the
effect of repetition of the tests. The pupils of Group X took Test B
without having read Article B, for the purpose of obtaining data on
previous knowledge.
RESULTS

Effect of Recall.—From a single frequency distribution of the scores

on Test A, the mean score and the standard deviation of scores of the
 Studies in Retention 645
population studied were determined. The means and standard
deviations of the scores of the pupils on Test A in each of Groups I-VIII
were then calculated. A comparison of these means with the mean
for the entire population showed that each of the groups was practically
representative of the population. However, some pupil scores were
deleted from two of the groups in order to make the groups more
nearly equal. The means of these equated groups are given in
Table II. The critical ratio of the largest difference between means
on Test A of any two of these equalized groups is only .18.
TABLE II.—TEST B RESULTS OF THE GROUPS THAT MADE EQUIVALENT SCORES ON
TEST A

Group N M test A Test* M test B SD test B

I 286 15.03 B, 13 23 4 69
I 284 B, 13 07 4 57
I 266 B, 12.18 4 59
II 338 15 05 B, 13 20 4.50
II 335 B, 11 84 4 64
II 312 B, 10 74 4.22
III 367 15.00 B, 9 56 4.24
III 349 Bt 8.93 4 06
IV 337 15.00 B, 7 87 3 56
IV 323 B, 8.15 3.83
V 371 15.04 B, 6 97 3 53
V 353 B, 7 10 3.21
VI 379 15 04 B, 6.49 2 91
VI 352 B, 7.07 3 08
VII 365 15 00 B, 6.80 3 03
VIII 350 15.03 B, 6 38 2.71
* For identification of various Test B's and the tune after learning that each
was taken, see Table I.
The mean scores of all groups of pupils who took Test B are shown
in Table II. In interpreting the data of this table the assumption is
made that the eight groups of pupils profited equally from the reading
and that the groups possessed equal ability to retain the effects of the
reading. This assumption is based on the fact that the groups were
equated on the sectioning test (Test A). According to the assumption
stated above, Group I would have made a mean score of approximately
9.56 (mean score of Group III on Test Bi) one day after reading the
article if the group had not been given the immediat erecall test. This
last statement is based on the fact that Groups I and III were originally
646 The Journal of Educational Psychology
equal. Therefore, had Group I delayed taking Test B until one day
after reading, the mean score made by the group would have been
the same as that made by Group III, or 9.56. On the same basis, had
Groups III, IV, V, VI, VII, and VIII been tested immediately after read-
ing Article B, each group would have made a mean score of approxi-
mately 13.22. The data of this table show that more is forgotten in one
day without recall than is forgotten in sixty-three days with the aid of
recall, as is shown by a comparison of the scores of II B3 and VIII Bt.
The differences between originally comparable groups and now differing
by only one recall and the critical ratio of these differences are shown
in Table III. These critical ratios would have been larger had the
standard error of difference formula for matched groups been used.*
This formula was not used because a statistically significant difference
was obtained without its use. The data summarized in Tables II
and III show clearly that retention benefited significantly by recall.
The effects of recall on retention are shown graphically in the
figures (see Figs. 1, 2,3, and 4). The points on the graph represent the
mean performance of the different groups. For example, the "Point
III Bi" represents the mean score of the pupils of Group III on their

TABLE III.—COMPARISON OP TEST B RESULTS FOB GROUPS DIFFERING BY ONLT

ONE RECALL

Group Test AM SE, AM Difference SE

CR
difference

I B, 13 07 .27
III B, 9.56 .22 3.51 346 10.10
I B, 12.18 28
IV Bs 8.15 .21 4.03 .350 11.51
II Bt 11.84 .25
IV B, 7.87 .19 3 97 317 12 50
II B, 10.74 .24
VI B, 7.07 .16 3 67 288 12.74
III B, 8.93 .22
V B, 6.97 .18 1 96 286 6.85
IV B, 8 15 .21
VI B, 6.49 .16 1 66 .264 6.29
V B, 7.10 18
VII Bx 6.80 .16 .30 241 1.25

* Lindquist, E. F. and Wilks, S. S.: "The Significance of a Difference between

Matched Groups." J. Educ. Psycho., Vol. xxn, 1931, pp. 197-204, 205-208.
 Studies in Retention 647

first attempt on Test B. This trial was taken one day after reading
Article B. Since the eight groups were equalized according to their
performance on Test A, which was quite similar to Test B, it is
assumed that the facts which Groups I and II acquired through reading
would have been forgotten at the rate shown by the solid line if they
had not been given the intervening recall tests. Thus, the space
separating a point on the solid line and a vertical point on any of the
broken lines represents the effect of the recall test on retention for that
particular situation. Figures 1 and 2, which are based on the entire
population, show that immediate forgetting unaided by recall was very
rapid and that in every case recall was beneficial to retention.
M
\w -
18

