Over the last decade, the conversation around cognitive science and psychology in education has grown ever louder, to the point at which these discourses have come to be seen as one of the dominant theories in contemporary education. Much of the discussion focuses on pedagogy including the role of memory and remembering, with theories of learning and teaching being based on the retrieval of information in the long term. Although the ability to remember information accurately is undoubtedly an important aspect of learning, forgetting is an important issue to consider when thinking about learning and seems to be not as widely discussed within education.
This blog will discuss the seminal work by Ebbinghaus and explore its role in the educational conversation and the many iterations of the forgetting curve which have emerged through teachers applying this to pedagogy.
Ebbinghaus was an experimental psychologist who was interested in finding a mathematical relationship between the elapsed time post learning and forgetting. He conducted a number of experiments in the early 1880s in order to establish this.
In his experiments, Ebbinghaus attempted to learn a row of thirteen nonsense syllables until he was able to freely recall each one in the correct order. After a preset time interval, he would relearn the syllables, given the fact he had forgotten them, until he could once again freely recall each one in the correct order.
It is important to recognise that Ebbinghaus’ view on forgetting was not a measure of how many syllables that could be recalled after a specific amount of time but the amount of time, or repetitions, it took to relearn the same list of syllables after forgetting. A measure he called savings. Savings can be presented as a decimal or a percentage and is calculated as follows:
If it took someone initially 10 minutes to learn the syllables but it only took them 8 minutes to relearn after a set time then the saving is 2 minutes. Savings is the 2/10 = 0.2
If the relearning took the same amount of time, then the savings would be 0 and if there was perfect recall without relearning, the saving would be 1 or 100%.
The original experimental results have been successfully replicated a number of times, but I am going to use data from the study by Murre and Dros in 2015 (paper can be found here) to discuss the forgetting curve due to the fidelity of their experiment. In their paper, Murre and Dros replicated Ebbinghaus’ experimental procedure and calculated savings using time. The resulting forgetting curve on a linear time scale is shown below:
The curve shows a general exponential decrease in savings. What is interesting is the higher than expected result for 1 day. Ebbinghaus also found this but he was able to fit the data point to the curve generated from his ‘forgetting equation’ so he overlooked this at the time. However, he did replicate, along with other subsequent researchers, this result after the publication of his work. This decrease or ‘slowing’ of forgetting from these experiments is thought to be due to the role of sleep in memory consolidation.
Interestingly, Murre and Dros recorded the number of correct responses (correct syllable in the correct position) during the relearning phases of their experiments. What this showed is that the proportion of correct answers after 20 minutes was marginally above 0.3 and this only decreased slightly at the longer time intervals.
Should we forget the curve?
From a position of experimental psychology the work of Ebbinghaus needs to be studied and remembered as it paved the way for psychology to have robust methods and rigour in the design of experiments that are still used today. The fact that the results of Ebbinghaus have been replicated a number of times is testament to this.
In terms of the educational conversation, it is useful to ask if we actually need a mathematical model (the graph with numbers) to tell us that learners forget. It is clear that what the Ebbinghaus’ forgetting curve does show is that:
1. a high proportion of information that is learnt is rapidly forgotten
2. the longer you leave before relearning something, the longer it will take you to relearn
I think I would be hard pushed to find a teacher that genuinely would disagree with these statements, with or without knowledge of the curve. The question we ask then, what use does awareness of Ebbinghaus’ curve brings to a teacher beyond the knowledge that forgetting takes place over a period of time after the point of learning?
Certainly, the misinterpretation and misrepresentation of the curve is not helpful. Making claims like “you only remember x% of information after y time” is clearly untrue if you are using Ebbinghaus as your evidence base. Applying ideas like this to education is widely problematic and can result in unhelpful numbered things about forgetting, models like the infamous learning pyramid.
Additionally, there is a danger with using a mathematical model rather than just having good awareness that forgetting takes place and that there are well researched methods to remedy this. For example, we might say we forget 50% of something we have learned within an hour. This sounds plausible and whilst you might worry about all the different permutations, that’s the least of the problems. Using that premise, I could simply say, well I’ll double the information learned at the start and then they won’t forget what I intended them to learn.And of course, the teacher in you will say that’s nonsense.
Being focused on forgetting is a good thing, but it is important to think critically in our application of science just like Ebbinghaus himself was.
