JOURNAL OF VERBAL LEARNING AND VERBAL BEHAVIOR21, 451-467 (1982)

When Forgetting Helps Memory: An Analysis of Repetition Effects

LAUREN J. C U D D Y AND LARRY L . JACOBY

McMaster University

For the processing of an item to be repeated, it is argued that the memory for a prior
presentation of that item must not be readily accessible when the item is repeated. Varying
the spacing ofrepetitions corresponds to a manipulation of retention interval and is only one
of several means of varying the accessibility of memory for a prior presentation. Factors
such as the similarity of repetitions, the type of intervening material, and cue effectiveness
are also shown to influence the processing of repetitions through their effects on accessibil-
ity. The experiments that are reported reveal interactions between each of these latter
variables and the spacing of repetitions. The results are discussed with relevance to previous
accounts of the effect of spacing repetitions, and a more general account of repetition effects
is offered.

In a classic paper, Melton (1967) de- titions is not the only means of producing
scribed the effect of spacing repetitions as forgetting or decreasing the accessibility of
paradoxical in that it seems to suggest that the memory for an item; effects similar to
forgetting helps memory. As the spacing of those of spacing repetitions might be pro-
repetition increases, a subject is less likely duced by manipulating other factors that
to recognize an item as being a repetition; influence the accessibility of the memory
however, when a later test of retention is for a prior presentation of an item when that
given, performance is higher when repeti- item is repeated.
tions of an item are spaced rather than Jacoby (1978) argued that remembering a
being massed during study. Since Melton's prior presentation of an item when that item
paper, a great deal of research and theoriz- is repeated allows a subject to bypass much
ing have been directed toward explaining of the processing of the repetition that
the effect of spacing repetitions (see would otherwise be required. A series of
Hintzman, 1974, for a review). Rather than experiments conducted by Jacoby demon-
focussing on the effect of spacing repeti- strated that remembering a prior presenta-
tions, however, we take Melton's paradox tion of an item can reduce subsequent re-
very seriously by investigating the effects tention performance. In a first phase of
of forgetting on the magnitude of repetition those experiments, a context word was
effects. It may simply be necessary for an presented along with letters and a series of
item to be partially forgotten or not readily blanks representing the missing letters of a
accessible for a repetition of that item to be word that was related to the context word
maximally effective. If so, theorizing (e.g., FOOT S--E). The subject's task was
should center around the question of why to solve the puzzle by reporting the related
repetition effects depend on forgetting word that could be produced by restoring
rather than around the effect of spacing the missing letters. In a second phase, the
repetitions. Increasing the spacing of repe- context word was given as a cue for recall
of the solution word. In some conditions,
This research is based upon a MA thesis conducted the task of giving a solution to the problem
by the first author, and was supported in part by Grant
in the first phase was trivialized by present-
A0281 from NSERC to the second author. Requests
for reprints should be sent to Larry L. Jacoby, De- ing the context word with the completed
partment of Psychology, University of Utah, Salt Lake solution word immediately prior to the puz-
City, Utah 84112. zle (e.g., FOOT SHOE: FOOT S--E). In
451
0022-5371/82/040451-17502.00/0
Copyright~) 1982by AcademicPress, Inc.
All rights of reproductionin any formreserved.
452 CUDDY AND JACOBY

those conditions, subjects did not have to tition was expected to serve as a better cue
actually solve the puzzle but could respond for retrieval of the prior presentation of that
by simply remembering the previously read problem than was presentation of an altered
solution. In other conditions, several prob- version of the problem. In all experiments,
lems intervened between the reading of a conditions that were less favorable for re-
solution and presentation of the problem membering a prior presentation of an item
requiring that solution. The memory of the when that item was repeated were expected
previously read solution should be less ac- to lead to higher subsequent cued recall. In
cessible in these latter conditions so that the general discussion, we consider the rel-
more processing is necessary to obtain a evance of the results of these experiments
solution for the presented problem; sub- to accounts of the effects of spacing repeti-
sequent retention performance was ex- tions and other memory phenomena.
pected to benefit from this additional pro-
cessing. In line with these expectations, EXPERIMENT 1
cued recall was found to increase with in-
creased spacing of the reading of a solution Method
and presentation of the problem requiring Design and subjects. The subjects were
that solution. Conditions that were less fa- 18 students enrolled in an introductory psy-
vorable for remembering the previously chology class who received course credit
read solution when a problem was pre- for participating in the experiment. Sub-
sented led to higher subsequent recall. jects were tested individually.
In the experiments to be reported, we In the first phase of the experiment, sub-
employed the procedure used by Jacoby jects read pairs of related words in which
(1978) but manipulated factors in addition the second member of the pair was pres-
to spacing to influence the accessibility of ented either intact or missing letters (e.g.,
the memory for a prior presentation of an TREE: BR--CH). When letters were miss-
item when that item was repeated. Spacing ing, the subject was to say the word that
of repetitions corresponds to a manipula- resulted from restoring the missing letters.
tion of the retention interval between the Within the list, some pairs of words were
presentation of an item and its repetition. It repeated. Half of these repeated pairs were
is well established that the effects of in- presented intact to be read on both presen-
creasing retention interval depend on the tations (Read-Read); the other half were
similarity of the material presented during intact on their first presentation but the
the retention interval to the material that is right-hand member of the pair was missing
to be remembered. In Experiment 1, the letters on its second presentation (Read-
similarity of the material that intervened Construct).
between the presentation of a solution and Repetitions of a pair were separated by
the problem requiring that solution was either zero, four, or eight intervening items.
manipulated. The two were separated by The intervening items at four and eight
the presentation of other word problems so spacings were either pairs of related words
that the intervening material was similar, or with letters missing from the right-hand
by addition problems so the intervening member of the pair, similar to repeated
material was dissimilar to the previously items, or addition problems. Type of inter-
presented solution. A second factor that in- vening item was factorially combined with
fluences the accessibility of a memory is the the longer spacings (four vs eight interven-
cues provided at the time of test. In Ex- ing items) and type of repetition (Read-
periments 2 and 3, problems were repeated Read vs Read-Construct). All factors were
with a problem being either identical or al- manipulated within subjects. In the second
tered across repetitions. An identical repe- phase of the experiment, subjects were
 REPETITION EFFECTS 453

