Metacognitive Control Over Distribution of Practice: When is Spacing Preferred?
Michael S. Cohen, Meghan Davis, Amy C. Moors, Thomas C. Toppino
Department of Psychology, Villanova University
Experiment 1
Introduction
Memory is generally better when study opportunities are distributed in time (spaced practice) rather
than when they occur in immediate succession (massed practice). However, when learners decide
practice schedules for themselves, do they choose massed practice or spaced practice?
• 34 participants
• Pairs of common words
• Duration of first presentation: 1s vs. 5s
Results - Proportion of restudied pairs selected for spaced practice.
Background
• Significantly higher for 5s than 1 s presentation duration
• No significant effect of difficulty or its interaction with duration.
Son (2004) presented pairs of GRE vocabulary words and common synonyms for 1 s each. After each
pair, learners judged how well they knew it (indexing difficulty) and chose to study it again for 3 s,
either now (massed), later (spaced) or never (done). For items chosen for restudy (massed or spaced),
Son found a negative correlation such that the preference for spaced practice declined as judged
difficulty increased.
The apparent trend for the proportion of spacing choices to increase
with difficulty in the 1s condition was reminiscent of Benjamin and
Bird’s (2006) findings, but was not reliable. The possibility of a similar
trend in the 5s condition was obscured by a ceiling effect.
Son’s results are consistent with the Region of Proximal Learning Model (Metcalfe & Kornell, 2005)
which predicts that, for items judged to be learnable, the perceived benefits of immediate study
(massed practice) should be greater for more difficult items.
The present study
Results - Proportion of restudied pairs selected for spaced practice.
Procedure. Following Son’s (2004) procedure, word pairs were presented on a computer screen to
college-student participants who then were given a choice between studying the pair again “now”
(immediately), studying it again “later” (after all word pairs were shown at least once), or being “done”
with it (not presented again). (See diagram.)
Now?
Later?
Done?
Cue word
---------Target word
x
Cue word
---------Target word
Duration of the second presentation of each re-studied pair was held constant at 3s. After completing
all study trials, participants solved arithmetic problems for 5 min. This was followed by a recall test in
which the cue words were presented successively, and participants attempted to type the corresponding
target word.
Design. Mixed factorial with Pair Difficulty (easy, medium, hard) varied within participants and
Duration of the Pairs’ Initial Presentation varied between participants.
Materials. Pairs of common words (Exp. 1) or GRE-synonym pairs (Exps. 2 & 3). Item difficulty
was determined a priori and manipulated experimentally.
Primary dependent variable. Proportion of re-studied items chosen for later, spaced
presentation (Proportion Spaced).
The 1s condition replicates Son (2004) and Exp. 2.
How many pairs did participants see?
• In the 1s condition, the number of items participants saw
declines significantly as a function of increasing difficulty.
• Although a similar trend is evident at the slower
presentation durations, the magnitude of the effect is much
smaller.
• 91 participants
• Son’s (2004) GRE-synonym pairs
• Duration of first presentation: 1s vs. 5s
• Greater for 5s than 1s presentation duration, but duration interacted
with difficulty.
• Declined significantly with increasing difficulty in the 1s initial
presentation condition (p = .035), replicating Son (2004).
• Increased significantly with increasing difficulty in the 5s initial
presentation condition (p = .002), reminiscent of Benjamin & Bird
(2006).
Son’s (2004) effect of pair difficulty was replicated with a 1s presentation, but the opposite
effect of difficulty was obtained with a 5s presentation.
General Method — All Experiments
Proportion of restudied pairs selected for spaced practice
• Higher in the 2s and 3s conditions than in the 1s condition
• Declines significantly with increasing difficulty in the 1s
condition (p = .026), but not in the 2s and 3s conditions.
Experiment 2
Method
We sought to resolve the conflict from these two earlier studies. Our procedure was similar to Son’s.
However, within experiments, we varied item difficulty experimentally (like Benjamin and Bird) rather
than correlationally (like Son), and we varied the duration of the initial presentation of each pair.
Across experiments, we varied the type of item: Pairs of common words (like Benjamin and Bird) in
Exp. 1, and GRE-synonym pairs (like Son) in Exps. 2 and 3.
All Re-studied Pairs (Seen and Unseen)
Son’s (2004) effect of pair difficulty was not replicated.
However, Benjamin and Bird (2006), using pairs of common words with related, but different,
methods, found a stronger preference for spaced practice on more difficult items. They interpreted their
results in terms of the Discrepancy Reduction Model, which predicts a constant or increasing
preference for the more effective strategy (spaced practice) as perceived difficulty increases.
