Cheat codes to intelligence

I’ve asked before why students don’t remember what they’ve learned: how we design instruction, the curriculum and assessment plays a large part.

On first discovering cognitive science, Kris Boulton said it was “like being given the cheat codes to intelligence”. The models of memory and the mind in seminal texts like Dan Willingham’s Why Don’t Students Like School, the models of how we learn and the insights for instruction: this research has a lot to offer teachers.


Touchpaper problem #7 is about retaining content in long-term memory. If cognitive scientists are correct, that ‘if nothing has been retained in long-term memory, nothing has been learned,’ this is a key issue for teachers.

Our team (Helene, Mark, Tim, Ben, Lucy, Jackie and I) tried in the touchpaper working party to turn what we know from cognitive science into what teachers can do about it, to work out exactly what we still need to know. Mark’s superb summary and collation of the research is here.

As a starting point, beginning with the end in mind, we set ourselves the simple constraint of one page. If we had one page that distilled and summarised the research for the classroom, we asked, what would we like teachers to have access to, free, online and neatly packaged?

So here it is.

What do we know about how memory works? What can teachers can do about it?



The research article that best synthesises a century of scientific research evidence is by five cognitive scientists: Dunlosky, Rawson, Marsh, Nathan and Willingham.

Here are their summaries of the two most effective strategies for retention, spacing or the distributed practice effect, and retrieving or the testing effect:


Distributed practice effect

Spreading of study across multiple, spaced, distributed sessions rather than cramming study into single sessions is highly effective for learning and retention.


Reviewing over 250 studies, Cepeda et al (2006) showed overall students recall more after spaced study (47%) than massed study (37%). Lag effects show the advantage of spacing with longer intervals over spacing with shorter intervals between practice sessions.



The schedule most strongly influences the benefits of distributed practice.

Schedule: How are learning episodes best spaced? Conditions with the most intersession forgetting yielded the greatest long-term retention: gaps of 30 days are shown to be better for learning than gaps of 1 day. Intervals of one month or more may be ideal for studying core content that needs to be retained for cumulative assessments.

Format: the distributed-practice effect works across free recall, multiple-choice questions and short-answer questions.



Practice testing has a high utility across learning conditions (reading, listening, writing etc), student characteristics (age, ability & prior knowledge), materials (problems, texts and questions across subjects) and tasks (recall, problem-solving and comprehension).


Testing effect

More than 100 years of research has yielded several hundred experiments showing that practice testing improves learning and retention. Practice testing is low-stakes or no-stakes, which pupils may even use on their own, outside class.



For example, final-test performance of word pairs is better for those that practice-tested (53% vs 36% after 10 minutes 35% vs 4% after one week).

Another instance is that 4 blocks of study with practice tests outperformed 8 blocks of study without (39% vs 17%).



How is testing best designed for retention? Format, dosage and timing strongly influence the benefits of practice testing.

On format, practice tests that require more generative responses (recall, short-answer) are more effective than those that require less generative responses (fill in the blank, recognition).

On dosage, more is better: final-test performance improves as the number of correct responses during practice increases, albeit with some diminishing returns as higher criterion levels are achieved.

On timing, longer is better: sizable benefits are observed when repeated tests are spaced: longer lags produce greater benefits. Testing effects are larger when final tests are given after longer delays, including intervals of 4 weeks, 4 months, 8 months, 11 months and even 1-5 years.

Practice testing outperforms restudying. The advantage of practice testing with feedback over restudy is extremely robust. Shorter and more frequent tests (one a week) are more effective than longer and less frequent (once every six weeks).

Distributed practice testing is more effective for retention than distributed practice alone.


Practice testing has a high utility across formats, materials, learner ages, learner abilities, outcome measures and retention intervals. It is the most effective learning technique in the research literature, according to these five cognitive scientists.



So, what do we still need to know? How can we guide research with the questions we want solved? As the cognitive scientists say, these ‘effective techniques are under-utilised’ by teachers and students, and ‘popular techniques (re-reading, highlighting) are ineffective’. They recognise the ‘disconnect between research and practice’ and say that ‘easy-to-use versions of the most promising techniques should be developed and evaluated in controlled investigations in educationally representative settings.’

