Practice

A study involving 60 undergraduate students confirms the value of even a single instance of retrieval practice in an everyday setting, and also confirms the value of cues for peripheral details, which are forgotten more readily.

In three experiments involving 20 undergraduate students, students were shown foreign or otherwise obscure movie clips that contained scenes of normal everyday events. The 24-second clips from 40 films were shown over a period of about half an hour. After a delay of either several minutes, three days, or seven days, the students were questioned on their memory of the general plot, as well as details such as sounds, colors, gestures, and background details that allow a person to re-experience an event in rich and vivid detail.

In the second experiment, students were given a brief visual cue, such as a simple glimpse of the title and a sliver of a screenshot, on testing. In the third experiment, students recalled the information soon after viewing, in addition to the later test.

Researcher found:

  • Peripheral details were, unsurprisingly, forgotten more quickly, and to a greater degree.
  • But those given cues did better at remembering peripheral details.
  • Cues didn’t significantly affect the memory of more substantial matters.
  • Those who retrieved their memories soon after viewing showed no forgetting of peripheral information.
  • Interestingly, these students still assumed they had forgotten a lot (confirming once again, that we're not great at judging our own memory)!

The finding confirms the value of even a single instance of retrieval practice, even without any delay. Note that memory was tested after a week. For longer recall, additional retrieval practice is likely to be needed — but it's probably fair to say that it's that first instance of retrieval that has the biggest effect. I discuss all this in much greater detail in my book on practice.

It's also worth thinking about this in conjunction with the earlier report that there's a special benefit in recounting the information to another person.

https://www.eurekalert.org/pub_releases/2017-01/bu-wta011717.php

Back when I was young, sleep learning was a popular idea. The idea was that a tape would play while you were asleep, and learning would seep into your brain effortlessly. It was particularly advocated for language learning. Subsequent research, unfortunately, rejected the idea, and gradually it has faded (although not completely). Now a new study may presage a come-back.

In the study, 16 young adults (mean age 21) learned how to ‘play’ two artificially-generated tunes by pressing four keys in time with repeating 12-item sequences of moving circles — the idea being to mimic the sort of sensorimotor integration that occurs when musicians learn to play music. They then took a 90-minute nap. During slow-wave sleep, one of the tunes was repeatedly played to them (20 times over four minutes). After the nap, participants were tested on their ability to play the tunes.

A separate group of 16 students experienced the same events, but without the playing of the tune during sleep. A third group stayed awake, during which 90-minute period they played a demanding working memory task. White noise was played in the background, and the melody was covertly embedded into it.

Consistent with the idea that sleep is particularly helpful for sensorimotor integration, and that reinstating information during sleep produces reactivation of those memories, the sequence ‘practiced’ during slow-wave sleep was remembered better than the unpracticed one. Moreover, the amount of improvement was positively correlated with the proportion of time spent in slow-wave sleep.

Among those who didn’t hear any sounds during sleep, improvement likewise correlated with the proportion of time spent in slow-wave sleep. The level of improvement for this group was intermediate to that of the practiced and unpracticed tunes in the sleep-learning group.

The findings add to growing evidence of the role of slow-wave sleep in memory consolidation. Whether the benefits for this very specific skill extend to other domains (such as language learning) remains to be seen.

However, another recent study carried out a similar procedure with object-location associations. Fifty everyday objects were associated with particular locations on a computer screen, and presented at the same time with characteristic sounds (e.g., a cat with a meow and a kettle with a whistle). The associations were learned to criterion, before participants slept for 2 hours in a MR scanner. During slow-wave sleep, auditory cues related to half the learned associations were played, as well as ‘control’ sounds that had not been played previously. Participants were tested after a short break and a shower.

A difference in brain activity was found for associated sounds and control sounds — associated sounds produced increased activation in the right parahippocampal cortex — demonstrating that even in deep sleep some sort of differential processing was going on. This region overlapped with the area involved in retrieval of the associations during the earlier, end-of-training test. Moreover, when the associated sounds were played during sleep, parahippocampal connectivity with the visual-processing regions increased.

All of this suggests that, indeed, memories are being reactivated during slow-wave sleep.

Additionally, brain activity in certain regions at the time of reactivation (mediotemporal lobe, thalamus, and cerebellum) was associated with better performance on the delayed test. That is, those who had greater activity in these regions when the associated sounds were played during slow-wave sleep remembered the associations best.

The researchers suggest that successful reactivation of memories depends on responses in the thalamus, which if activated feeds forward into the mediotemporal lobe, reinstating the memories and starting the consolidation process. The role of the cerebellum may have to do with the procedural skill component.

The findings are consistent with other research.

All of this is very exciting, but of course this is not a strategy for learning without effort! You still have to do your conscious, attentive learning. But these findings suggest that we can increase our chances of consolidating the material by replaying it during sleep. Of course, there are two practical problems with this: the material needs an auditory component, and you somehow have to replay it at the right time in your sleep cycle.

