Distributed Rereading can Hurt the Spacing Effect inText Memory

PETER P. J. L. VERKOEIJEN*,REMY M. J. P. RIKERS and BINNUR OZSOY

Department of Psychology, Erasmus University Rotterdam, The Netherlands

SUMMARY

The spacing effect refers to the commonly observed phenomenon that memory for spaced repetitionsis better than for massed repetitions. In the present study, we examined this effect in students’memory for a lengthy expository text. Participants read the text twice, either in immediate succession(massed repetition), with a 4-day interstudy interval (spaced short), or with a 3.5-weeks interstudyinterval (spaced long). Two days after the second study trial, all participants were tested. The resultsdemonstrated that students in the spaced-short condition remembered more of the content than thosein the massed condition. By contrast, students in the spaced-long condition remembered as much asstudents in the massed condition. These results were interpreted in terms of a theoretical framework,which combines mechanisms of encoding variability and study-phase retrieval to account for thespacing effect. Copyright # 2007 John Wiley & Sons, Ltd.

Memory for repeated items improves when learning trials are separated by other items

or events (i.e. spaced repetitions) than when they are presented in immediate succession

(i.e. massed repetitions). This phenomenon has been dubbed as the spacing effect and it

has been demonstrated with diverse populations, with a broad range of stimulus

materials, and in both explicit and implicit memory tasks (e.g. Challis, 1993; Glover

& Corkilll, 1987; Greene, 1989, 1990; Krug, Davis, & Glover, 1990; Mammarella,

Russo, & Avons, 2002; for reviews see Crowder, 1976; Dempster, 1996; Hintzman, 1974,

1976).

Recently, however, Rawson and Kintsch (2005) identified a boundary condition on the

spacing effect in memory of expository texts. In each of the two experiments in this study,

participants read a lengthy expository text twice, either in a massed fashion or in a spaced/

distributed fashion with a 1-week delay between repeated trials. Following the second

study trial, they received a memory test on the text after a retention interval of 5minutes

(short retention interval) or after a retention interval of 2 days (long retention interval). The

results showed that the spacing effect interacts with the length of the retention interval.

That is, with a short retention interval, massed restudying yielded a better memory

performance than spaced restudying. By contrast, with a long retention interval, this pattern

was reversed and performance was better after spaced restudying than massed restudying.

APPLIED COGNITIVE PSYCHOLOGYAppl. Cognit. Psychol. 22: 685–695 (2008)Published online 10 August 2007 in Wiley InterScience(www.interscience.wiley.com) DOI: 10.1002/acp.1388

*Correspondence to: Peter P. J. L. Verkoeijen, Department of Psychology, Erasmus University Rotterdam, P.O.Box 1738, NL-3000 DR Rotterdam, The Netherlands. E-mail: verkoeijen@fsw.eur.nl

Copyright # 2007 John Wiley & Sons, Ltd.

On the basis of these findings, Rawson and Kintsch (2005) formulated the following advice

regarding the application of the spacing effect: ‘given the argument that students are most

likely to face delayed testing conditions in real educational settings, we would still

recommend distributed study as the most effective rereading schedule for students (p. 79)’.

Although useful, the recommendation of Rawson and Kintsch leaves an important issue

unaddressed: how far should an educator space the first and the second presentation of a

text to obtain a maximal memory performance? Interestingly, the answer to this question

depends on what theoretical approach to the spacing effect is taken.

In the text rereading literature, the spacing effect has often been explained in terms of the

deactivation hypothesis (e.g. Krug et al., 1990; Rawson & Kintsch, 2005). According to

this hypothesis, full processing of a text will only occur on learning trials in which readers’

representation of the text is no longer in working memory. By contrast, when a text

representation is still available in working memory, or activated, readers will employ this

availability to merely skim over the material. Regarding twice presented texts, the

deactivation hypothesis predicts that under massed repetition, the text representation

formed during the first study trial is still activated at the second trial. As a result, the

processing of the second presentation will be rather superficial and this in turn will exert a

negative influence on the memory of the text. However, the activation of the first trial’s text

representation will decrease with the length of the interrepetition interval until at a certain

point the first trial’s representation has been completely deactivated. Regarding memory

for the repeated text, this implies that performance will increase as a function of the

interrepetition interval, eventually reaching asymptote at the deactivation lag (i.e. the

interrepetition interval corresponding with the complete deactivation of the first trial’s text

representation). Thus, the deactivation hypothesis predicts that forgetting between trials

should benefit memory of repeated text presentations. Hence, on the basis of this account,

one would advise educators to use an interrepetition interval that is sufficiently long to

allow for a complete deactivation of the first text presentation. Under such conditions, the

memory performance is expected to reach a maximum level.