12 SB,

10 1 M7B,
SB,

a g JVS,

»-
r
- . 1
P « Unaided tv
* .
to
Aided ty

to
0
14 21
TIME IN DATS (

FIG CURVES OF RETENTION FOR THE ENTIRE POPULATION WHEN THE

AMOUNT RETAINED IS EXPRESSED IN RAW SCORES

When the papers were corrected for guessing through use of the
W
formula, S = R — (N _ ,y the critical ratios of differences remained
practically the same as those shown in Table III.
The one hundred sixty-nine children of Group IX who were given
a second Test B immediately after taking the first test improved their
mean score only .03. Thus, an immediate repetition of the test did not
result in a large increase in the number of facts acquired.
The mean score of the three hundred one children of Group X who
were tested for previous knowledge was 5.27. The assumption that
this score represents the previous knowledge of Groups I-VIII is
rather questionable since the foils or wrong responses of items in the
test were not as plausible for those who had not read the article
(group tested for previous knowledge) as these responses were for those
648 The Journal of Educational Psychology
pupils who had read the article. The previous knowledge of Groups
I-VIII, then, was probably less than 5.27. Since no other measure of
previous knowledge was available, this amount was subtracted from
the mean score of each group in obtaining the data from which Fig. 2
was constructed. This figure, subject to the limitation given above,
shows the curves of retention when the amount retained is given as a
per cent of the amount learned. The figure shows the same trend as
that shown in Fig. 1. In this case, however, forgetting is more rapid

• h—
100

' • *

••5, 1
JUS*
A'c/rtY ty

»-O
•
re, va,
VA — _
s52.

7 M 21
TIME IN DAYS

FIG. 2 CURVES OF RETENTION FOR THE ENTIRE POPULATION WHEN

THE AMOUNT RETAINED IS EXPRESSED AS A PER CENT OF THE
AMOUNT LEARNED

and the curve of retention unaided by recall begins to level much

nearer the zero line.

LEARNING ABILITY AND RETENTION

In order to determine whether ability to learn affected the shape of
the retention curve, the Test B scores of those pupils who scored in the
upper one third on Test A were compared with the Test B scores of
those in the lower one third on Test A. The data for making these
comparisons are shown in Tables IV and V. The data on the upper
and lower thirds are shown graphically in Figs. 3 and 4. An examina-
tion of these figures shows that the curve of retention of pupils of
inferior learning ability begins to level or to have reached the horizontal
seven days after learning. At the same time the curve for the pupils
with superior learning ability does not begin to level until twenty-one
days after the learning period.
 Studies in Retention 649
TABLE IV.—TEST B RESULTS OF PUPILS WHO MADE SCORES IN THE UPPER AND
LOWER THIRDS ON TEST A

Group Division AM AM SDof

N Test
test A test B B scores

I Upper 95 19.62 B, 17.02 3.57

I Upper 95 Bt 16.76 3.36
I Upper 88 B, 15.47 3.87
I Lower 95 10.42 B, 9.67 3.40
I Lower 95 B, 9.67 3.21
I Lower 89 B, 9 19 3.36
II Upper 112 19.17 B, 17.18 2.70*
II Upper 120 B, 16.10 3.64f
II Upper 111 B, 14.57 3.42f
II Lower 112 10.92 B, 10.01 3.23
II Lower 120 B, 8.55 3.25
II Lower 111 B, 7.39 2.49
III Upper 123 19.33 B, 12.58 4.02
III Lower 123 10.90 B, 7.09 2.52
IV Upper 113 19.27 B, 10.39 3.36
IV Lower 113 10.92 B, 5.48 2.43
V Upper 124 19.54 B, 9.28 3.47
V Lower 124 10.84 B, 5.22 2.30
VI Upper 130 19.58 B, 8.46 3.21
VI Lower 130 10.76 B, 6.33 2.03
VII Upper 122 19.47 B, 8.50 3 11
VII Lower 122 10.57 B, 5.57 2.13
VIII Upper 117 19.43 B, 7.74 2.84
VIII Lower 117 10.79 B, 5.16 1.95
* Unusually small because high scores were deleted in the equalization
procedure.
t Not equalized.
TABLE V.—COMPARISON or TEST B RESULTS OF UPPER AND LOWER THIRDS FOB
GROUPS DIFFERING BY ONLY ONE RECALL