provided with the left-hand members of of these lists were presented to two subjects
repeated pairs, and asked to recall the while the remaining six lists were presented
right-hand members of those pairs. to one subject. An attempt was made to
Materials. The materials included 144 evenly distribute items representing the
pairs of related words that were chosen different forms and spacing of repetitions
from the C o n n e c t i c u t f r e e - a s s o c i a t i o n throughout a study list so that no class of
norms (Bousfield, Cohen, Whitmarsh, & items would differentially benefit from any
Kincaid, 1961). The right-hand member of serial position effects. A cued-recall test
each pair was approximately the third high- sheet contained the left-hand member of
est associate of the left-hand member of the 48 critical repeated pairs as a cue for
that pair and was from 4 to 8 letters in recall of the right-hand member of those
length. Of these pairs, 48 pairs were critical pairs, The order of items in the test list
in that they were repeated within a list. was random.
Twenty-four of the pairs were presented Procedure. In the first phase of the ex-
intact to be read on each of their two pre- periment, subjects were informed that we
sentations; the remaining 24 critical pairs were investigating their reaction time for
were each presented intact for their first solving problems of different types. They
presentation but with two letters deleted were instructed to go through the deck of
from their right-hand member for their sec- cards one by one, pushing a button to allow
ond presentation. Deleted letters were re- recording of their reaction time as they
placed by dashes. For both critical and announced the solution to the problem on
noncritical pairs, neither the first nor the each card. Subjects read both words aloud
last letter of a word was ever deleted. A set for the verbal problems, restoring deleted
of 96 addition problems completed the ma- letters when necessary, and gave the sum
terials employed to construct lists. The ad- for addition problems. A timer placed sub-
dition problems required the addition of a jects through the cards by producing a tone
single-digit number to a two-digit number; every 6 seconds; subjects were to use the
problems were selected to be nonoverlap- tone as a signal to turn the next card. In
ping in two-digit numbers and to be at an reality, reaction times for solving problems
approximately uniform level of difficulty. were not recorded.
Study lists were constructed by typing After subjects had progressed through all
pairs of related words and addition prob- of the cards, there was a brief period during
lems on notecards. A study deck contained which the experimenter recorded the sub-
288 cards: 24 pairs repeated intact (Read- ject's name and student number. Next, the
Read), 24 pairs presented intact and then subject was given an unexpected test of
presented with letters from the right-hand cued recall. Completion of the cued-recall
member of the pair deleted (Read-Con- test was subject paced.
struct); 96 noncritical pairs; and 96 addi- The level of significance for all statistical
tion problems. Among repeated pairs, 16 tests was set at p < .05.
pairs were presented at each of the three
levels of spacing of repetitions, (zero, four,
and eight). At each of the two longer spac- Results and Discussion
ings, half of the repeated pairs were sep- In the first phase of the experiment, only
arated by noncritical pairs and half were one error occurred in restoring missing let-
separated by addition problems. Twelve list ters to construct a response for pairs re-
formats were constructed such that across peated in the Read-Construct condition.
formats each pair equally often represented For noncritical items, more errors were
each of the combinations of spacing, type of made solving verbal intervening items (.~ =
intervening item, and type of repetition. Six 6.67) than solving math problems that in-
454 CUDDY AND JACOBY

tervened between repetitions (X = 2.39), of memory for a prior presentation of an

t(17) = 4.27. item on its later processing. In agreement
Of greater interest were the c u e d - r e c a l l with others (e.g., Slamecka & Graf, 1978),
results from the second phase of the ex- we argue that retention performance shows
periment. For a first analysis of those re- a benefit when subjects are required to con-
sults, performance was collapsed across the struct a response or an encoding for an
different types of intervening material at the item. As the spacing of presentations is in-
two longer spacings of repetitions to pro- creased, the previously read solution be-
duce only spacing of repetitions and rep- comes less accessible in the R e a d - C o n -
etition condition ( R e a d - R e a d vs R e a d - struct condition so that more problem-
Construct) as factors, This analysis revealed solving activity is required to arrive at
a significant effect of spacing repetitions, a solution when a problem is presented.
F(2,34) = 5.65, M S e = 1.24, and rep- The processing required to read a word
etition condition, F(1,17) = 16.28, M S e in the R e a d - R e a d condition, in contrast,
= 1.60, as well as a significant interaction is likely to be less than that r e q u i r e d
between these two variables, F(2,34) = to construct a solution, and apparently re-
64.71, M S e = .44. As shown in Figure 1, mains relatively constant across spacings of
there was an effect of spacing repetitions repetitions. The interaction between spac-
only in the R e a d - C o n s t r u c t condition. ing of repetitions and repetition conditions
Recall levels were identical for the R e a d - can be interpreted in terms of differences in
Construct and the R e a d - R e a d conditions the effectiveness of the second presentation
at zero spacing; h o w e v e r , the level of of an item as a cue for retrieval of the mem-
recall performance increased across the ory of its first presentation. According to
longer spacings of repetition in the R e a d - this suggestion, the second presentation of
Construct condition while remaining rela- an item in the R e a d - R e a d condition is
tively constant in the R e a d - R e a d con- nominally identical to its first presentation
dition. and so corresponds to a test of recognition
The above results replicate results re- memory, whereas a pair is altered prior to
ported earlier by Jacoby (1978), and can be its repetition in the R e a d - C o n s t r u c t condi-
interpreted as showing that the effects of tion and so its second presentation corre-
spacing repetitions depend on an influence sponds to a test of cued recall for memory
of its earlier presentation. Due to greater
cue effectiveness, accessibility provided by
a test of recognition memory is less influ-
'80 enced by increases in retention interval
(spacing) than is that provided by a test of
~ .7O cued recall. The greater influence of spac-
F--
zo. . . . . . . . . . o ~C
ing on the accessibility of the memory for a
I '60 / p r i o r p r e s e n t a t i o n o f an item in the
R e a d - C o n s t r u c t condition produces larger
u
.50 / j
differences in processing and, thereby, a
Q
>- more pronounced spacing effect than is
...a found in the R e a d - R e a d condition. The
I11 interaction of spacing repetitions and dif-
.3 I I I
ferences in cue effectiveness is further ex-
o 4 8
SPACING
amined in experiments reported later in this
FIG. 1. Probabilities of c o r r e c t cued recall from Ex-
paper.
p e r i m e n t 1 for R e a d - R e a d (RR) and R e a d - C o n s t r u c t As well as depending on the spacing of
(RC) conditions as a function of spacing interval. repetitions, the accessibility of a previously
 REPETITION EFFECTS 455