We know that we remember more 12 hours after learning if we have slept during that 12 hours rather than been awake throughout, but is this because sleep is actively helping us remember, or because being awake makes it harder to remember (because of interference and over-writing from other experiences). A new study aimed to disentangle these effects.

In the study, 207 students were randomly assigned to study 40 related or unrelated word pairs at 9 a.m. or 9 p.m., returning for testing either 30 minutes, 12 hours or 24 hours later.

As expected, at the 12-hour retest, those who had had a night’s sleep (Evening group) remembered more than those who had spent the 12 hours awake (Morning group). But this result was because memory for unrelated word pairs had deteriorated badly during 12 hours of wakefulness; performance on the related pairs was the same for the two groups. Performance on the related and unrelated pairs was the same for those who slept.

For those tested at 24 hours (participants from both groups having received both a full night of sleep and a full day of wakefulness), those in the Evening group (who had slept before experiencing a full day’s wakefulness) remembered significantly more than the Morning group. Specifically, the Evening group showed a very slight improvement over training, while the Morning group showed a pronounced deterioration.

This time, both groups showed a difference for related versus unrelated pairs: the Evening group showed some deterioration for unrelated pairs and a slightly larger improvement for related pairs; the Morning group showed a very small deterioration for related pairs and a much greater one for unrelated pairs. The difference between recall of related pairs and recall of unrelated pairs was, however, about the same for both groups.

In other words, unrelated pairs are just that much harder to learn than related ones (which we already know) — over time, learning them just before sleep vs learning early in the day doesn’t make any difference to that essential truth. But the former strategy will produce better learning for both types of information.

A comparison of the 12-hour and 24-hour results (this is the bit that will help us disentangle the effects of sleep and wakefulness) reveals that twice as much forgetting of unrelated pairs occurred during wakefulness in the first 12 hours, compared to wakefulness in the second 12 hours (after sleep), and 3.4 times more forgetting of related pairs (although this didn’t reach significance, the amount of forgetting being so much smaller).

In other words, sleep appears to slow the rate of forgetting that will occur when you are next awake; it stabilizes and thus protects the memories. But the amount of forgetting that occurred during sleep was the same for both word types, and the same whether that sleep occurred in the first 12 hours or the second.

Participants in the Morning and Evening groups took a similar number of training trials to reach criterion (60% correct), and there was no difference in the time it took to learn unrelated compared to related word pairs.

It’s worth noting that there was no difference between the two groups, or for the type of word pair, at the 30-minutes test either. In other words, your ability to remember something shortly after learning it is not a good guide for whether you have learned it ‘properly’, i.e., as an enduring memory.

The study tells us that the different types of information are differentially affected by wakefulness, that is, perhaps, they are more easily interfered with. This is encouraging, because semantically related information is far more common than unrelated information! But this may well serve as a reminder that integrating new material — making sure it is well understood and embedded into your existing database — is vital for effective learning.

The findings also confirm earlier evidence that running through any information (or skills) you want to learn just before going to bed is a good idea — and this is especially true if you are trying to learn information that is more arbitrary or less well understood (i.e., the sort of information for which you are likely to use mnemonic strategies, or, horror of horrors, rote repetition).

Students come into classrooms filled with inaccurate knowledge they are confident is correct, and overcoming these misconceptions is notoriously difficult. In recent years, research has shown that such false knowledge can be corrected with feedback. The hypercorrection effect, as it has been termed, expresses the finding that when students are more confident of a wrong answer, they are more likely to remember the right answer if corrected.

This is somewhat against intuition and experience, which would suggest that it is harder to correct more confidently held misconceptions.

A new study tells us how to reconcile experimental evidence and belief: false knowledge is more likely to be corrected in the short-term, but also more likely to return once the correction is forgotten.

In the study, 50 undergraduate students were tested on basic science facts. After rating their confidence in each answer, they were told the correct answer. Half the students were then retested almost immediately (after a 6 minute filler task), while the other half were retested a week later.

There were 120 questions in the test. Examples include: What is stored in a camel's hump? How many chromosomes do humans have? What is the driest area on Earth? The average percentage of correct responses on the initial test was 38%, and as expected, for the second test, performance was significantly better on the immediate compared to the delayed (90% vs 71%).

Students who were retested immediately gave the correct answer on 86% of their previous errors, and they were more likely to correct their high-confidence errors than those made with little confidence (the hypercorrection effect). Those retested a week later also showed the hypercorrection effect, albeit at a much lower level: they only corrected 56% of their previous errors. (More precisely, on the immediate test, corrected answers rose from 79% for the lowest confidence level to 92% for the highest confidence. On the delayed test, corrected answers rose from 43% to 70% on the second highest confidence level, 64% for the highest.)