Recently, @TeacherTapp recommended our blog titled: ‘Cognitive Science v Neuroscience: retrieval at the start of a lesson or not?‘ based on research in a neuroscience paper on how memory is formed and it has produced quite a strong, but positive reaction. Experienced and new teachers alike have said how the key process of ‘priming‘, a process of how memory formation happens ready for retrieval or constructing new knowledge or skill (a process which we established in our blog), resonates with their experiences and indeed CPD offered by many in education. It changes the debate around memory, the formation of it and how we approach education – if we believe that we can appropriate ideas from both cognitive science and neuroscience into education (note the many limitations with doing exactly that). This further blog aims to revisit key aspects of educational ideas and policies which are reliant on the notion of how memory is formed through this new lens of priming. It must be said that this blog is theoretical speculation and is done to give you some scope of where we will be looking to research the concept of priming and to explore if there is evidence to support this idea.
Enjoying this blog? Please consider tweeting it out to help share the knowledge in our community.
The immediate response from the original blog was that interleaved retrieval practice could have more limitations than we first thought. For example, doing interleaved retrieval practice at the start of a lesson in which the retrieved schema is not going to be used in the lesson would be working with the wrong memory cells (containing the schema) if we follow the neuroscience research outcomes. Instead of readying the reformulated architecture or strengthening the memory cells for the lesson, it is readying unrelated memory cells that do not have reformulated architecture ready for gene expression. In other words, the start of a lesson isn’t the right place for interleaved retrieval practice. The good news is that online asynchronous learning has been accelerated in its development and uptake over the last year thanks to the reduction in the number of children going to schools during so called ‘lockdowns’ (schools never closed(!!)). Interleaved retrieval practice therefore could still have a place online and asynchronously. However, the practice would need to change from being cold retrieval practice to a two step process of warm reactivation and then retrieval practice. This could be achieved, for example, through undertaking reflection, watching a short video, viewing some modelling or perhaps writing out a non-assessed overview e.g. a synopsis of a play. There are clearly implications here for how online learning is constructed and so that area is something to revisit separately. Further, the gap in time between revisiting learning and the way we revisit learning is affected. Too soon and the cells have not reformulated their architecture ready for expression of the arc gene. Too late and with no priming then no expression of the arc gene takes place.
There are further areas where our ideas about memory are predicated in policy and practice. Take, for example, the notion that ‘learning is a change in long term memory’. This is a prevalent idea found in the OFSTED research framework and the OFSTED inspection handbook which talks about teachers ensuring that pupils ’embed key concepts…and apply them fluently’ (p.44) as well as ‘transfer key knowledge to long term memory’ (ibid). These ideas are now in the Core Curriculum Framework (CCF) for trainee teachers upon which the Early Career Framework (ECF) is founded and also the NPQ suite of qualifications (for first teaching 2021 onwards) have been built. The CCF and the NPQ suite of qualifications sets out ideas such as ‘…committing some key facts to their long-term memory is likely to help pupils learn more complex ideas.’ (p.11) and very importantly, ‘Requiring pupils to retrieve information from memory, and spacing practice so that pupils revisit ideas after a gap are also likely to strengthen recall’ (p.12). All of this language is clearly from ideas that have emerged through cognitive science. The fundamental ideas here stems from the paper from Kirchner, Sweller and Clark which focuses on the concept of working memory and the limitations of asking working memory to learn or problem solve without prior instruction (discovery learning). The neuroscience model of memory formation adds to this language and enables us to revisit some of these core concepts. Memory, according to the research in the neuroscience paper cited at the start of this article, is formed through expression of the arc gene. By controlling the priming process large increases in gene expression can be produced at the points of memory formation which leads to enhanced remembering. Why this process has evolved is unclear, but a workable analogy is that memory can be used as almost an immune response to threats (retrieval and/or problem solving/constructing new knowledge or skill (creativity?)). Problem solving, we could theorise then, requires a series of processes to be successful. An initial activation event, a period of time, and then a warm reactivation event in which both retrieval and construction of knowledge happen simultaneously – there would be initial activation and expression of the arc gene happening simultaneously. The key concept is that remembering works like an immune response to a threat. How strong that response is relies on the original gene expression at the point of memory formation. To respond to this threat requires both strong retrieval and constructivist thinking. If the memory has not been strengthened by expression of the arc gene prior to retrieval then the resultant retrieval will be weaker than if the full priming process had been used. Further, if you are undertaking retrieval practice to strengthen the memory (in cognitive science words, to put knowledge into long term memory) then cold retrieval practice does not necessarily lead to the expression of the arc gene necessary for a subsequent strengthening of the memory. Cold problem solving, cold questioning, cold retrieval practice; all these do not fit with the neuroscience evidence of how memory cells are strengthened (memory formation with gene expression) or used effectively to deal with the subsequent activities which rely on memory (remembering). I suppose it’s a little like the athlete who spends time visualising and reflecting prior to an event – in effect, getting themselves ready ahead of the event (threat) following winter training. Controlling the priming process enhances both the remembering (what to do) alongside the constructivist problem solving (how to resolve an unexpected threat). What ‘learning’ is then, if we follow this train of thought, is not necessarily ‘a change in long term memory’ – long term memory is not a static schema that remains the same. You learn it and then you strengthen it ready for future use. You then also are aware of the priming process for its use at a point in the future. Retrieval practice alone, then, does not strengthen as effectively as memory which has been formed through a priming process. The memories therefore require the full priming process if they are to be strengthened and in order to be able to work at peak effect at some unknown point in the future. Memory can wane just like your immune system can wane. Yet, given the right priming it can be ready to retrieve and problem solve a threat a long time after the original learning and priming process actually strengthened the gene expression.
