Making long-term memories in minutes – too good to be true?

Twitter posts can lead in all sorts of intriguing directions. It’s a cornucopia of endless nourishment for the educational soul but you do have to taste with caution. A few days ago, this tweet caught my eye.

The idea of spaced learning has been moderately prominent for a while now. I can’t remember where I first came across it but Pashler et al (2007) p.5 is probably the most accessible source and this summary comes from the What Works Clearing House so the underpinning research will be pretty robust.

Typical spaced learning is based on the concept of the Ebbinghaus forgetting curve (Ebbinghaus 1885) and this Class Teaching blog post is a nice summary of sources that you might have come across, and how the concept might be applied in the classroom.

So far, so good, but the Tweet was about something I hadn’t come across before. This approach is termed Spaced Learning (with capitals). Rory Gallagher @eddiekayshun ran a session at #ililc5 that seems to have really inspired teachers who participated. Alex Bellars @bellaale blogged about this. It’s a very clear post that describes the session in detail. Rory’s conference session will, of course, have been adapted for the situation and I don’t think went through the full three learning cycles so the clearest description is this guide to Spaced Learning at Monkseaton High School produced by Learning Futures.

The key idea is that factual knowledge is delivered at pace and then repeated (with more challenge to thinking and perhaps the use of association techniques) after a break of about ten minutes, then re-visited a third time after another, similar break. The whole three cycles could potentially fit into a normal school lesson. The suggestion is that this approach is particularly effective at long-term memory encoding i.e. it’s a particularly efficient way of learning factual knowledge.

Initial outcomes from the programme are reported very positively. The Learning Futures guide just contains anecdotal reports from teachers and children but the academic paper written by the creator(s) of the Spaced Learning programme contains research evidence from the Monkseaton High School trial. The paper is by Kelley and Whatson (2013) and provides a detailed neuroscientific background, references to supporting education and psychology research, and describes the research carried out during the trial at Monkseaton High School. This seems to have carried some weight since the EEF and Wellcome Trust are jointly funding a pilot study working with 15 schools around the Hallam Teaching School Alliance to develop resources and explore different ways of implementing the programme in schools.

I thought all of this was worth a more detailed look, so have taken the time to read Kelley and Whatson (2013) properly. The paper starts out with references to neuroscience research that show rapid stimuli followed by breaks increases processes, like protein production, in the brain, that lead to long-term memory formation. Further research is cited that shows how experimental work has determined that the timing, particularly of the breaks, makes a big difference. I’m perfectly happy to accept this at face value because it doesn’t seem very important to me. I think that there is a lot of scope for innovative ideas in education to be developed from neuroscience and psychology findings as long as there is no suggestion that, just because there is a lab-based finding behind it, the education approach ‘must’ work. However, the Spaced Learning programme choice of rapid study (by normal school standards) and breaks of 10 minutes, does seem to be largely based on research on honeybees and rats. Forgive me if it sounds trite but given that the lifespan of a honeybee is about 6-7 weeks I do wonder if their brains might function on a different time-scale to a typical teenager. Of course, it’s perfectly possible that, even if the origin of the timing choices is spurious, they could by chance be appropriate. My main misgiving is that everything else I’ve seen about spaced learning (no capitals) suggests that longer intervals are better than short ones.

This hasn’t been ignored; Kelley and Whatson have cited other research on spaced learning quite extensively in their paper. Unfortunately, as far as I can tell, these references to research in psychology or education settings, refer to papers looking at learning spaced out by at least 24 hours. That doesn’t mean it isn’t relevant but my feeling is that the way the paper presents these references sort of implies there might be an element of support in there but actually some of them are quite contradictory.