read solution in the Read-Construct con- studies have found increased retention
dition was expected to depend on the mate- when a difficult task rather than an easy one
rial that intervened between presentation of intervened between repetitions of an item
a problem and its solution. A second (Bjork & Allen, 1970; Proctor, 1980; Rob-
analysis included only items that had been bins & Wise, 1972; Tzeng, 1973). Bjork and
presented at longer spacings in the R e a d - Allen accounted for their results by sug-
Construct condition so as to examine the gesting that a more difficult intervening task
effect of similarity of intervening material. increased the probability that the second
This analysis revealed both a significant encoding of an item would be different from
effect of intervening material, F(1,17) = its first encoding. The primary difference
5.28, M S e = .01, and a significant inter- between our interpretation and that of
action between type of intervening ma- Bjork and Allen is that we place greater
terial and the spacing of repetitions, F (1,17) emphasis on the processing of later pre-
= 6.83, M S e = .01. In general, subsequent sentations of a repeated item. The impor-
cued recall performance was higher when tance of this difference between interpreta-
verbal items rather than addition prob- tions will be described in the general dis-
lems intervened between a puzzle and its cussion.
solution; that is, when the intervening ma-
terial was similar to the previously pre- EXPERIMENT 2
sented solution. The advantage in probabil- Experiment 2 employed the same general
ity of cued recall produced by presenting procedure as did Experiment 1. Rather than
verbal intervening material rather than manipulating the material intervening be-
math problems was larger at a spacing of tween presentations of a problem and its
four intervening items (.69 vs .51) than at a solution, however, the similarity of the pre-
spacing of eight intervening items (.65 vs sentations of repeated problems was ma-
.58). The interaction between spacing and nipulated in Experiment 2. The memory for
type of intervening material would be ex- a previously obtained solution to a problem
plained if similar intervening material pro- was expected to be more accessible when
duced a steeper forgetting function than did the later presentation of the problem was
dissimilar intervening material but a rela- similar rather than dissimilar to the earlier
tively small difference in the asymptotic presentation of that problem. An example
level of accessibility. Differences in acessi- from each of the four repetition conditions
bility as a function of type of intervening employed in the first phase of Experiment 2
material would then be maximal at shorter is given in Table 1. These four repetition
retention intervals (spacing) and decrease conditions were factorially combined with
as retention interval was lengthened to ap- two levels of spacing repetitions (0 vs 20
proach the asymptotic level of accessibility. intervening items) to produce eight experi-
Again, by this analysis, conditions that are mental conditions.
least conducive to gaining access to a pre- As shown in Table 1, items presented in
viously read solution when a problem is re- the first phase of the experiment were di-
peated produce higher subsequent cued re- vided among four repetition conditions, re-
call. quiring construction of a solution for a
Since the verbal intervening problems problem on the first presentation, the sec-
were more difficult in that they produced ond presentation, or on both presentations
more errors than did the addition problems, of a problem. In a Construct-Construct/
as well as being closer in similarity to the Different condition, different letters were
critical items, it is possible that differences deleted upon each presentation of a prob-
in task difficulty are at least partially re- lem, while the same letters were deleted
sponsible for retention differences. Other upon each presentation of a problem in the
456 CUDDY AND JACOBY

TABLE 1
PRESENTATION CONDITIONS AND EXAMPLES FROM EACH CONDITION FOR EXPERIMENT 2

First Second
Condition presentation presentation

Read-Construct (RC) LAWYER COURT LAWYER C--RT

Construct-Read (CR) LAWYER C--RT LAWYER COURT
Construct-Construct/Same (CCs) LAWYER C--RT LAWYER C--RT
Construct-Construct/Different(CCo) LAWYER C--RT LAWYER COU--

C o n s t r u c t - C o n s t r u c t / S a m e condition. condition, in contrast, subjects were first

Changing the form of a problem between its presented with the fragmented version of
presentations was expected to make the the pair, being required to construct a solu-
previously constructed solution less acces- tion, and then read the intact version of the
sible when the problem was repeated. That pair. This second condition had the same
is, the altered form of the problem was ex- number of exposures of the solution as did
pected to be a poorer cue for retrieval of the the first condition, but reading the solution
previously constructed solution than would came later so it could not trivialize the
be an exact repetition of the problem. Due solving of the problem and, thereby, lower
to this difference in accessibility, construc- subsequent cued-recall performance. When
tive processes are more likely to be involved a solution to a problem is read long before
in solving the second presentation of a presentation of the problem requiring that
problem in the C o n s t r u c t - C o n s t r u c t / solution, subsequent cued-recall perfor-
Different condition than in the Construct- mance in the R e a d - C o n s t r u c t condition
Construct/Same condition, at least when might converge with that in the Construct-
repetitions of a problem are widely spaced. Read condition. This is because at longer
Even when repetitions are widely spaced, intervals the previously read solution may
an exact repetition of a problem may be no longer be easily remembered, so it will
so effective as a cue for retrieval of an not trivialize responding to the later prob-
earlier constructed solution that responding lem. To account for effects of the order
to the problem for a second time involves of events, influences on processing must
little more processing than would simply be considered. Any advantage in subse-
reading a solution for that problem. This quent cued-recall performance of the Con-
possibility leads to the prediction that the s t r u c t - R e a d condition over the R e a d -
processing required to deal with a rep- Construct condition must be due to an in-
etition, and consequently, subsequent reten- fluence of m e m o r y for a prior presen-
tion performance will be approximately tation of an item on its later processing.
equal in the C o n s t r u c t - C o n s t r u c t / S a m e
and the C o n s t r u c t - R e a d conditions ex- Method
emplified in Table 1. Design and subjects. E x p e r i m e n t 2
Additional comparisons between condi- employed the same crossword puzzle task
tions in Table 1 allow assessment of the as did Experiment 1. The four repetition
persistence of the effect of reading a solu- conditions illustrated in Table 1 were facto-
tion to a problem prior to being presented rially combined with two levels of spacing
with that problem. In the Read-Construct of repetitions (0 vs 20 intervening items).
condition, subjects first read an intact ver- Both factors were manipulated within sub-
sion of a pair and were then presented with jects so that each subject served in each of
a fragmented version of that pair as a prob- the eight experimental conditions.
lem to be solved. In the Construct-Read The subjects were 16 students enrolled in
 REPETITION EFFECTS 457

an introductory psychology course who the e x p e r i m e n t , subjects were given a

served in the experiment for course credit. cued-recall test which contained the con-
Materials and procedure. The 96 pairs of text words from the 96 critical items as cues
words that served as critical items were for recall of the solution words from those
drawn from the same word pool as items in items.
Experiment 1. Each of the eight experi- Analyses. The manipulation of repetition
mental conditions resulting from the com- conditions in the first phase of the experi-
bination of repetition condition and spacing ment would be expected to influence the
of repetitions was represented by 12 pairs probability of a subject correctly solving a
within a list. To produce word problems, presented problem. For example, when the
two letters were deleted from the right-hand solution to a problem is presented prior to
member of a pair; the first letter of a word that problem as in the R e a d - C o n s t r u c t
was never among the deleted letters. To condition, subjects should be more likely to
produce problems for the C o n s t r u c t - C o n - give the correct solution for the problem
struct/Different condition, pairs were used than when the solution has not been previ-
to produce two problems that differed only ously presented. Consequently, differences
in the letters that were deleted. The con- in cued recall may reflect differential expo-
struction of lists employed 96 additional sure to solution words produced by dif-
pairs of words that were noncritical in that ferences in the probability of correct l y
their recall was not tested; they appeared solving problems. At the extreme, subjects
in a study list intervening between repeti- cannot be expected to recall a word that
tions of a critical item. The left-hand mem- they failed to generate and, therefore, never
ber of each of these noncritical pairs was encountered in the first phase of the ex-
repeated and repetitions were made to ap- periment. To counter this difficulty, the
pear similar to those in the C o n s t r u c t - cued-recall data that are report ed were
Construct/Different condition. For noncrit- conditionalized on correct responding in
ical pairs, however, each presentation of a phase 1 of the experiment. Uncondition-
pair containing a particular context word alized data were also analyzed but are
required a different solution word. For not reported since both the direction of re-
example, intervening pairs might be "RIGID sults and conclusions drawn are the same
S T - - F " followed by " R I G I D S T - - C T " ; for conditionalized and unconditionalized
solution words for these pairs are STIFF data.
and STRICT, respectively. The similarity
of noncritical pairs to critical pairs was Results and Discussion
expected to encourage subjects to attend There were very few errors in solving
to the letters of a solution word that were problems in the first phase of the experi-
provided when a problem was repeated. ment. Differences in the probability of an
Eight list formats were produced such that error, however, did provide some evidence
across formats each pair of critical items of an effect of memory for a previously
equally often represented each of the com- presented solution. The probability of an
binations of spacing and form of repeti- error in the C o n s t r u c t - R e a d condition was
tions. The noncritical items intervening be- .09 while that in the R e a d - C o n s t r u c t con-
tween repetitions of critical items were held dition was .01, so previously reading a so-
constant across formats. lution to a problem did decrease the proba-
The procedure followed was the same as bility of an error when the problem was
that in Experiment 1. Subjects were paced later presented. The probability of an error
through the 288 cards, with one pair of on the second presentation of a problem in
items presented on each card, at a rate of 6 the C o n s t r u c t - C o n s t r u c t / S a m e and the
seconds per card. In the second phase of C o n s t r u c t - Construct/Different condition
458 C U D D Y A N D JACOBY