At a visual level, imagine that there are ten cells. One becomes populated with memory through the initial activation, but no reactivation through a process of priming. Through conditioning (retrieval practice), you can make the one cell very efficient at producing the memory at will to a specific stimulus. Imagine now that controlling priming to enhance gene expression (the arc gene which is associated with memory) brings 9 more cells into play. The memory now sits in ten cells and thus the original formation was much stronger. The resultant remembering is supercharged and thus is able to be used more effectively within future learning opportunities. It also requires less future retrieval practice conditioning. Indeed, the science says that subsequent revisits to the memory are not having the same impact on the expression of the arc gene as the second encounter with the learning. That second encounter and how it happens is where the majority of gene expression is happening.
Learning, then, happens as part of a process. New knowledge is constructed into pre-existing knowledge but then needs to go through a priming process to be expressed in the form of the arc gene. The whole priming process is essential as this is what causes the amplification of expression of the arc gene to happen. Although the whole process is important, the ‘learning as a change in long term memory’, if defined by gene expression, is happening predominantly during the reactivation event. But alone, it is insufficient: it needs to be taught, a period of time allowed, warmly reactivated and then it is ready for strong remembering for ‘threats’. Those threats should be both retrieval (known threat) and constructivist retrieval (unknown threat). In short, it is a moving sequence vulnerable to time delays and cold threats rather than a static schema which sits in the long term memory.
What does this look like at the level of lesson or learning episode? Well it begins to adapt some of the pedagogical tools we use – in particular formative and summative assessment (threats) as well problem solving or creativity and the length of time between the initial encounter with learning and when and how it is revisited. A first teaching should be followed by a period of time to allow for cell reformulation . The warm reactivation becomes the super important event. It suggests that before introducing a ‘threat’ during the revisiting for the second time – whether that be Year 1 painting in primary school or leading a Q&A session in English, you would undertake a non-threatening priming activity as a pre-activity. Not low stakes retrieval practice, but low stakes warm reactivation. To be very honest, this is not wholly new – doing a pre-questioning session to make a Q&A session better than a cold Q&A session is something teachers already do. Questions like ‘Do you like Lady Macbeth?’ would be a clear warm reactivation question. There isn’t any right answer, but the cells containing the memories would be readied. If asked later on for Lady Macbeth quotations the pupil’s memory cells would retrieve these successfully and also this would lead to expression of the arc gene. This is a reversal of what is currently happening.
I suppose at this point we begin to reflect on cold assessment. Every teacher knows that cold assessment is never as successful as assessment where the learning has been warmly reactivated first. It makes us consider if our wholly cold national assessment system is an accurate way to measure learning amongst children as well as the quality of teaching in schools. It would be interesting to see what effect controlling priming would have on large scale assessment systems such as mock examinations. Warm reactivation before mock examinations could be more effective at strengthening memory that simply delivering cold mock examinations. There is something to be said there about how a teacher knows better what a student’s knowledge or skill is like because they see the student in operation when they are primed rather than in the cold examination hall. It possibly also explains how a student can prime themselves more effectively for an exam than they ever did in class and thus score more effectively in the cold formal exam than they did at school (e.g. in a mock examination where they were not primed nor motivated to prime themselves).