However, the results of a study by Toppino, Hara, and Hackman (2002) argue against the

deactivation hypothesis. In their critical second experiment, they asked participants to

study a list of semantically related words for an upcoming memory test. This list contained

once-presented words as well as massed and spaced repetitions. Furthermore, the spacing

of repetitions was varied between participants: in the spacing/short lag condition, the

interrepetition interval entailed four intervening items, whereas in the spacing/long lag

condition, eight items separated the two occurrences. The analysis of the performance on

the free-recall test revealed that memory for spaced-short items surpassed that of massed

items. Yet contrary to what the deactivation account would predict, the beneficial effect of

spacing disappeared completely when relatively long interrepetition intervals were used. In

that case, memory performance did not differ between massed and spaced items.

The findings of Toppino et al. (2002) are consistent with a two-factor model, which

incorporates mechanisms of encoding variability and study-phase retrieval to account for the

spacing effect in free recall (e.g. Greene, 1989; Toppino & Bloom, 2002). The

encoding-variability component of this model dictates that an item is encoded differentially

at its first and second occurrence, thereby adding additional retrieval cues to the repeated

item’s memory trace. However, the study-phase retrieval component of the two-factor model

states that encoding variability will enrich a repeated item’s memory trace only if the first

occurrence of a repeated item is retrieved from long-term store at its second occurrence.

Combining these two components leads to the prediction that memory will vary with spacing

Copyright # 2007 John Wiley & Sons, Ltd. Appl. Cognit. Psychol. 22: 685–695 (2008)

DOI: 10.1002/acp

686 P. P. J. L. Verkoeijen et al.

in the following way. On the one hand, encoding variability, and therefore the number of

encoded retrieval cues, increases as a function of interrepetition spacing. On the other hand,

the probability of successfully retrieving a repeated item’s first presentation at its second

occurrence decreases with the length of the interrepetition interval. Initially, the negative

effect of the second process on free-recall performance will be cancelled out by the first

process, giving rise to the spacing effect. But at a certain spacing interval, the balance must

reverse, and the second process must start to outweigh the first. From this spacing interval

onwards, free-recall performance must decline with further interrepetition spacing. As a

consequence, the two-factor model predicts a kind of inverted-U-shaped relationship

between interrepetition spacing and free-recall performance.

On the basis of the above-described two-factor model (e.g. Greene, 1989; Toppino &

Bloom, 2002) the findings from the second experiment of Toppino et al. (2002) can be

readily explained. In the long-lag condition, the spacing interval may have been too wide,

resulting in a study-phase retrieval failure for the majority of the items. By contrast, in the

short-lag condition, study-phase retrieval was probably successful for most items. For that

reason, a spacing effect was observed in the short-lag condition, but not in the long-lag

condition. Additional evidence for the notion that study-phase retrieval is a necessary

condition for the spacing effect to occur has been obtained in several other studies (e.g.

Johnston & Uhl, 1976; Toppino & Bloom, 2002; Verkoeijen, Rikers, & Schmidt, 2004).