Carpenter et al (2012) looked at different spacing and concluded that the optimum was when spacing was 10% of the time to the final test, so teaching a Y10 topic it would be 1-2 months spacing, and for Y11 revision it would be 1-4 days maybe. Two other references that are relevant are: Karpicke and Bauernschmidt (2011) – no measurable difference if the gaps between repeats increase in line with the Ebbinghaus forgetting curve; Sobel, Cepeda, Kapler (2011) – one week spacing is better than no spacing. However, the Pashler et al (2007) review is even more specific:

“Although a few inconsistencies have been found, by far the most common finding is that when the time between study sessions is very brief relative to the amount of time to the final test, students do not do as well on the final test” (p.5)

Perhaps slightly suprisingly, the EEF and Wellcome Trust funding for further research into Spaced Learning via The Education and Neuroscience Initiative is based on a EEF literature review which reaches the same conclusion. In specific reference to the Monkseaton High School programme it states:

“However, rather than converging on one optimal amount of time for the spacing, recent work shows the optimal gap increases as the time to test also increases. This study [reference is to two studies, Cepeda (2008), Cepeda (2009)] indicates the penalty for a too-short gap is far greater than the penalty for a too-long gap” (Howard-Jones, 2014 p.29)

I don’t think that any of this is necessarily a terminal sentence for the Spaced Learning programme but it certainly makes me wonder why Kelley and Whatson haven’t been clearer in explaining why the existing, contradictory literature doesn’t apply. It would seem simple to argue that the assumption by other researchers is that ‘spaced’ means over days/weeks/months (in line with the Ebbinghaus forgetting curve) and therefore it hasn’t occurred to anyone to try minutes.

So, anyway, the final question is obviously what does the Monkseaton High School research look like. My first problem has been that my limited intellect struggled a bit to follow the protocols but I think I’ve got this now and I like the experimental design as at least one of the protocols (1) is potentially biaised against Spaced Learning and another is essentially the opposite way round, just in case. What do the results show?

The comparison between a Biology module taught to one class using a single session of Spaced Learning and taught to other classes over four months showed no significant difference. There is then a chart showing that the effect per hour of instruction for Spaced Learning was about 20 times!! higher (I’ll come back to this). Finally, classes that had done a Physics module with a lesson of traditional intensive revision performed in line with the national cohort whereas the same classes outperformed the national cohort in a Biology module taught normally but with a lesson of Spaced Learning revision (mean marks of 63% against 55%).

The first finding seems tremendously impressive as it suggests that it’s possible to replace four months of teaching with one hour of Spaced Learning. No Effect Size is reported but that must be well over Hattie’s 0.40 bar (actually it’s probably a bit like Sergei Bubka taking on a slightly high kerb). My only query is that we don’t know what the scores would have been like if this module hadn’t been taught at all – pupils just given a revision guide or something. It was a high-stakes GCSE module exam and there is bound to have been some prior learning as well. Hopeful, though.

The second finding is ridiculous (I said I’d come back to it!) If you teach the same thing in an hour, and in 23 hours, it’s no surprise that you get more progress per hour from the first option. Let’s move on.

The third finding seems more reasonable. The implication is that an hour of Spaced Learning is more effective revision than ‘normal’ revision, maybe by as much as a GCSE grade (depending on the boundaries for that module). On the other hand, I don’t think we know anything about who taught the Physics and Biology modules and it’s important to remember that a really significant effort had gone into preparing the Spaced Learning materials for Biology, by the teachers implementing it.

I’m not trying to dismiss Spaced Learning out of hand but it seems to me that the link from the neuroscience to Spaced Learning is tenuous, the existing research in psychology and education is contradictory, and if the Spaced Learning was implemented by a small group of highly-motivated teachers, excited by the innovation, with pupils anxious to get the best possible grades, more aware of Biology revision than other subjects because of the Spaced Learning, and believing that Spaced Learning might help them to achieve more highly, then some – maybe all – of the effect could be spurious.

But this certainly isn’t a Ben Goldacre – Bad Science post. A tentative, small-scale pilot is exactly what should have come out of the original idea, and this is just what the Monkseaton High School work has been. Excellent! The positive effect might not be real but nor is this a negative outcome. Something on a slightly larger scale is appropriate, with further, independent evaluation. Again, this is what is happening with the Hallam Teaching School Alliance, EEF/Wellcome Trust funded project. If the evidence continues to build up then that would be good news for science teachers, who are in a constant battle with a gargantuan curriculum. It’s so hard to cover the whole thing without moving on from topics before children are ready. What I really want to see is a load of the more random disconnected factual content taken out and far more depth and mastery of the parts that matter in actually understanding science. In the meantime – whilst waiting for the cows to assemble in my back garden – an approach that allowed the making of long-term memories in minutes would be a good compromise. But it does sound too good to be true, doesn’t it?