Another area worth revisiting is Rosenshine’s Principles of Instruction. Here we see the idea that we should be reteaching that which has not been retrieved accurately. However, if the reason that the retrieval was unsuccessful was that there was weak memory formation then this reteaching could be simply relaying down foundational knowledge rather than creating the priming process for expression of the arc gene needed for a stronger future remembering. With the idea that memories can get better at responding to threats (retrievals) by being warmly reactivated first and then exposed to the threat, simply retrieving at the start is not the right idea. Indeed, one of Rosenshine’s key principles ‘…the review at the start of the lesson…’ (Rosenshine, 1982, p.8), is very nearly there. It’s just been conflated with contemporary ideas of retrieval practice and sometimes reduced to cold quizzes at the start of a lesson – further more, not enough attention has been paid to whether it is a second revisiting and how much time has lapsed between the initial encounter with the learning and the second revisiting.
It’s important to think about how we could further support children with learning difficulties through priming. By offering CPD to support teaching assistants in the theory of priming they could better understand just how important theories such as Porges’ work on the Automatic Nervous System (ANS). In his modelling, some learners who have had adverse childhood experiences (which could well include negative school experiences due to early issues with learning needs) have amplified ANS responses. In other words, when faced with ‘cold’ unprimed ‘threats’ (cold questioning, cold retrieval practice, examinations, etc.) some children respond with ANS driven fight or flight responses. By introducing warm reactivation ahead of such situations, support assistants could modify the ANS response, reduce behaviour-led responses and increase successful retrieval and construction of new knowledge or skills. Those with SEND can also have higher absence rates or simply have insufficient adaption in a lesson – this could affect the priming process: the initial event, the period of reformulation and then the expression of the arc gene are all susceptible to absences or issues with access to learning (barriers to learning).
I’m sure we could go on, but it is worth thinking about how enshrined our ideas about memory are into the way we train teachers, leaders, inspect schools and so forth. The new ideas from neuroscience both complements some of these ideas and challenges them, but also refines the language as well as lends criticality to the way we understand the processes. However, it does need a jolt of reality. Much of this is theorising and it is important always to be critical of everything we meet in education – every idea in education, after all, has limitations.
We have launched a two phase project to investigate the concept of priming and enhancing the formation of memory using these ideas. The active part of this project will run from September 2021 to July 2022. We have recruited 40 schools who are currently participating in this research. If you are interested in being part of this research (we still have space for further schools in phase 2) or any other research projects then drop me an email at firstname.lastname@example.org or you can find me on Twitter at @englishspecial.
Dr James Shea, Principal Lecturer in Teacher Education
It’s a simple premise: if you think memory is important to learning, then memory formation is, by this very argument, important as well. And memory formation comes before memory conditioning (e.g., retrieval practice). With the publication of a neuroscience paper on memory formation we now have convergence between cognitive science’s research in how memory works [in education] with neuroscience’s research in how memory works. Remembering knowledge over time and how to do something, after a gap of time, are very much vogue in education right now and rightly so. However, this article is about the publication of neuroscience research and what this neuroscience paper suggests about the way some in education will approach their lesson design as a consequence (a longer write up of these ideas can be found here).
Let’s get some caveats out the way. The research in this paper is about cells in the brain and formation of memory through the expression of genes. It’s not about the more social concept of mind, nor indeed does the paper suggest what teachers might do with this new knowledge of the brain. I’m not claiming that this paper proves anything about how we should teach, but I’m aware that it does bring criticality to the way that some think that memories are formed and need to be conditioned (especially within education) and so it is worth investigation. It’s such a simple premise: do you think memory is important to learning? If the answer is yes, then memory formation is just as important as memory conditioning, but more importantly comes first in the time line. And, you might ask yourself, how long is that time line?
Enjoying this blog? Please consider tweeting it out to help share the knowledge in our community.
So what does this neuroscience paper say? Well it said that the formation and recall of memory works a bit like a body reacting to a vaccine. The first jab makes the body receptive to it and gets it ready for the booster. When the booster arrives a surge of gene expression takes place, the strength of which controls the number of cells ready to be recalled in case they are required (e.g. if an infection shows up then the body is ready to react). So there are three parts of the process. The initial activation is the first step – this is where memory is formed for the first time. The next part of the priming process is at a genetic level – a reformulation of the cell architecture which is effectively ‘readying architecture for gene expression’. Finally, a warm reactivation event in which expression of the arc gene happens across the reformulated architecture ready for future remembering. The reactivation event is the crucial event. We can speculate then, that when you teach you are laying down the first activation- a first pass of the schema* itself, but the gene expression at this point will be small. Indeed, over time, the effectiveness will wane as the potential memory cells have not had a second reactivation event (weak remembering (forgetting) is why teachers are using the conditioning approach of retrieval practice so much). When you do enter the reactivation window that is when you supercharge the memory cell formation. Further, should you need the schema to be ready for recall and to be produced on demand (e.g. an examination), you undertake a warm reactivation event yet again to prepare for the recall (retrieval) event itself. Instead of going straight into a full retrieval of memory cold, there should be a warm up process where you are getting the memory ready to access the schema in case it is required. Then you should ask for a full retrieval. It is very important to note the distinction between retrieval (accessing memory cells) and retrieval practice (conditioning a rote response to stimuli through the ‘testing effect’). Our work here is focused on the strong formation of memory and subsequent strong remembering, not on conditioning a response.