were .06 and .02, respectively. Contrary to viously read solution were to become to-
expectations, reducing the similarity of tally inaccessible, the processing required
repetitions apparently did not reduce ac- to solve a problem in the Read-Construct
cess to memory for a prior presentation of condition should not differ from that re-
an item and, thereby, produce more errors quired to solve a problem in the Con-
when that item was repeated. struct-Read condition, and, consequently,
The cued-recall results from the second cued-recall performance in the two condi-
phase of the experiment are plotted in Fig- tions should be equal. Even the longer
ure 2. Analysis of these results revealed a spacing of presentations employed in the
significant effect of repetition condition, experiment, however, was insufficient to
F(3,45) = 23.04, MSe = .02, and an effect produce equality of these two conditions in
of spacing repetitions, F(1,14) = 111.38, cued-recall performance. The effect of pre-
M Se = .01, as well as a significant interac- viously reading a solution can apparently
tion between repetition condition and the persist over a relatively long period of time
spacing of repetitions, F(3,45) = 7.02, to influence the processing required to
M Se = .01. solve a later problem, and reduce sub-
Consistent with predictions made earlier, sequent retention performance.
the effect of spacing repetitions was more It was earlier suggested that an exact
pronounced in the Read-Construct condi- repetition of a problem can be a very effec-
tion than in the other repetition conditions. tive cue for retrieval of a previously con-
As in Experiment 1, the spacing effect in structed solution so that little processing
the Read-Construct condition can be in- may be required to allow responding to a
terpreted as being due to the previously second presentation of a problem. In line
read solution becoming less accessible due with this suggestion, cued-recall perfor-
to increases in spacing. Further processing mance in the Construct-Construct/Same
is required to solve a problem as spacing is condition did not differ substantially from
increased and retention benefits from this that in the C o n s t r u c t - R e a d condition.
further processing. If memory for the pre- Even when repetitions were spaced, a pre-
viously constructed solution was appar-
,..1 ently so readily accessible that the pro-

~
•"J "85 cessing required to respond to an identical
CC D
I1:
repetition of a problem differed little from
. - " o CC s
.75 that required to simply read the solution to
W that problem. Varying the form of a prob-
o .65 lem between its repetitions, in contrast,
F-
W was expected to reduce the effectiveness of
n- "55
n- the second presentation of the problem as a
O
cue for retrieval of the solution constructed
LL '45 on its first presentation, particularly when
0
>- repetitions were widely spaced. This re-
.35 duced accessibility was expected to neces-
..I
sitate further processing at the time of the
m -25 second presentation of a problem to arrive
I I
0 2O at a solution, and, consequently, enhance
SPACING subsequent retention performance. Perfor-
Fio. 2. Probability of correct cued recall for con- mance in the C o n s t r u c t - C o n s t r u c t / D i f -
ditionalized data of Experiment 2 for R e a d - C o n s t r u c t
(RC), C o n s t r u c t - R e a d (CR), C o n s t r u c t - C o n s t r u c t /
ferent condition was marginally superior
S a m e (CCs), and C o n s t r u c t - C o n s t r u c t / D i f f e r e n t but not significantly different from that in
(CCD) conditions as a function of spacing interval. the Construct-Construct/Same condition.
 REPETITION EFFECTS 459

The effect of similarity of repetitions of a respond with that of repetitions in the

problem was further investigated in Ex- C o n s t r u c t - C o n s t r u c t / D i f f e r e n t condition
periment 3 in an attempt to magnify the ef- o f E x p e r i m e n t 3 (i.e., E x p e r i m e n t 2:
fects observed in Experiment 2. " J O S T L E SH-K-" and JOSTLE SH-V-";
E x p e r i m e n t 3: " J O S T L E S H - K - " and
EXPERIMENT 3
" J O S T L E --OVE").
Experiment 3 was identical to Experi- The procedure, instructions, and cued-
ment 2 with the exception that problems re- recall test were identical for Experiments 2
p e a t e d in the C o n s t r u c t - C o n s t r u c t / and 3. As in Experiment 2, cued-recall
Different condition were altered to make performance was conditionalized on cor-
their repetitions even more dissimilar. This rectly responding to a problem in the first
change in materials was expected to in- phase of the experiment. Unconditionalized
crease cued recall of items presented in the scores were also analyzed but since the di-
C o n s t r u c t - C o n s t r u c t / D i f f e r e n t condition rection of d i f f e r e n c e s and c o n c l u s i o n s
while having little influence on recall from drawn were the same only conditionalized
the other conditions. data are reported.