The study-phase retrieval assumption in the two-factor model could have important

implications for educational practice. That is, if distributed practice is used to enhance

students’ memory for to-be-mastered materials, successive trials should not be spaced too

far apart since intertrial forgetting will eventually reduce the beneficial effect of distributed

practice. However, the two-factor model has thus far only been tested with relatively

simple stimulus materials (i.e. word lists), and with short retention intervals. Therefore, the

purpose of the present study was to assess the merits of the model in a more ecologically

valid setting. Specifically, we aimed at extending Toppino and colleagues’ (2002) findings

to educationally relevant conditions. Similar to the procedure used by Rawson and Kintsch

(2005), participants studied an expository text twice, either in a massed fashion, or in a

spaced fashion with a 4-day delay (spaced short), or in a spaced fashion with a 3.5-weeks

delay (spaced long). Subsequently, all participants were tested 2 days after the second study

opportunity. The test consisted of a free-recall measure and some short-answer questions

tapping on the memory of the text. Given the results obtained by Toppino and colleagues’

we predicted that the magnitude of the spacing effect would vary as a function of the

interrepetition spacing. Specifically, with a relatively short interrepetition lag, a spacing

effect ought to emerge, whereas with a relatively long interrepetition lag, the spacing effect

should be smaller or even absent.

METHOD

Participants

Participants were 61 psychology undergraduates from the Erasmus University Rotterdam,

the Netherlands (mean age¼ 20.39 years, SD¼ 3.73; 71% female) who took part in the

experiment to fulfil a course requirement. Participants were randomly assigned to the three

repetition conditions (i.e. massed vs. spaced short vs. spaced long). The three conditions

contained respectively 18, 20 and 23 participants.

Copyright # 2007 John Wiley & Sons, Ltd. Appl. Cognit. Psychol. 22: 685–695 (2008)

DOI: 10.1002/acp

Distributed rereading and the spacing effect 687

Materials

The stimulus material consisted of a Dutch translation of the English text used in Rawson

and Kintsch’s (2005) second experiment, which was originally taken from a non-fiction

reading book (Carnes, 1995) about the way Hollywood films represent history. The

experimental text discussed how films portray history inaccurately, explained the reasons

for these inaccuracies and then illustrated these arguments by comparing a particular film

(The Charge of the Light Brigade [Curtiz, 1936]) to the actual historical facts of the

Crimean War on which this film was based. The experimental text was 1755 words in

length with a Flesch reading ease score of 31.6 and a Flesch-Kincaid reading level of 13.9.

The text was divided into 12 sections, each section comprising on average 14 sentences1.

Fifteen short-answer questions were constructed to tap participants memory of

information explicitly stated in the text. These 15 questions were Dutch translations of

questions used by Rawson and Kintsch (2005) in their second experiment. In Appendix A,

a few examples of the original English versions of the questions as well as the Dutch

translations are listed.

Procedure

At the start of the experiment, participants were given a sheet on which they were asked to

indicate what they knew about the Crimean War and if they had ever seen the movie The

Charge of the Light Brigade. The analysis of participants’ responses revealed that almost

no participant knew anything about the Crimean War and that none of them had seen the

movie. Therefore, the results of this prior knowledge assessment will not be discussed any

further in the remainder of the article. Subsequently, all participants were instructed to

study a text twice about the way in which Hollywood portrays history. Participants in the

massed-repetition condition were told that they would receive the second study trial

immediately after the first. Participants in the two spaced-repetition conditions were

informed that they would receive the second study trial respectively after 4 days (spaced

short) or after 3.5 weeks (spaced long). Also, all participants were told that they would be

tested on their memory and comprehension of the text 2 days after the second study trial.

Next, all participants read the text once. The 12 sections of the text were presented one

by one on a computer screen in clearly readable black print. Participants could advance to

the next section by pushing the space bar; when they had advanced to a next section of the

text, they were not permitted to return to the previous section. After reading the text once,

participants in the massed condition immediately reread the text in the same manner as in

the first trial. In the two spaced conditions, participants were dismissed after the first study

trial. They were instructed to return either 4 days later or 3.5 weeks later for the second

study trial, at which time they reread the text in the same way as in the first trial. During

both study trials, the total time required to read the entire experimental text was measured

for each participant. By doing so, we could rule out alternative time-on-task explanations

of any spacing effect differences between the short-lag condition and the long-lag

condition that might emerge in memory performance. All participants returned 2 days after

the second study trial to take the test.

The test phase started with a free-recall task that required participants to write down on a

blank sheet everything they could remember from the text they had studied. There were no

1The complete experimental text can be obtained from the first author upon request (verkoeijen@fsw.eur.nl).