In those instances where students had forgotten the correct answer, they were much more likely to reproduce the original error if their confidence had been high. Indeed, on the immediate test, the same error was rarely repeated, regardless of confidence level (the proportion of repeated errors hovered at 3-4% pretty much across the board). On the delayed test, on the other hand, there was a linear increase, with repeated errors steadily increasing from 14% to 23% as confidence level rose (with the same odd exception — at the second highest confidence level, proportion of repeated errors suddenly fell).

Overall, students were more likely to correct their errors if they remembered their error than if they didn’t (72% vs 65%). Unsurprisingly, those in the immediate group were much more likely to remember their initial errors than those in the delayed group (85% vs 61%).

In other words, it’s all about relative strength of the memories. While high-confidence errors are more likely to be corrected if the correct answer is readily accessible, they are also more likely to be repeated once the correct answer becomes less accessible. The trick to replacing false knowledge, then, is to improve the strength of the correct information.

Thus, as recency fades, you need to engage frequency to make the new memory stronger. So the finding points to the special need for multiple repetition, if you are hoping to correct entrenched false knowledge. The success of immediate testing indicates that properly spaced retrieval practice is probably the best way of replacing incorrect knowledge.

Of course, these findings apply well beyond the classroom!

In the first study, undergraduates studied English-Lithuanian word pairs, which were displayed on a screen one by one for 10 seconds. After studying the list, the students practiced retrieving the English words — they had 8 seconds to type in the English word as each Lithuanian word appeared, and those that were correct went to the end of the list to be asked again, and those wrong had to be restudied. Each item was pre-assigned a "criterion level" from one to five — the number of times it needed to be correctly recalled during practice.

In the first experiment, participants took one of four recall tests and one of three recognition tests after a 2-day delay. In the second experiment, in order to eliminate the reminder effect of the recall test, participants were only given a recognition test, after a 1-week delay.

Both experiments found that higher criterion levels led to better memory. More importantly, through the variety of tests, they showed that this occurred on all three kinds of memory tested: associative memory; target memory; cue memory. That is, practicing retrieval of the English word didn’t just improve memory for that word (the target), but also for the Lithuanian word (the cue), and the pairing (association).

While this may seem self-evident to some, it has been thought that only the information being retrieved is strengthened by retrieval practice. The results also emphasize that it is the correct retrieval of the information that improves memory, not the number of times the information is studied.

In a related study, 533 students learned conceptual material via retrieval practice across three experiments. Criterion levels varied from one to four correct retrievals in the initial session. Items also varied in how many subsequent sessions they were exposed to. In one to five testing/relearning sessions, the items were practiced until they were correctly recalled once. Memory was tested one to four months later.

It was found that the number of times items were correctly retrieved on the initial session had a strong initial effect, but this weakened as relearning increased. Relearning had pronounced effects on long-term retention with a relatively minimal cost in terms of additional practice trials.

On the basis of their findings, the researchers recommend that students practice recalling concepts to an initial criterion of three correct recalls and then relearn them three times at widely spaced intervals.

[2457] Vaughn KE, Rawson KA. Diagnosing Criterion-Level Effects on Memory. Psychological Science [Internet]. 2011 . Available from: http://pss.sagepub.com/content/early/2011/08/02/0956797611417724.abstract

Rawson, K.A. & Dunlosky, J. 2011. Optimizing schedules of retrieval practice for durable and efficient learning: How much is enough? Journal of Experimental Psychology: General, Jun 27, 2011, No Pagination Specified. doi: 10.1037/a0023956

I’ve talked about the importance of retrieval practice at length, so I’m pleased to report on the latest study to confirm its value. Indeed, this study demonstrates that practicing retrieval is a more effective strategy than elaborative studying.

In two studies, a total of 200 students studied texts on topics from different science disciplines. One group engaged in elaborative studying by creating concept maps. The second group read the texts, then put the material away and practiced recalling the concepts from the text. Both groups performed at about the same level on a test at the end of the study period. However, when the students were tested again a week later, the group that studied by practicing retrieval performed 50% better than the group that studied by creating concept maps.

The test involved understanding as well as memory, with some of the questions asking them to draw connections between things that weren't explicitly stated in the material.

The study also confirms that most students are poor at judging the success of their study habits. Asked to predict which technique would produce better results, most thought that concept mapping would be superior.

The findings should certainly not be taken as a slur on concept mapping, which is a study strategy of proven effectiveness. Moreover, while concept mapping can be used solely as an elaborative study method (as it was in these experiments), it can also be used as a retrieval practice technique.