Method Results and Discussion

Design and subjects. The design of Ex- As in Experiment 2, the probability of an
periment 3 was identical to that of Experi- error in solving a problem during the first
ment 2. The subjects were 16 students en- phase of the experiment reflected memory
rolled in an introductory psychology class for a previously encountered solution. The
who served in the experiment for course probability of an error was higher in the
credit. C o n s t r u c t - R e a d condition (.086) than in
Materials and procedure. Stimuli were the R e a d - C o n s t r u c t condition (.005). The
identical to those employed in Experiment probability of an error on the second pre-
2 with the following exceptions: First, the sentation of a problem in the C o n s t r u c t -
s e c o n d p r e s e n t a t i o n of an item in the Construct/Same and Construct-Construct/
C o n s t r u c t - C o n s t r u c t / D i f f e r e n t condition Different conditions were .062 and. 10. The
was altered to appear as different as possi- probability of an error in these latter two
ble from the first presentation of that item. conditions was conditionalized on correct
In contrast to Experiment 2, the first letter responding to a problem on its first pre-
of the solution word was always present in sentation for further analyses. An analysis
the first presentation of a problem and de- of those conditionalized probabilities of an
leted from the second presentation of a error revealed both an effect of spacing rep-
problem in Experiment 3 (i.e., Experiment etitions, F(1,15) = 8.93, M S e = .003 and an
2: " L A W Y E R C--RT" and " L A W Y E R effect of repetition condition, F(1,15) =
COU--; Experiment 3: " L A W Y E R C--RT" 5.36, MSe = .001. Given that a problem
and " L A W Y E R - O U R - " ) . Second, items was solved correctly on its first presenta-
that appeared in the study list intervening tion, changing the problem prior to its rep-
between repetitions were altered to make etition served to increase the probability of
them appear similar to items repeated in the an error as did increasing the spacing of
Construct-Construct/Different condition. repetitions. The conditionalized probability
This change between experiments corre- of an error at 0 and at 20 spacing in the
sponded with the change in critical items. Construct-Construct/Same condition were
The second presentation of a noncritical .004 and .022; the corresponding prob-
context word was paired with a new solu- abilities in the C o n s t r u c t - C o n s t r u c t /
tion word as in Experiment 2 but the pat- Different condition were .047 and .069.
tern of deleted letters was changed to cor- Mean probabilities of cued recall are dis-
460 C U D D Y A N D JACOBY

played in Figure 3. The results of Experi- condition, F(1,15) = 9.47, M S e = .01. This
ment 3 generally support the findings of difference between the Same and Different
Experiments 2. Analysis of the cued-recall repetition conditions was also significant in
data revealed significant effects of repeti- the analysis of unconditionalized cued re-
tion condition, F(3,45) = 28.43, MSe = .02, call. As in Experiment 2, the Construct-
and the spacing of repetitions, F(1,15) = Construct/Same and the Construct-Read
97.76, M S e = .01, as well as a significant conditions produced approximately equal
interaction between those two variables, levels of cued-recall performance.
F(3,45) = 3.12, M S e = .02. As in Experi- The important comparison to make in
ment 2, cued-recall performance in the Experiment 3 is of conditions that allow a
Read-Construct condition was lower than solution to a problem to be easily retrieved
that in the Construct-Read condition even or read (Construct-Construct/Same and
at the longer level of spacing of repetitions. Construct-Read) with conditions that re-
Although the previously read solution ap- quire more extensive processing of the sec-
parently became less accessible in the ond presentation of an item when repeti-
Read-Construct condition as the spacing tions are spaced ( C o n s t r u c t - C o n s t r u c t /
of repetitions was increased, memory for Different and Read-Construct). It appears
the solution was still having some influence that when exactly the same problem is en-
on processing of the later presented prob- countered twice, the second presentation is
lem even at the longer spacing of repeti- treated essentially like items that are pre-
tions. Of greater concern, the further re- sented to be read. Processing of the second
duction of the similarity of repetitions in the presentation is minimal in these conditions
Construct-Construct/Different conditions and that presentation does relatively little
had the predicted effect. At the longer to enhance subsequent retention. On the
spacing of repetitions, the Construct-Con- other hand, the results of solving different
struct/Different condition produced sub- forms of the same problem are much the
stantially higher cued-recall performance same as those of constructing a solution
than did the Construct-Construct/Same that has been previously read. The effect of
spacing repetitions was similar in the
.J C o n s t r u c t - C o n s t r u c t / D i f f e r e n t and the
"85! R e a d - C o n s t r u c t conditions although the
CCD
Construct-Construct/Different condition
n..75
CCS held a large overall advantage due to the
.-t- :8 CR subjects having solved problems on their
u .65 first presentation as well as their second. In
I--
(J both conditions, increasing the spacing of
ILl
n- .55 repetitions is seen as having the effect of
O i
making a previously encountered solution
u. .45 less accessible so that more processing is
O
>- necessary to deal with a repetition of a
I-- -35
problem; this additional processing en-
hances retention performance.
m~ "25
O I I It should be noted that repetition condi-
0 20 tion was confounded with problem diffi-
SPACING
culty in the present experiment. For the
FIG. 3. Probability of correct cued recall for con-
second presentation of a problem in the
ditionalized data of Experiment 3 for R e a d - C o n s t r u c t
(RC), C o n s t r u c t - R e a d (CR), C o n s t r u c t - C o n s t r u c t /
Construct-Construct/Different condition,
S a m e (CCs), and C o n s t r u c t - C o n s t r u c t / D i f f e r e n t terminal letters were deleted while it was
(CC,) conditions as a function of spacing interval. always interior letters that were deleted
 REPETITION EFFECTS 461

from the solution word to form problems in tion of spacing should have been larger in
the Construct-Construct/Same condition. the Different than in the Same repetition
It might be argued that the deletion of ter- condition. The probability of an error in
minal letters produced more difficult prob- solving problems was low in all conditions
lems, and that the difference in problem so that measures may have simply been to
difficulty rather than a difference in simi- insensitive to detect an interaction between
larity of repetitions was responsible for the spacing and repetition conditions.
observed effects. Indeed, the error data By the cue effectiveness account, the
from the first phase of the experiment pro- second presentation of a problem serves as
vide evidence that the second presentation a cue for memory of its first presentation
of a problem in the Different repetition and the similarity of repetitions influences
condition was more difficult than that in the cue effectiveness. By the problem difficulty
Same repetition condition. To account for account, in contrast, it might be claimed
the cued recall results in terms of differ- that at the wider spacing of repetitions a
ences in problem difficulty, however, it problem is solved without reference to its
must be claimed that differences in diffi- prior presentation, and it is the difficulty of
culty were effective only when repetitions the problem, solved in isolation, that is im-
were widely spaced. It was only at the portant for later retention. Results reported
wider spacing that cued recall produced by by Jacoby (1978, Expt. 2) are useful for
the Construct-Construct/Different condi- choosing between these two alternative ac-
tion was superior to that produced by the counts. That experiment employed proce-
Construct- Construct/Same condition. Any dures similar to those employed in the pres-
differences in problem difficulty as a func- ent experiment but deleted either one or
tion of the spacing of repetitions is likely to two interior letters from target words to
be attributable to differential accessibility vary problem difficulty. Problems were
to memory for a previously constructed presented only once to be solved or were
solution to a problem when that problem is preceded by a presentation of an intact ver-
repeated. That is, when one has just con- sion of the pair that was to be read prior to
structed a solution to a problem prior to its the problem being presented to be solved,
being repeated, the solution to the problem as in the R e a d - C o n s t r u c t condition of
is likely to be readily accessible so deleting the present experiment. In the Read-Con-
initial letters from the solution word to pro- struct conditions, the spacing of repeti-
duce a second version of the problem will tions was also varied. The manipulation
not effectively increase problem difficulty. of problem difficulty was successful in
At the wider spacing of repetitions, mem- that deleting two interior letters produced
ory for the previously constructed solution more errors than did deleting one interior
may be less accessible so the same manipu- letter when pairs were presented only as a
lation does produce an effective increase problem to be solved. Despite this effect on
in problem difficulty. When expanded in errors, subsequent cued recall of those
this fashion to account for the interaction of once presented items was not influenced by
repetition condition and spacing, the ac- the manipulation of problem difficulty. In
count of results in terms of problem diffi- the Read-Construct conditions, memory
culty is similar to an account in terms of of a prior presentation influenced the later
differential cue effectiveness. For both ac- solving of the more difficult problems as
counts, the lack of a significant interaction shown by a reduction in errors in solving
between repetition condition and spacing in those problems as compared to the corre-
the error data from the problem-solving sponding o n c e - p r e s e n t e d condition in
phase of the experiment poses a difficulty. which a pair was not previously read. In
The increase in number of errors as a func- contrast to the results obtained for once-
462 CUDDY AND JACOBY