Copyright # 2007 John Wiley & Sons, Ltd. Appl. Cognit. Psychol. 22: 685–695 (2008)

DOI: 10.1002/acp

688 P. P. J. L. Verkoeijen et al.

time restrictions for this free-recall test. After participants had indicated they had

completed the free-recall test, the experimenter provided them with a sheet containing the

15 short-answer questions.

RESULTS

To score the free-recall protocols, we used a procedure similar to that used by Krug et al.

(1990). That is, the experimental text was segmented into 79 phrases, each phrase

representing 1, 2 or 4 idea units. For example, the Dutch phrase ‘Deze conflicten omvatten

de herbewapening van Nazi-Duitsland, de agressieve houding van de Sovjet-Unie, een

invasie in de derde wereld door het fascistische Italie, en Japanse aanvallen op Chinees

grondgebied’ [The conflicts included German remilitarisation, aggressive posturing by the

Soviets, third-world invasions by fascist Italy and Japanese advances in China] consists of

four idea units–the conflicts included (1) German remilitarisation, (2) aggressive posturing

by the Soviets, (3) third-world invasions by fascist Italy, and (4) Japanese advances in

China. Alternatively, the Dutch phrase ‘Helaas is de meeste ‘‘historie’’ die weergegeven

wordt in films echter fictie en geen werkelijkheid’ [Unfortunately, much of the ‘history’ that

is portrayed on screen is fiction, not fact] represents only one idea unit. The free-recall for

each participant was calculated by comparing the participant’s free-recall protocol with the

79 target phrases. When a target phrase was entirely and correctly reproduced (verbatim or

paraphrase) a full credit of 1 point was given, whereas an entirely incorrect answer resulted

in 0 points. In case a target phrase contained either two or four idea units, partial credits

were awarded when the free-recall protocol contained some but not all of the correct

information. For target phrase containing two idea units, partial credit implied that a

participant received 0.5 points. By contrast, for target phrases containing four idea units,

partial credit was 0.25, 0.5, or 0.75 points. For each participant, the points in the free-recall

protocol were summed. Subsequently, the overall free-recall score was expressed in terms

of a proportion of the maximum number of points (i.e. 79). All free-recall protocols were

scored blind with respect to the experimental group. Two independent graders scored

approximately 16% of the protocols. The interrater reliability calculated on the pairs of

scores in this subset was high (r¼ 0.97) and therefore one of the two graders rated the

remaining protocols.

Responses to the 15 short-answer questions were assigned a score of 0 (no answer or

entirely incorrect answer) to 1 (complete and correct answer), with partial credits given

when some but not all of the correct information was provided. Scores on this short-answer

test are reported in terms of the proportion of the total possible points, i.e. 15. The

reliability analysis on the scores of the short-answer question revealed a Cronbach’s alpha

coefficient of 0.72.

Although not of primary interest in the present study, reading times are presented in

Appendix B as a function of repetition type and study trial.

Recall performance

Figure 1 depicts the mean overall free-recall performance as a function of repetition type.

Tabachnik and Fidell (2001) indicate that in case of planned comparisons an omnibus

ANOVA (analysis of variance) should not be performed. Instead, the planned comparisons

should directly be conducted. Because we were interested in two specific contracts

Copyright # 2007 John Wiley & Sons, Ltd. Appl. Cognit. Psychol. 22: 685–695 (2008)

DOI: 10.1002/acp

Distributed rereading and the spacing effect 689

(i.e. massed vs. spaced short and massed vs. spaced long), two one-way ANOVA’s were

performed on the free-recall scores. The criterion for statistical significance was set at

p¼ 0.05 for these and all subsequent analyses reported in this article. Furthermore, in each

analysis, the effect size (h2) was obtained. The following guidelines provided by Cohen

(1988) were used to interpret these effect sizes: small (0.01� h2� 0.06), medium

(0.06< h2� 0.14) and large (h2> 0.14).