Differ- SE
SE,
Group Division Test AM AM ence of differ- CR
AM's ence

I Upper B, 16.76 .34

III Upper B, 12.58 .36 4.18 .495 8.44
I Lower B, 9.67 .33
III Lower B, 7.09 .23 2.58 .402 6.41
II Upper B, 16.10 .33
IV Upper B, 10.39 31 5.71 .463 12.33
II Lower B, 8.55 .30
IV Lower B, 5.48 .23 3.07 .378 8.12
650 The Journal of Educational Psychology
Because of the fact that the data of Tables IV and V and Figs. 3
and 4 are based on scores neither corrected for guessing nor corrected
for previous knowledge, the rate of forgetting shown is probably much
less rapid than the true rate of forgetting. According to the data as
presented the pupils in the upper one third forgot twenty-six per cent
' \\l&MB,
' l?4-~-
'% — MO,

\l!B,

.aw1
zae,

Una/dec/ £y

7 14 21
TIME IN DAYS

FIS. 3 CURVES OF RETENTION FOR THE PUPILS IN THE UPPER ONE

THIRD OF SCORES ON TEST A

1
14

ire,

W 21
TIME IN DAYS

FIG 4 CURVES OF RETENTION FOR PUPILS IN THE LOWER ONE

THIRD OF SCORES ON TEST A

of their original score in one day while the pupils in the lower one third
forgot twenty-eight per cent. For one week, the figures were thirty-
nine per cent and forty-four per cent. When scores were corrected
for guessing the upper one third forgot within one day thirty-three
per cent while the lower one third forgot forty-nine per cent. It
should be remembered that the upper and lower thirds referred to are
 Studies in Retention 651
based on the pupils' scores on Test A. The data show that the pupils
in the lower third tend to have a more rapid initial rate of forgetting.
Additional data on the relationship between learning ability and
retention were secured by correlating the immediate score of pupils on
Test A with their delayed score on Test B. The score on Test A was
considered a measure of the pupils' ability to learn, while the delayed
score on Test B was considered as a measure of the pupils' ability to
retain. The " r ' s " obtained in the above manner ranged from .60 for
one day to .44 for sixty-three days. Since the correlation between the
two tests on immediate recall was only .76, the " r ' s " given for the rela-
tionship between immediate and delayed recall, or between ability to

~- "-—1

— _ - ^ —
_

14 21 63 PREVIOUS
KNOWLEBGC
- TIME IN DAYS -

FIG S CURVES OF FORGETTING FOR ITEMS NUMBER 9, I I , 15, 0 23.

(PER CENTS UNCORRECTEO)

learn and retention, are probably very conservative. In connection

with the relationship between immediate and delayed recall, the atten-
tion of the reader is called to the fact that in some previous studies this
relationship was found by correlating the immediate and delayed scores
on the same test. The results of such a procedure are almost certain
to be spuriously high because of the effect of recall on retention. When
this procedure was used in the present investigation, a correlation of
.91 between the scores on immediate recall and recall after one day
was obtained.
EFFECT OF ITEM DIFFICULTY ON THE FORM OF RETENTION CURVES

From an item analysis of the papers of all pupils who took Test B,
the data on forgetting of individual items shown in Table VI were
652 The Journal of Educational Psychology
obtained. The first seven columns represent the per cent of pupils
tested at various times who answered each item correctly. The last
column shows the per cent of the pupils tested for previous knowledge
who answered the items correctly. An examination of the data in
Table VI will show that there is little relationship between the rate of
forgetting of items and the initial difficulty of the items. This fact is
shown graphically in Fig. 5. The four items (9, 11, 15, and 23) for
which curves of retention are shown were approximately of equal
difficulty on the initial test. Two of the items (11 and 23) were also
about equal on the previous knowledge score. In spite of these simi-
larities, widely differing curves of retention were found. The one
general characteristic of the data on each item is the fact that the rate
TABLE VI.—THB P E R CENT OP CORRECT RESPONSES FOR EACH OF THE ITEMS ON
TEST B