presented items, the manipulation of prob- G E N E R A L DISCUSSION

lem difficulty did influence subsequent
cued recall in the Read-Construct condi- Melton (1967) described the effect of
tions. At the wider spacing of repetitions, spacing repetitions as presenting a paradox
the more difficult problems produced sub- in that it seemed to show that forgetting is
stantially higher subsequent cued recall beneficial for memory. Results of the pres-
than did the easier problems. That the ef- ent studies provided support for Melton's
fect of problem difficulty in subsequent contention that the effect of repetition is
cued recall was confined to the repetition greater when memory for an earlier pre-
conditions and was then only substantial at sentation of the repeated item is less acces-
wider spacings, implies that cue effective- sible. Experiment 1 led to the conclusion
ness played a critical role in producing that similar material intervening between
those results, and supports the argument item presentations results in higher recall at
that it was cue effectiveness rather than a later test. Experiments 2 and 3 led to two
problem difficulty without reference to a main conclusions: First, memory for a prior
prior presentation of a problem that was presentation of an item can act over at least
important for producing the results of the a few minutes to influence the processing
present experiment. involved in solving a problem. That it was
Regardless of whether the cue effective- processing of the problem that was in-
ness or problem difficulty account of the volved was shown by the differential effects
results is favored, it is clear that memory of reading the solution to a problem prior to
for a prior presentation of an item did play rather than after solving the problem. Sec-
some role in producing the results. By both ond, varying the form of a problem between
accounts, it is the accessibility of memory its presentations renders memory for a
for a prior presentation that was responsi- prior presentation of the problem less ac-
ble for eliminating the difference between cessible when that problem is repeated so
the Same and Different conditions when that repetition of the problem engenders
repetitions were massed. A useful direction greater processing and, consequently, sub-
for future research may be to further inves- sequent retention performance is enhanced.
tigate the effect of problem difficulty and Emphasizing the role of forgetting in pro-
the sequencing of problems. In an investi- ducing the effect of spacing repetitions has
gation of the effects of spacing repetitions, the advantage of relating the spacing effect
Landauer and Bjork (1978) found that a se- to other memory phenomena. One effect of
quence that involved successively increas- forgetting a prior presentation of an item is
ing the spacing o f repetition produced to make the processing of a later presenta-
higher subsequent retention performance tion of that item more difficult, and several
than did a sequence in which the spacing of experiments have demonstrated that task
repetition was initially larger and then suc- difficulty influences subsequent retention
cessively reduced across presentations. performance. Jacoby, Craik, and Begg
Similarly, successively increasing the diffi- (1979) reported three new experiments and
culty of a problem as recommended by the reviewed other experiments to show that
"fading" procedure employed by operant increasing the difficulty of judgments can
psychologists may enhance retention per- enhance subsequent retention of items in-
formance. Results of the present experi- volved in those j u d g m e n t s . Similarly,
ment and those reported by Jacoby (1978) Slamecka and Graf (1978) showed that the
can be used to suggest that variations in cue more difficult task of generating a response
effectiveness play a central role so that both to an item produces higher subsequent re-
the spacing of repetitions and differences in tention performance than does reading a re-
problem difficulty will be important. sponse to an item. Effects of task difficulty
 REPETITION EFFECTS 463

are also found when the effects of a prior repeated each time that item is rehearsed as
test on subsequent retention performance a member of a small set of items. Similar to
are considered. Similar to the effects of experimenter-provided repetitions, the ease
spacing repetitions, increasing the spacing of the accessibility to memory for a prior
of the presentation and test of an item en- rehearsal of an item is likely to reduce the
hances subsequent retention performance. processing that is necessary to produce a
A prior test of an item does more to en- later rehearsal of that item. That is, the in-
hance later retention if that test is delayed effectiveness of maintenance rehearsal may
by the presentation of intervening material reflect a drastic reduction in processing
rather than being directly preceded by pre- between the original encoding of an item
sentation of the item that is to be tested and subsequent rehearsals of that item.
(e.g., Landauer & Eldridge, 1967; Gotz & The question remains as to why forget-
Jacoby, 1974; Whitten & Bjork, 1977). ting between repetitions or increasing task
Further, later retention performance shows difficulty by other means should enhance
a greater benefit if an earlier test of reten- memory performance. The most popular
tion is made more difficult by giving a recall account of the effect of spacing repetitions
rather than a recognition test of memory attributes that effect to an influence of
(e.g., Bjork & Whitten, 1974). This latter spacing on encoding variability. Theories of
effect of test difficulty can be seen as par- this form hold that repetition enhances re-
alleling the effects of similarity of repeti- tention performance only to the extent that
tions investigated in Experiments 2 and 3 the encoding of a later presentation of an
reported earlier; increasing the cues for re- item does not repeat that of an earlier pre-
trieval provided either at an earlier test or sentation. The spacing effect is attributed to
by the repetition of an item reduces sub- spaced repetition producing greater dissim-
sequent retention performance. ilarity among the encoded versions of a re-
The ineffectiveness of maintenance re- peated item. This greater variability in en-
hearsal for improving long-term retention coding is said to increase the number of re-
can also be interpreted as being due to an trieval routes to memory of a repeated item,
influence of memory for a prior encounter and to increase the probability of overlap
on the processing of an item when it is re- between information provided by a re-
peated. Manipulations of maintenance vs trieval cue and that contained in a memory
elaborative rehearsal are similar to manip- trace of repeated item. The most sophisti-
ulations of spacing of repetitions except cated account of the spacing effect in terms
that in the former paradigm it is the spacing of encoding variability has been put for-
of rehearsals rather than the spacing of ward by Glenberg (1979). The account of
experimenter-provided repetitions that is the ineffectiveness of maintenance rehear-
varied. To show the ineffectiveness of sal for long-term retention offered by Craik
maintenance rehearsal, it has been demon- and Lockhart (1972) is similar to the en-
strated that rehearsing a small set of items coding variability theory used by others to
in order to maintain those items in primary account for the effect of spacing repeti-
memory does not enhance performance on tions. According to Craik and Lockhart,
a delayed recall test (e.g., Craik & Watkins, repeatedly processing an item at the "same
1973; Jacoby & Bartz, 1972). This mainte- level" serves to maintain that item in pri-
nance rehearsal involves massed rehearsal mary memory but will not benefit sub-
of items and is likely to reflect influences on sequent memory performance. To benefit
processing that are similar to those pro- secondary memory, it is said that further
duced by the presentation of massed repe- study of an item must involve deeper pro-
titions. It is unlikely that the processing re- cessing or serve to elaborate the memory
quired to initially encode an item must be trace of the item. As claimed by encoding
464 C U D D Y AND JACOBY