The first analysis demonstrated a large spacing effect in the overall free-recall of the text

F(1, 36)¼ 8.60,MSE¼ 0.01, p< 0.01, h2¼ 0.19, indicating that the text was remembered

better when study trials were separated by a short lag than when they were presented

consecutively (massed repetition). However, the second analysis revealed, consistent with

our predictions, that the spacing effect in free-recall was much smaller when the massed

condition was compared with the spaced-long condition. In fact, overall free-recall did not

differ significantly between these two conditions F(1, 39)¼ 2.22, MSE¼ 0.01, p> 0.10,

h2¼ 0.05.

A reviewer suggested that it would be informative to report the inferential statistics

comparing the free-recall performance in the short-lag condition and in the long-lag

condition. Although this comparison was not of primary interest in this paper, we

calculated the requested ANOVA. This analysis did not show a reliable difference between

the free-recall performance in the two conditions F(1, 41)¼ 1.45, MSE¼ 0.01, p> 0.10,

h2¼ 0.03.

Short-answer questions

Figure 2 represents the mean performance on the short-answer questions as a function of

repetition type. The results on these scores revealed a pattern comparable to that observed

Figure 1. Mean proportion of accurate overall free-recall as a function of repetition type. Barsrepresent 95% confidence interval of the mean

Copyright # 2007 John Wiley & Sons, Ltd. Appl. Cognit. Psychol. 22: 685–695 (2008)

DOI: 10.1002/acp

690 P. P. J. L. Verkoeijen et al.

in the free-recall scores. That is, performance was better in the spaced-short condition than

in the massed condition F(1, 36)¼ 4.29, MSE¼ 0.03, p< 0.05, h2¼ 0.11, whereas

performance did not differ between the massed condition and the spaced-long condition

F< 1, MSE¼ 0.03, p> 0.50, h2¼ 0.005. Further, an additional ANOVA, which was

performed in response to a reviewer’s request, failed to uncover a reliable difference

between the performance in the short-lag condition and in the long-lag condition F< 1,

MSE¼ 0.06, p> 0.50, h2¼ 0.02.

DISCUSSION

The findings of the present study suggest that the beneficial effect of distributed rereading

depends on the length of the interval separating the two occurrences of spaced repetitions.

More specifically, a relatively short lag between spaced repetitions improved memory over

massed repetitions, whereas a relatively long lag failed to produce a statistically significant

spacing effect. In addition, and in line with our predictions, the magnitude of the spacing

effect was considerably smaller for spaced-long repetitions than for spaced-short

repetitions. To the best of our knowledge, the present study is the first to demonstrate under

educationally relevant conditions that a shorter interrepetition lag is more effective than a

longer interrepetition lag. This observation has both important theoretical and practical

implications.

Theoretical implications

As mentioned in the introduction of this article, researchers have frequently proposed (e.g.

Krug et al., 1990; Rawson &Kintsch, 2005) that the spacing effect should be attributed to a

specific kind of deficient processing mechanism. According to this so-called deactivation

hypothesis, a text representation is formed during the first study trial, and it remains active

spaced-longspaced-shortmassed

Repetition type

0,00

0,20

0,40

0,60

Pr

(co

rrec

tsh

ort

answ

ers)

Figure 2. Mean proportion of correct short-answer questions as a function of repetition type. Barsrepresent 95% confidence interval of the mean

Copyright # 2007 John Wiley & Sons, Ltd. Appl. Cognit. Psychol. 22: 685–695 (2008)

DOI: 10.1002/acp

Distributed rereading and the spacing effect 691

for a short period. Under massed, but not under spaced repetition, the second study trial will

take place while the text representation of the first trial is still elevated. Consequently, less

processing will be directed at the second occurrence of a massed repetition than at the

second occurrence of a spaced repetition. Therefore, the total amount of processing time

allocated to repeated text presentations will be less extensive for massed repetitions than

for spaced repetitions. With respect to memory for the repeated text, this implies that

performance will increase as a function of the interrepetition interval, eventually reaching

asymptote. However, the outcomes of the present study contradict the deactivation

hypothesis by demonstrating that, instead of reaching asymptote, the magnitude of the

spacing effect actually decays with further interrepetition spacing.