Cue word
---------Target word
Experiment 3 - Results
Method
The Question
Summary of Experiments 1 and 2
Son’s (2004) findings (declining choice of spacing as a function of increasing pair difficulty) were replicated only
when we used conditions identical to hers (GRE-synonym pairs presented for a very brief 1s duration).
Using pairs of common words or a long (5s) presentation duration produced either no effect of pair difficulty or a
tendency for the choice of spaced practice to increase with increasing difficulty. Overall, spacing was chosen
more often when the presentation duration was long (5s) rather than short (1s).
Experiment 3
Question -- Why are Sonʼs (2004) results replicable only with GRE-synonym pairs and a very fast
presentation?
Hypothesis - A 1s presentation duration strains participantsʼ ability to perceive and initially encode
a pair. As GRE-synonym pairs (but not pairs of common words) increase in difficulty, the cue word
tends to become longer and less familiar. Thus, increasing difficulty in GRE-synonym pairs would
exacerbate any problems participants have in establishing initial encodings of the pairs.
How to test the hypothesis - Ask participants whether they saw the words.
Secondary Results — Common to All Experiments
Method
Proportion of “done” responses (pairs not selected for re-study).
•  Decreased with increasing difficulty in all experiments
•  Did not vary as a function of the duration of initial pair presentation.
This confirms that we successfully manipulated pair difficulty and suggests that participants used the
“done” response primarily for pairs that they felt they knew from the outset.
Recall on final test. Spaced pairs had higher recall than massed pairs; thus, spacing was probably the
superior strategy
• 
• 
• 
• 
79 participants
Son’s (2004) GRE-synonym pairs
Duration of the first presentation: 1s vs. 2s vs. 3s
The procedure was identical to the previous experiments with one critical exception.
Immediately after participants decided whether a pair should be massed, spaced, or done, a screen appeared asking
them whether they had seen both words in the pair. They responded “yes” or “no” with a keystroke.
In the 1s condition, the likelihood of seeing pairs declines
as difficulty increases. This parallels the tendency to
choose spaced practice less (and massed practice more)
as difficulty increases.
Seen Pairs Only
Proportion of restudied pairs selected for spaced practice
• No significant decline in the likelihood of choosing spaced
practice with increasing difficulty
• Effect of initial presentation duration is also reduced and is not
significant.
When results were restricted to pairs that were seen, Son’s
(2004) results were not replicated. The effect of difficulty
on participants’ tendency to choose spaced practice was
largely eliminated. The residual trend was not statistically
reliable.
Discussion
Son (2004) reported that as item difficulty increased, learners became less likely to choose spaced practice rather than massed
practice. She attributed this to metacognitive judgments in which the perceived benefit of massed practice is assumed to increase
with decreases in how well the pairs are judged to be learned. Her account was consistent with the Region of Proximal Learning
Model (Metcalfe & Kornell, 2005). However, our findings suggest that Son’s (2004) results were largely an artifact of conditions
that prevented learners from fully encoding some items, especially the most difficult ones, on the initial study opportunity. When
unseen items were excluded from the analysis, the effect of difficulty on choice of spaced practice was largely eliminated. It was
similarly eliminated when pairs involved common words (Exp. 1) in which length and familiarity did not vary as a function of
difficulty.
It appears that if learners are unable to fully encode a pair initially, they will want to see it again immediately. However, when
this is not the case, learners generally seem to prefer spaced practice. When they do show a preference based on difficulty, it
appears to be a preference to space the more difficult items. These findings are in accordance with both Benjamin and Bird’s
(2006) results and the Discrepancy Reduction Model.
Further work is necessary to determine whether the preference we observed for spaced practice actually reflects a metacognitive
appreciation of its benefits. An alternate possibility is that learners who opt for spaced practice believe their choice will allow
them to re-study the pairs closer to the time of the test. Thus, learners’ choices could reflect metacognitive judgments about the
effect of the length of the retention interval rather than about the effect of spaced practice. This possibility is being tested.
References
Benjamin, A.S., & Bird, R.D. (2006). Metacognitive control of the spacing of study repetitions.
Journal of Memory and Language, 55, 126-137.
Metcalfe, J., & Kornell, N. (2005). A Region of Proximal Learning model of study time allocation.
Journal of Memory and Language, 52, 463-477.
Son, L.K. (2004). Spacing one’s study: Evidence for a metacognitive control strategy. JEP:
Learning, Memory, and Cognition, 30, 601-604.
Author Notes
•  We thank Lisa Son for providing us with helpful information, including a copy of
the items that she used in her 2004 article.
•  For further information, contact [email protected] or
[email protected].
•  A PDF version of this poster is available at:
http://www.sccs.swarthmore.edu/~mcohen/Metacognition_and_Spacing_APS.pdf