variability theory, then, Craik and Lock- suggested that the effect of repetition is di-
hart see rehearsal as being effective only minished if an item is r e p r e s e n t e d in
to the extent that later processing or en- short-term memory when it is repeated
coding of an item does not repeat that of (e.g., Greeno, 1967; Whitten & Bjork,
earlier presentations or rehearsals. The two 1977). Rather than employing the distinc-
views are made even more similar by the tion between short-term and long-term
suggestion that maintenance and elabora- memory, we emphasize differences in ac-
tive rehearsal are better seen as end points cessibility, and the influence of memory for
on a continuum rather than as discrete a prior presentation of an item on the pro-
categories (Craik & Jacoby, 1975; Glenberg cessing of a later presentation. We believe
& Adams, 1978). Elaboration and variabil- that repeated processing of an item can en-
ity in encoding are similar notions in that hance memory performance but that pro-
both are described as having their effect by cessing will only be repeated if memory for
increasing the number of access routes to a prior presentation of an item is not readily
the memory for an item. The effects of de- accessible. Maintenance rehearsal and
cision difficulty can also be attributed to an massed repetition are seen as being similar
influence of decision difficulty on elabora- in that relatively few of the operations
tion of the memory for a presented item. A originally required to encode an item are
more difficult decision requires further pro- likely to be repeated. Any strengthening
cessing and, consequently, a more elabo- effect will be limited to those operations
rate or distinctive memory trace results that are actually repeated. By our view,
(Jacoby et al., 1979). The notion of distinc- then, it is not that repeated processing does
tiveness differs from that of elaboration nothing to enhance memory but rather that
only in that distinctiveness requires that much of the processing of an item is not
conditions of test as well as those of study repeated when repetitions are massed.
be taken into account. An implication of the emphasis on
We agree that learning more about an forgetting between presentations is that it is
item as a result of encoding variability or the encoding of later presentations of an
elaboration can enhance memory perfor- item that is influenced by the spacing of
mance. However, we would also like to repetitions. Hintzman, Black, and Sum-
entertain the possibility that repetition can mers (1973) demonstrate effects of spacing
produce a "strength-like" effect on mem- on the encoding of later presentations and
ory. By an encoding variability theory, the suggest a habituation account of the spacing
effect of repeating an item should be maxi- effect that is compatible with the view that
mal when there is no overlap between the we propose. By our view, accessibility of
encodings of a repeated item so that inde- memory for a prior presentation can result
pendent traces are formed. However, the in the "dropping out" of some encoding
effect of repeating an item has been found operations and, t h e r e b y , a more im-
to be larger than could be produced by in- •poverished trace. If the processing of later
dependent traces (Jacoby, Bartz, & Evans, presentations is abbreviated, one would
1978; Ross & Landauer, 1978). One way to expect a corresponding reduction in the
account for this too-large effect of repetition amount of effort invested in the processing
is to assume that traces of an item can some- of those later presentations. In line with this
times act in concert to have a strength-like possibility, Johnston and Uhl (1976) have
effect (Jacoby et al., 1978). Further, we reported a positive correlation between
suggest that neither elaboration nor a processing effort indexed by performance
strength-like effect will occur if memory for on a subsidiary task and subsequent reten-
a prior presentation is too readily accessible tion performance. Massed repetitions re-
when that item is repeated. Others have quired less effort to process and produced
 REPETITION EFFECTS 465

poorer retention performance than did are no more similar than would be produced
spaced repetitions. by the presentation of two unrelated words.
The effect of changing modifiers between The emphasis on forgetting between rep-
repetitions has been interpreted as evidence etitions has substantial heuristic value. For
for the encoding variability account of the the spacing effect, the implication is that
spacing effect (e.g., Madigan, 1969) but is, the details of that effect should change
perhaps, better interpreted in terms of its across other manipulations that influence
influence on the accessibility of memory for the accessibility of memory for a prior pre-
a prior presentation when an item is re- sentation of an item when that item is re-
peated. It is sometimes found that the peated. In the experiments reported earlier,
spacing effect is attenuated when a change we found interactions of spacing with type
in modifiers is used to bias a different of intervening material and the similarity of
meaning of a repeated word upon each of its repetitions. The manipulation of type of
presentations (Madigan, 1969). In Experi- intervening material c o r r e s p o n d s to a
ment 3, however, changing an item between manipulation of retroactive inhibition.
its repetitions produced a more pronounced Spacing would also be expected to interact
spacing effect rather than attenuating the with manipulations of proactive inhibition.
spacing effect as would be predicted by an Other factors that would be expected to be
encoding variability theory. This exaggera- important are factors such as modality of
tion of the spacing effect is understandable presentation and the level-of-processing of
as an interaction between the cues for re- prior presentations. In sum, the manipula-
trieval provided by a repetition and length tion of spacing repetitions corresponds to a
of the retention interval (spacing). At zero manipulation of retention interval and the
spacing, memory for the prior presentation effect of that manipulation should be mod-
of an item is readily accessible so the re- erated by other factors that also influence
duction in cues for retrieval produced by acessibility to memory for a repeated item's
changing an item prior to its repetitions was prior presentation.
ineffectual. A change in cues, however, did Glenberg and Smith (1981) criticized an
have an effect at the longer retention inter- earlier paper by Jacoby (1978) as attributing
val produced by wider spacing, presumably the spacing effect totally to differences in
because memory for the prior presentation encoding and as ignoring the importance of
was less accessible by other means so the retrieval. The encoding variability view ad-
cues for retrieval provided by the presented vanced by Glenberg, in contrast, was de-
version of the repeated item were more im- scribed as taking both differences in en-
portant. To eliminate the effect of spacing coding and differences in retrieval into ac-
repetitions by changing modifiers, our view count. Rather than ignoring retrieval, we
suggests that the change in an item between see the emphasis on forgetting between
its repetitions must be so drastic that the repetitions as saying that retreival is im-
second presentation of an item will not pro- portant during study as well as at the time
vide access to memory for its first presen- of test. During study, retrieval is seen as
tation even when the two presentations are operating to influence the processing of
contiguous in the list. It seems reasonable later presentations of a repeated item. If the
that a change in modifiers can sometimes trace of a prior presentation is too readily
have such profound effects. When homo- accessible when an item is repeated, few of
graphs are employed as stimuli a change in the operations originally required to en-
modifiers between repetitions of an item of code that item will be repeated and the re-
the sort employed in some experiments sult will be an impoverished trace of the
may have so large an effect on encoding as later presentation. A more telling criticism
to produce traces of the repeated item that might be that our approach allows so many
466 CUDDY AND JACOBY