A two-factor model, which incorporates mechanisms of encoding variability and

study-phase retrieval to account for the spacing effect (e.g. Greene, 1989; Toppino &

Bloom, 2002) seems to be more readily reconciled with our findings. The encoding-

variability component of this model states that an item is encoded differentially at its

first and second occurrence, adding additional retrieval cues to the repeated item’s

memory trace. However, an essential second assumption model is that encoding

variability will only enrich a repeated item’s memory trace if the first occurrence of a

repeated item is retrieved from long-term store at its second occurrence, that is, if

study-phase retrieval takes place. When the interval between successive study trials

lengthens, encoding variability will be enhanced, but at the same time the probability of

study-phase retrieval will decrease. Initially, the first process will dominate, leading to a

spacing effect in memory performance. However, at a certain spacing interval, the second

process will start to outweigh the beneficial effect of encoding variability. From this

spacing interval onwards, free-recall performance must decline with further spacing. Thus,

the two-factor model predicts a kind of inverted-U-shaped relationship between

interrepetition spacing and memory performance. The finding in the present study that

a longer interrepetition lag is less effective than a shorter interreptition lag is consistent

with this prediction.

The above-presented two-factor model can not only explain the current findings, but it

can also accommodate a range of other findings reported in the spacing-effect literature.

Raaijmakers (2003) provided a mathematical version of this model on the basis of the

search of associative memory (SAM) theory of memory (Raaijmakers & Shiffrin, 1981),

and he showed that the SAM model fitted well to the majority of the standard data on the

spacing effect. Conversely, the SAMmodel could not account for an important observation

reported by Glenberg (1976, Experiment 1) that the optimal spacing level increases as a

function of the length of the retention interval. Furthermore, it is not clear how the

two-factor model relates to learning situations in which participants receive feedback about

the effectiveness of used encoding strategies (e.g. Bahrick & Hall, 2005, Experiment 1; see

also Bahrick, 1979).

In Bahrick and Hall’s (2005) investigation, participants studied a list of Swahili-English

translation pairs for four training sessions with intertraining intervals of 0 (massed), 1 day

and 14 days. In each training session, study and test trials were alternated in combination

with a drop-out procedure so that correctly recalled pairs were no longer studied or tested

on subsequent trials. In addition, the second, third and fourth training sessions started with

a test on the word pairs that were studied in a previous session. A final long-term retention

test was administered to each participant 14 days after the last training session. The

employment of a test during the training sessions provided students with feedback about

Copyright # 2007 John Wiley & Sons, Ltd. Appl. Cognit. Psychol. 22: 685–695 (2008)

DOI: 10.1002/acp

692 P. P. J. L. Verkoeijen et al.

the effect of their encoding strategies and this is relevant in the light of Bahrick and Hall’s

metacognitive explanation of the spacing effect.

According to Bahrick and Hall’s (2005) approach, encoding strategies differ in terms of

retrieval effectiveness; some facilitate retrieval for only minutes, while others ensure

effective retrieval after weeks or even months. Also, when during a training session a

learner employed an encoding strategy that only produces effect of short duration, this

sub-optimal strategy may be identified and modified on the basis of retrieval failures in the

test phase of the subsequent training session. However, if training sessions are massed,

even short-duration encoding strategies will be effective, and the learner will not replace

this strategy during the restudy phase. In opposition, if training sessions are spread apart,

short-duration strategies will yield many retrieval failures, and these will be replaced with

more fruitful long-duration strategies. Therefore, at a final long-term retention test spacing

will produce a superior performance compared to massing. Thus, the metacognitive

hypothesis proposes that when participants are explicitly made aware of their retrieval

failures during training and when they are given sufficient time to correct their learning (i.e.

during test-study sessions) subsequent long-term retention will be enhanced. Alternatively,

on the basis of the two-factor model of the spacing effect (e.g. Greene, 1989; Toppino &

Bloom, 2002; Raaijmakers, 2003) it is expected that when participants are not explicitly

informed about their retrieval failures during training (i.e. during study-only sessions)

long-term memory retention will suffer. Hence, upcoming research should investigate

whether the relationship between spacing and long-term retention differs between

test-study sessions and study-only sessions.