Item I and IV
number II III V VI VII VIII X

1 40 28 21 19 17 20 14 10
2 61 30 20 15 16 16 15 14
3 81 62 52 45 43 43 50 60
4 80 56 38 36 31 33 28 18
5 55 42 40 37 39 41 46 56
6 76 57 57 55 59 61 61 66
7 43 40 33 29 25 24 25 17
8 32 24 18 19 22 18 19 9
9 75 60 53 46 45 44 40 26
10 55 28 21 20 14 18 13 11
11 72 50 29 23 23 23 31 38
12 29 22 19 16 15 14 10 6
13 67 30 24 19 20 25 26 19
14 77 64 57 44 46 41 33 20
15 73 46 34 24 26 22 14 9
16 41 28 27 17 13 17 14 7
17 40 28 22 18 19 19 13 8
18 37 20 19 17 14 14 9 4
19 68 39 32 22 17 17 9 4
20 53 39 32 34 33 27 32 21
21 63 27 16 14 16 13 13 14
22 24 17 14 19 16 13 17 9
23 73 56 54 52 50 51 50 34
24 35 31 27 27 26 29 28 22
25 31 21 21 19 17 21 20 10

SE of the per cents varies from 2.1 to 2.7.

 Studies in Retention 653

of forgetting is more rapid during the first day than during any sub-
sequent period.
The data for items 3, 5, 6, and 11 (Table VI) show that reading or
study of material can be detrimental to success on a test over the
content of that material. Children who had read the article made
lower scores on some tests over these four items than those who had not
read the articles. In the case of these four items, the difference
ascribed to this interference was statistically significant.
The data on items 13, 24, and 25 seem to be evidence for reminis-
cence. However, none of the differences or gains ascribed to reminis-
cence are statistically significant.
When the amount of previous knowledge (last column in Table VI)
is subtracted from the per cents given in the other columns of Table VI,
a very different picture is presented. (See Table VII.) The difficulty
TABLE VII.—ITEMS IN RANK ORDER OF PER CENT ANSWERED CORRECTLY AFTER
SUBTRACTION OP THE AMOTJNT CREDITED TO PREVIOUS KNOWLEDGE

Item number I and II III IV V VI VII VIII

15 64 35 25 15 17 13 5
19 64 35 28 18 13 13 5
4 62 38 20 18 13 15 10
14 57 44 37 24 26 21 13
9 49 34 27 20 19 18 14
13 48 11 5 0 1 6 7
21 47 13 2 0 2 - 1 - 1
2 47 16 6 1 2 2 1
10 44 17 10 9 3 7 2
23 39 22 20 18 16 17 16
11 34 12 - 9 -15 -15 -15 - 7
16 34 21 20 10 6 10 7
18 33 16 15 13 10 10 5
17 32 20 14 10 11 11 5
20 32 18 11 13 12 6 11
1 30 18 11 9 7 10 4
7 26 23 16 12 8 7 8
8 23 15 9 10 13 9 10
12 23 16 13 10 9 8 4
3 21 2 - 8 -15 -17 -17 -10
25 21 11 11 9 7 11 10
22 13 8 5 10 7 4 8
24 13 9 5 5 4 7 6
6 10 - 9 - 9 -11 - 7 - 5 - 5
5 - 1 -14 -16 -19 -17 -15 -10
654 The Journal of Educational Psychology

rank of some items changes considerably. However, even after this

change, no relationship between difficulty and rate of forgetting is
evident.

TABLE VIII.—THE EFFECT OF RECALL ON RETENTION FOR INDIVIDUAL ITEMS

Group Group
Group
II per II per
VTTT
Differ- Per cent Differ- Per cent
V XXX
Item number cent cent ence for- ence for-
correct correct per cent Bt-B, gotten B.-VIII gotten
on Bi on B,

1 39 25 14 14 36 25 64
2 61 40 15 21 34 46 75
3 81 63 50 18 22 31 38
4 80 67 28 13 16 52 65
5 55 61 46 - 6 -11 9 16
6 76 81 61 - 5 - 7 15 20
7 42 33 25 9 21 17 40
8 31 17 19 14 45 12 39
9 74 65 40 9 12 34 46
10 51 37 13 14 27 38 75
11 72 41 31 31 43 41 57
12 28 21 10 7 25 18 64
13 63 43 26 20 32 37 59
14 76 65 33 11 14 43 57
15 72 55 14 17 24 58 81
16 41 33 14 8 20 27 66
17 40 35 13 5 13 27 68
18 37 37 9 0 0 28 76
19 68 52 9 16 24 58 85
20 54 36 32 18 33 22 41
21 59 37 13 22 37 46 78
22 26 22 17 2 8 9 35
23 69 68 50 1 1 19 28
24 35 37 28 - 2 - 6 7 20
25 31 26 20 5 16 11 35