factors to operate that it predicts that al- (Eds.), Cognitive Theory, Vol. I. Potomac, Md.:
most anything can happen. What is needed Erlbaum, 1975.
CRAIK, F. I. M., & LOCKHART, R. S. Levels of pro-
is some measure of forgetting between rep-
cessing: A framework for memory research.
etitions that is independent of subsequent Journal of Verbal Learning and Verbal Behavior,
retention performance. In this regard, inci- 1972, 11, 671-684.
dental learning procedures of the sort CRAIK, F. I. M., & WATKINS, M . J . The role of re-
employed in the present experiments are hearsal in short term memory. Journal of Verbal
likely to be particularly useful. It may be Learning and Verbal Behavior, 1973, 12,
599- 607.
possible to use effects on the probability of GLENBERG, A. M. Component-levels theory of the ef-
correctly solving a problem and effects on fects of spacing of repetitions and recall and rec-
reaction time to analyze the processing of ognition. Memory and Cognition, 1979, 7,
later presentations of a problem so as to 95-112.
arrive at an independent measure of forget- GLENBERG, A. M., & ADAMS, F. Type 1 Rehearsal
and recognition. Journal of Verbal Learning and
ting between repetitions that can then be Verbal Behavior, 1978, 17, 455-463.
related to subsequent retention perfor- GLEN~ER6, A. M., & SMITH, S.M. Spacing repeti-
mance. tion, and solving problems are not the same.
The message of a considerable body of Journal of Verbal Learning and Verbal Behavior,
1981, 20, 110-119.
recent research has been that it is not words
GoTz, A., & JACOBY, L. L. Encoding and retrieval
per se that are remembered but rather what processes in long-term retention. Journal of Ex-
is remembered in some product of a pre- perimental Psychology, 1974, 102, 291-297.
sented word and the processing activities of GREENO, J. G. Paired-associate learning with short-
the learner (e.g., Craik & Lockhart, 1972; term retention: Mathematical model and data re-
garding identification of parameters. Journal of
Jenkins, 1974) or perhaps the operations
Mathematical Psychology, 1967, 4, 430-472.
performed on a presented word (Kolers, HINTZMAN, D . L . Theoretical implications of the
1976). The approach to analyzing repetition spacing effect. In R. L. Solso (Ed.), Theories of
effects that we propose continues in the cognitive psychology: The Loyola Symposium.
above vein. To specify the effect of repeti- New York: Edbaum, 1974.
HINTZMAN, D. L., BLACK, R. A., & SUMMERS, J. J.
tions, it must be determined which, if any,
Modality tags and memory for repetitions: Locus
processing is repeated across presentations of the spacing effect. Journal of Verbal Learning
of an item, and then plot performance and Verbal Behavior, 1973, 12, 229-238.
against that which is truly repeated. One JACOBY, L. L. On interpreting the effects of repetition:
factor that likely influences the probability Solving a problem versus remembering a solu-
tion. Journal of Verbal Learning and Verbal Be-
of processing being repeated is the accessi- havior, 1978, 17, 649-667.
bility of the memory for a prior presentation JACOBY, L. L., & BARTZ, W. H. Rehearsal and trans-
of an item when that item is repeated. fer to LTM. Journal of Verbal Learning and Ver-
bal Behavior, 1972, 11, 561-565.
REFERENCES JACOBY, L. L., BARTZ, W. H., & EVANS, J. D. A
BJORK, R. A., 8~ ALLEN, T. W. The spacing effect: functional approach to levels of processing. Jour-
Consolidation or differential encoding? Journal of nal of Experimental Psychology: Human Learn-
Verbal Learning and Verbal Behavior, 1970, 9, ing and Memory, 1978, 4, 331-346.
567-572. JACOBY, L. L., CRAIK, F. I. M., & BEGG, I. Effects of
BJORK, R. A., & WHITTEN, W. B. Recency sensitive decision difficulty on recognition and recall. Jour-
retrieval processes in long term free recall. Cog- nal of Verbal Learning and Verbal Behavior,
nitive Psychology, 1974, 6, 173-189. 1979, 18, 585-600.
BOUSFIELD, W. A., Cohen, B. H., WHITMARSH, JENKINS, J. J. Remember that old theory of memory?
G. A., & KINCAID, W. D. The Connecticut free Well forget it. American Psychologist, 1974, 29,
association norms. Studies on the mediation o f 785-795.
verbal behavior. Technical report No. 35, Univer- JOHNSTON, W. A., & UHL, C. N. The contributions of
sity of Connecticut, November 1961. encoding effort and variability to the spacing ef-
CRAIK, F. I. M., & JACOBY, L. L. A process view of fect and free recall. Journal of Experimental Psy-
short-term retention. In F. Restle, R. N. Shiffrin, chology: Human Learning & Memory, 1976, 2,
H. J. Castellan, M. R. Lindman, & D. B. Pisoni 123-160.
 REPETITION EFFECTS 467

KOLERS, P. A. Reading a year later. Journal of Ex- ties of spacing effects. Journal of Verbal Learning
perimental Psychology: Human Learning and and Verbal Behavior, 1978, 17, 669-680.
Memory, 1976, 2, 554-565. SLAMECKA,N. J., & GRAF, P. The generation effect:
LANDAUER, T. K., • ELDRIDGE, L. Effects of tests Delineation of a phenomenon. Journal of Experi-
without feedback and presentation-test interval in mental Psychology: Human Learning & Memory,
paired-ass0ciated learning. Journal of Experi- 1978, 4, 592-604.
mental Psychology, 1967, 75, 290-298. TZENG, O. J. L. Stimulus meaningfulness, encoding
MADIGAN, S. A. Intraserial repetition and coding pro- variability, and the spacing effect. Journal of Ex-
cesses in free recall. Journal of Verbal Learning perimental Psychology, 1973, 99, 162-166.
and Verbal Behavior, 1969, 8, 828-835. WHITTEN, W. B., & BJORK, R . A . Learning from
MELTON, A. W. Repetition and retrieval from mem- tests: Effects of spacing. Journal of Verbal
or3/. Science, 1967, 158, 532. Learning and Verbal Behavior, 1977, 16,
PROCTOR, R. W. The influence of intervening tasks on 465-478.
the spacing effect for frequency judgments. Jour-
nal of Experimental Psychology: Human Learn-
ing and Memory, 1980, 6, 254-266. REFERENCE N O T E
ROBBINS, D., & WISE, P. S. Encoding variability and 1. LANDAUER, T. K., & BJORK, R. A. Optimum re-
imagery: Evidence for a spacing-type effect with- hearsal patterns and name learning. Paper pre-
out spacing. Journal of Experimental Psychology, sented at the International Symposium on Practi-
1972, 95, 229-230. cal Aspects of Memory, Cardiff, Wales, 1978.
Ross, B. H., & LANDAUER, Z. K. Memory for at least
one of two items: Test and failure of several theo- (Received March 10, 1981)