Practical implications

The results reported in our study are of practical relevance as they indicate that merely

providing students with distributed rereading opportunities does not always guarantee

the most favourable learning outcome. Instead, the present findings suggest that

educators need to identify the length of the interrepetition interval at which the

maximummemory performance is obtained. The next question then is what the length of

the optimal interval should be. The present study does not provide an answer to this

question, but Cepeda, Pashler, and Vul (2005) present a potentially useful guideline. In

their study, participants learned during the first session some obscure facts until they

achieved a 100% correct score at a memory test. Subsequently, these facts were

relearned with feedback after an interrepetition interval of 0 (massed), 1, 7, 21 or 105

days. Following this second study trial, both a free-recall and a recognition test were

administered after 1, 5, 10 or 50 weeks. For both types of test, memory performance

displayed a type of inverted-U-shaped relationship with the optimal interstimulus

interval (ISI) being dependent on the length of the retention interval (RI). However, and

more interesting for the present discussion was the observation that the optimal ISI/RI

ratio always had a value equal to about 0.1. Hence, when teaching material, educators

can optimise study time by using a distributed study interval corresponding with 0.1

times the anticipated retention interval. Having said that, if the goal is for the learner to

remember information a very long period, booster sessions may be needed because a

great deal of forgetting occurs within such time frame. To conclude the discussion, it

should be noted that Cepeda and colleagues (2005) based the optimal ISI/RI ratio on

memory of information with a relatively low level of complexity. Therefore, further

Copyright # 2007 John Wiley & Sons, Ltd. Appl. Cognit. Psychol. 22: 685–695 (2008)

DOI: 10.1002/acp

Distributed rereading and the spacing effect 693

research is required to determine whether this ratio also applies to the rereading of more

complex materials, such as instructional texts.

ACKNOWLEDGEMENTS

This study was supported in part by a TopTalent grant of the Erasmus University Rotterdam

to the first author. We thank Katherine Rawson for providing the text and the short-answer

questions that were used in the present study.

REFERENCES

Bahrick, H. P. (1979). Maintenance of knowledge: Questions about memory we forgot to ask. Journalof Experimental Psychology: General, 108, 296–308.

Bahrick, H. P., & Hall, L. K. (2005). The importance of retrieval failures to long-term retention: Ametacognitive explanation of the spacing effect. Journal of Memory and Language, 52, 566–577.

Carnes, M. C. (1995). Past imperfect: History according to the movies. New York: Holt.Cepeda, N. J., Pashler, H., & Vul, E. (2005). november. Distributed practice and long-term retention:A ratio rule for optimal inter-study interval to retention interval. Poster session presented at thePsychonomic Society Annual Meeting, Toronto, ON.

Challis, B. H. (1993). Spacing effects on cued-memory tests depend on level of processing. Journal ofExperimental Psychology: Learning, Memory, and Cognition, 19, 389–396.

Cohen, J. (1988). Statistical power analysis for the behavioral sciences. Hilllsdale, NJ: ErlbaumPublishers.

Crowder, R. G. (1976). Principles of learning and memory. Hillsdale, NJ: Erlbaum Publishers.Curtiz, M. (Director). (1936). The Charge of the Light Brigade [Motion picture]. United States:Warner Bros.

Dempster, F. (1996). Distributing and managing the conditions of encoding and practice. In E. L.Bjork, & R. A. Bjork (Eds.), Memory (pp. 317–344). San Diego, CA: Academic Press.

Glenberg, A.M. (1976). Monotonic and nonmonotonic lag effects in paired-associate and recognitionmemory paradigms. Journal of Verbal Learning and Verbal Behavior, 15, 1–16.

Glover, J. A., & Corkill, A. J. (1987). Influence of paraphrased repetitions on the spacing effect.Journal of Educational Psychology, 79, 198–199.

Greene, R. L. (1989). Spacing effects in memory: Evidence for a two-process account. Journal ofExperimental Psychology: Learning, Memory, and Cognition, 15, 371–377.

Greene, R. L. (1990). Spacing effects on implicit memory tests. Journal of Experimental Psychology:Learning, Memory, and Cognition, 16, 1004–1011.