In order to determine what effect recall had on the retention of

individual items, a comparison of the amount forgotten by Group II
(with recall) and Group VIII (without recall) was made. In each case
the time elapsed was sixty-three days. The data for making thi6
comparison are given in Table VIII. The data show that recall was
beneficial to retention for every item. When the pupils were divided
into an upper and a lower group on the basis of their Test A scores and
 Studies in Retention 655
data similar to that of Table VIII prepared, it was found that the
superior pupils benefited more from recall.
The high scores made by the pupils of Groups I and II on the
repeated tests might be attributed to the pupils' remembering how the
test was marked the first time. If this assumption were valid, then,
both correct and incorrect responses would be repeated with equal
consistency. In order to test the assumption, one hundred of the
Test Bi and Test B2 papers of Group II were analyzed to determine the
consistency of responses. It was found that right answers were
repeated seventy-nine per cent of the time; while wrong answers were
repeated only fifty per cent. Since right answers were repeated more
consistently than wrong answers, only a small portion of the unusually
high scores on the second trial of the test can be ascribed to the pupils'
remembering how the test was marked the first time. Chance alone
would give twenty per cent repetition of responses. The fact that
wrong answers were repeated more often than chance would allow is
not in conflict with the assumption that recall and not mere repetition
is the factor that aids retention. In giving some wrong responses, the
pupils were sometimes recalling information given in the article, for
many of the false responses were taken from the content of the article.
Thus, recall can also aid the retention of erroneous ideas.
Limitations.—The findings of this study are subject to a number
of limitations. The most important of these are the following: (1)
The method of measurement employed a type of cue not used in
recalling information in everyday school or life situations. (2) The
tests were repeated. (3) The learning was of little practical use to
the children. (4) The children were given no opportunity to refer to
the material after the initial learning period. (5) The pupils did not
know whether their responses were correct or incorrect. (6) The
effect of reacting to one item upon the response to other items is not
known.
EDUCATIONAL IMPLICATIONS

The findings of this study seem to warrant the following educa-

tional implications: (1) Immediate recall in the form of a test is an
effective method of aiding the retention of learning and should, there-
fore, be employed more frequently in the elementary school. (2)
Because recall can aid in fixing erroneous ideas, all tests or examina-
tions should be returned to the pupil corrected, or the pupil should be
given an opportunity to correct his own paper. (3) Achievement
656 The Journal of Educational Psychology
tests or examinations are learning devices and should not be considered
only as tools for measuring achievement of pupils. (4) Since reading
about a topic can interfere with the knowledge the subject already
possesses, careful consideration should be given to what is supposed
to be learned and to the thoroughness of learning. (5) In all studies
where the same tests are repeated, the possible effects of recall on
retention should be recognized.

BIBLIOGRAPHY

1. Dietze, A. G., and Jones, G. E.: "Factual Memory of Secondary-School Pupils

for a Short Article Which They Read a Single Time." / . Ed. Psycho., Vol.
xxn, 1931, pp. 586-598, and 667-676.
2. Gates, A. I.: "Correlation of Immediate and Delayed Recall." / . Ed.
Psycho., Vol. ix, 1918, pp. 489-496.
3. Jones, H. E.: "Experimental Studies of College Teaching." Archives of
Psychology, No. 68, 1923.
4. Keys, Noel: "The Influence on Learning of Weekly as Opposed to Monthly
Tests." J. Exp. Psycho., Vol. xxv, 1934, pp. 427-436.
5. Myers, G. C : "Recall in Relation to Retention." J. Ed. Psycho., Vol. v, 1914,
pp. 119-130.
6. Raffel, Gertrude: "The Effect of Recall on Recognition." J. Exp. Psycho.,
Vol. xvn, 1931, pp. 828-837.
7. Yoakam, G. A.: The Effect of a Single Beading on the Retention of Various Types
of Materials in the Content Subjects of the Elementary-School Curriculum as
Measured by Immediate and Delayed Recall. Ph. D. Thesis, State University
of Iowa, 1922, p. 254.
8. Young, William E.: The Relation of Reading Comprehension and Retention to
Hearing Comprehension and Retention. Ph. D. Thesis, State University of
Iowa, 1930.