As a starting point, the question we got onto by the end of the touchpaper party was: what is the optimal frequency and type of quiz for students to best retain content?

Much is still to be done: in particular, as Dr Becky Allen suggests, a much more extensive literature review.


As a very brief starting point, and to see how well the findings of cognitive science align, it’s worth summarising some key research papers in a single sentence:


Distributing studying over time, overlearning (continuing to study even after knowing the material), and testing frequent self-testing are the most important ways to improve students’ memory.


Spacing rather than massing practice, interleaving rather than blocking content, and testing as retrieval to aid future recall are the best ways of enhancing learning.

Dunlosky et al

A century of research shows that practice testing (low-stakes quizzing) and distributed practice (spaced, spread out study) are the most effective techniques for learning, evidenced across ages, abilities, tasks and contexts.

Cepeda et  al

A century of research evidence shows that cumulative review and increasing the gap between study episodes can enhance retention.

About Joe Kirby

School leader, education writer, Director of Education and co-founder, Athena Learning Trust, Deputy head and co-founder, Michaela Community School, English teacher
This entry was posted in Education, Psychology. Bookmark the permalink.

12 Responses to Cheat codes to intelligence

  1. debrakidd says:

    This is really interesting Joe – is there any follow up research to check whether learning had been retained after the final course exam? i.e. if it was ‘dumped’ after it was no longer deemed to be useful?

    • I haven’t read the post yet, but just off the top of my head, I think that kind of ‘dumping’ typically happens due to cramming, i.e. building short-term retrieval strength of memories but not their storage strength. It’s not the immediate sense of utility of the knowledge that guarantees its later retrieval, but its ‘retrieval strength;’ this degrades much more slowly over time (and is more readily rebuilt) if ‘storage strength’ is high.

      Storage strength can only be built over a long period of time.

  2. Tom Berend says:

    Thank you. Dunlofsky et al. is a 50-page paper, it’s going to take a while to absorb. But it seems that there are at least one more element you could add: ATTENTION, which is nicely described in Logan’s paper:

    According to Instance Theory, forming a neural connection is conditional on having sufficient attention, it’s an all-or-nothing event and it only happens when there is sufficient focus. The consequence for the classroom is that learning should take place with without distractions, background noise, competing ideas, fatigue, or discomforts.

    Of course, you want to lay down multiple connections, and assess progress. Logan argues that Instance Theory predicts a learning curve (power law curve of practice) merely as a consequence of forming connections. This lets you measure learning in some cases – simply test until the curve flattens out sufficiently.

  3. Pingback: Interesting read: how can we improve how much students remember | From experience to meaning...

  4. michaelslav says:

    Hi Joe (and Helene, Mark, Tim, Ben, Lucy and Jackie!),

    Thank you for a phenomenal post. I had not encountered the article that you linked to before and it is a source of immense interest. Did your discussions touch at all on how research such as this might inform a course of teacher training or CPD? It seems that the 10 techniques reviewed in the article might make for 5/6 excellent CPD sessions of their own right!

    Thanks again,

  5. Rory says:

    Here’s a 5-minute audio that solves a great deal of the problem of retention:

  6. Rory says:

    The problem with what’s being learned is that it’s all overwritten by new material almost instantly. This is especially true of information-rich information like video. I have a 5-minute audio that solves a great deal of this problem:

  7. I do not create many responses, however i did some searching and wound up here Cheat codes to intelligence:
    touchpaper#7 | Pragmatic Education. And I actually do have a couple of questions
    for you if you tend not to mind. Could it be just me or does it look like some of the responses look like
    coming from brain dead individuals? 😛 And, if you are
    posting on additional places, I’d like to keep up
    with everything new you have to post. Would you list of the complete urls of your community sites like
    your Facebook page, twitter feed, or linkedin profile?

  8. Pingback: A guide to this blog | Pragmatic Education

  9. Pingback: Why homework is bad for you - Mr Thomas' Blog

Leave a Reply