Hintzman, D. L. (1974). Theoretical implications of the spacing effect. In R. L. Solso (Ed.), Theoriesin cognitive psychology: The Loyola Symposium (pp. 77–99). Potomac, MD: Erlbaum Publishers.

Hintzman, D. L. (1976). Repetition and memory. In G. H. Bower (Ed.), The psychology of learningand motivation: Advances in research and theory (Vol. 10, pp. 47–91). NewYork: Academic Press.

Johnston, W. A., & Uhl, C. N. (1976). The contributions of encoding effort and variability to thespacing effect on free recall. Journal of Experimental Psychology: Human Learning and Memory,2, 153–160.

Krug, D., Davis, B., & Glover, J. A. (1990). Massed versus distributed repeated reading: A case offorgetting helping recall? Journal of Educational Psychology, 82, 366–371.

Mammarella, N., Russo, R., & Avons, S. E. (2002). Spacing effects in cued-memory tasks forunfamiliar faces and nonwords. Memory and Cognition, 30, 1238–1251.

Raaijmakers, J. G. W. (2003). Spacing and repetition effects in human memory: Application of theSAM model. Cognitive Science, 27, 431–452.

Raaijmakers, J. G. W., & Shiffrin, R. M. (1981). Search of associative memory. PsychologicalReview, 88, 93–134.

Copyright # 2007 John Wiley & Sons, Ltd. Appl. Cognit. Psychol. 22: 685–695 (2008)

DOI: 10.1002/acp

694 P. P. J. L. Verkoeijen et al.

Rawson, K. A., & Kintsch, W. (2005). Rereading effects depend on the time of test. Journal ofEducational Psychology, 97, 70–80.

Tabachnik, B. G., & Fidell, L. S. (2001). Using multivariate statistics (4th ed.). Boston: Allyn &Bacon.

Toppino, T. C., & Bloom, L. C. (2002). The spacing effect, free recall, and two-process theory: Acloser look. Journal of Experimental Psychology: Learning, Memory, and Cognition, 28, 437–444.

Toppino, T. C., Hara, Y., & Hackman, J. (2002). The spacing effect in the free recall of homogenouslists: Present and accounted for. Memory and Cognition, 30, 601–606.

Verkoeijen, P. P. J. L., Rikers, R. M. J. P., & Schmidt, H. G. (2004). Detrimental influence ofcontextual change on spacing effects in free recall. Journal of Experimental Psychology: Learning,Memory, and Cognition, 30, 796–800.

APPENDIX A

Examples of short-answer questions, taken from Rawson and Kintsch (2005). The Dutch

translation is presented in italics.

Explain the two reasons why filmmakers choose to fictionalise history. Geef twee

redenen waarom filmmakers ervoor kiezen om van historie fictie te maken.

What was the political idea that The Charge of the Light Brigade was intended to

promote? Welke politiek ideaal diende de film ‘The charge of the light brigade’ te

promoten?

The text listed particular political conflicts happening at the time The Charge of the Light

Brigadewas made that threatened to start a world war. What were they? In de tekst werden

enkele politieke conflicten aangegeven die speelden toen de film ‘The charge of the light

brigade’ gemaakt werd. Noem de politieke conflicten.

APPENDIX B

Mean Reading Times (minutes) as a Function of Repetition Type and Study Trial. Standard Errors areBetween Brackets.

Repetition type Study trial 1 Study trial 2

Massed 11.86 (0.83) 9.24 (0.77)Spaced short 11.09 (0.78) 10.62 (0.73)Spaced long 10.58 (0.73) 11.17 (0.69)

Note: exploratory ANOVA’s indicated that mean reading time did not differ between the three experimentalconditions at the first study trial F< 1, h2¼ 0.02, and at the second study trial F (2, 58)¼ 1.80, MSE ¼10.77,p¼ 0.17, h2¼ 0.06.

Copyright # 2007 John Wiley & Sons, Ltd. Appl. Cognit. Psychol. 22: 685–695 (2008)

DOI: 10.1002/acp

Distributed rereading and the spacing effect 695