In an experiment to investigate why testing might improve learning, 118 students were given 48 English-Swahili translation pairs. An initial study trialwas followed by three blocks of practice trials. For one group, the practice trial involved a cued recall test followed by restudy. For the other group, they weren’t tested, but were simply presented with the information again (restudy-only). On both study and restudy trials, participants created keywords to help them remember the association. Presumably the 48 word pairs were chosen to make this relatively easy (the example given in the paper is the easy one of wingu-cloud). A final test was given one week later. In this final test, participants received either the cue only (e.g. wingu), or the cue plus keyword, or the cue plus a prompt to remember their keyword.

The group that were tested on their practice trials performed almost three times better on the final test compared to those given restudy only (providing more evidence for the thesis that testing improves learning). Supporting the hypothesis that this has to do with having more effective keywords, keywords were remembered on the cue+prompt trials more often for the test-restudy group than the restudy-only group (51% vs 34%). Moreover, providing the keywords on the final test significantly improved recall for the restudy-only group, but not the test-restudy group (the implication being that they didn’t need the help of having the keywords provided).

The researchers suggest that practice tests lead learners to develop better keywords, both by increasing the strength of the keywords and by encouraging people to change keywords that aren’t working well.

A new study explains why variable practice improves your memory of most skills better than practice focused on a single task. The study compared skill learning between those asked to practice one particular challenging arm movement, and those who practiced the movement with other related tasks in a variable practice structure. Using magnetic stimulation applied to different parts of the brain after training (which interferes with memory consolidation), it was found that interference to the dorsolateral prefrontal cortex, but not to the primary motor cortex, affected skill learning for those engaged in variable practice, whereas interference to the motor cortex, but not to the prefrontal cortex, affected learning in those engaged in constant practice.

These findings indicate that variable practice involves working memory (which happens in the prefrontal cortex) rather than motor memory, and that the need to re-engage with the task each time underlies the better learning produced by variable practice (which involves repeatedly switching between tasks). The experiment also helps set a time frame for this consolidation — interference four hours after training had no effect.

A study involving 117 six year old children and 104 eight year old children has found that the ability to preserve information in working memory begins at a much younger age than had previously been thought. Moreover the study revealed that, while any distraction between learning the words and having to recall them hindered recall, having to perform a verbal task was particularly damaging. This suggests that their remembering was based on “phonological rehearsal”, that is, verbally repeating the names of the items to themselves. Consistent with the research suggesting children begin to phonologically rehearse at around 7 years of age, the verbal task hindered the 8 year olds more than the 6 year olds.

Loss of memory and problems with judgment in dementia patients can cause difficulties in relation to eating and nutrition; these problems in turn can lead to poor quality of life, pressure ulcers and infections. A study used two different step-by-step training programs to help dementia patients regain eating skills. Three institutions, involving 85 patients, were assigned to one of three programs: spaced retrieval training; Montessori-based training; control. Training consisted of three 30-40 min sessions per week, for 8 weeks. Both training programs resulted in significantly improved feeding skills, however the Montessori group needed more physical and verbal assistance. Nutritional status was significantly higher in the spaced-retrieval group compared to the control.

Lin, L., Huang, Y., Su, S., Watson, R., Tsai, B. W., & Wu, S. (2010). Using spaced retrieval and Montessori-based activities in improving eating ability for residents with dementia. International Journal of Geriatric Psychiatry, 9999(9999), n/a. doi: 10.1002/gps.2433.

It has long been known that spacing practice (reviewing learning or practicing a skill at spaced intervals) is far more effective than massed practice (in one heavy session). It is also well-known that people commonly over-estimate the value of massed practice, and tend not to give due recognition to the value of spacing practice, despite the fact that most memory improvement and study programs advise it.

Many learning strategies require extensive training. The advantage of spaced practice is that it does not. Experience with it may also result in better self-appraisal about how well information has been learned.

In this study, a class of 708 students were given instruction sheets on which was written a 7-digit number purporting to be a phone number. The students were instructed to memorize the number and told their recall would be tested later in the term. Half the class were told to memorize the number however they usually would. The other half were told to post the number where they would see it, and look at it once or twice a day for a week. They were told this would be an effective way of learning the number.

A significantly greater number of students from the spaced-practice group remembered the number correctly two weeks later (72.7% compared to 61% of the control group). According to the questions they answered, some 11.6% of the spaced-practice group in fact did all their studying in a single session, and only 46.4% studied the number on 3 or more days. Some 18% of the control group also studied the number on 3 or more days. In other words, being in the spaced-practice group doesn't necessarily mean spaced practice was used, nor does being a member of the control group mean that spaced-practice wasn't used.

Clearly, simple instructions to use spaced practice improve memory, but equally clearly, many people are not necessarily going to follow those instructions, for whatever reason.

Landauer, T.K. & Ross, B.H. (1977). Can simple instructions to use spaced practice improve ability to remember a fact? An experimental test using telephone numbers. Bulletin of the Psychonomic Society, 10, 215-218.

Older news items (pre-2010) brought over from the old website