Stage 1 – teaching and re-teaching of new/schema through an activation phase (followed by a delay of some days to allow the biological reformulation of architecture). Update: Phase 1 trials suggest that a 3-7 days gap following activation is the optimum time span.
Stage 2 – warm reactivation of the schema leading to revisiting the schema from Stage 1 (this triggers the expression of the arc gene needed for memory formation)
Stage 3 – remembering of schema associated to Stage 2 alongside teaching of new schema
How does this affect teaching then? Well, currently, there is a lot of focus on retrieval practice, a psychological conditioning process – retrieving knowledge repeatedly with the view to making the recall stronger through something called the testing effect. Yet this paper is suggesting that there is a big elephant in the room. Memory formation needs to happen first. And memory formation takes much longer than you think. If you go straight for a retrieval quiz then that’s the equivalent of asking the schema to be recalled without having reformulated cells ready to express the arc genes. You are still in the activation phase. It’s not necessarily strengthening the memory formation. Cold questions and retrieval quizzes at the start of the lesson and very soon after the first learning don’t reflect what this neuroscience research says. What you should be doing is something else – controlling the priming event.
Controlling the priming event means knowing architecture has formed and is only available for a short window of time and that you can take advantage of this knowledge through a variety of pedagogies. Think about discussions and recaps of the topic – quite wide ranging discussions rather than small minutiae. Then, in the main part of the lesson, the small minutiae will more accurately be recalled and more importantly, the process of priming and the subsequent remembering (producing memory to face the challenge) will work more effectively. It is important to see the nuances here of some of the things we do in teaching. Retrieval at the start of a lesson of content that is not going to be used is reactivating the wrong (cells containing the) schema. In addition, that’s a different process altogether – that’s conditioning – (retrieval and retrieval practice are two different things). According to the science in this paper, you want to be ensuring that you actually work with the schema for which the cellular architecture was created as that will lead to expression of arc genes which are responsible for creating more memories. In some ways the paper informs us on the gaps between cognitive science and neuroscience.
The main gap is this – much of the work on what we do as teachers is once memory has been formed: reducing extraneous load, conditioning memory through revisiting and so forth. However, all of this practice is reliant on the idea that memory is formed in a one off event and that we begin the conditioning process immediately. What the science in this paper says is that is not how memory is formed. And this has implications for the way we as teachers approach learning. There is a period of time where memory is formed. It’s not instantaneous and nor is it the time for conditioning. Instead, it’s the time for supercharging memory formation. If memory is important to learning, then creating strong memories could, we speculate, lead to more efficient use of time by a teacher later on once memory has been formed. Less time spent conditioning memory would free up time for more learning of further knowldge to take place – but that’s something for our research study to consider.
The future could be, one could speculate – not starting the lesson with the conditioning process of retrieval practice, but starting the lesson with warm reactivation of the schema and with a teacher awareness of who is not meeting this learning for the second time (e.g. absences, barriers to learning) and how long it has been since the activation phase. In particular, you should not start the lesson with retrieval of schema which won’t be used in the lesson as this does not lead to expression of the arc gene necessary for memory formation. The lesson itself should contain both remembering and new learning together as the brain constructs the new knowledge (or skill) into gene expression and starts to get further architecture ready for the next warm reactivation event. When that happens, it will be ready to swing into action with both the original knowledge and the new knowledge constructed into a single schema. This stronger formation of memory will lead to better remembering and require less future retrieval practice (because we currently use conditioning to supplement weak memory formation). Retrieval should still form part of the lesson, but if incorporated into the main lesson following a warm reactivation event at the start of the lesson then it will be more effective. And lastly, remember, the schema used should be relevant to the lesson.
It is a potential change in the sequence of learning that we have come to see become quite mainstream. First, an activation event, then a gap of time to allow for reformulation of the cell architecture (Phase 1 trials suggest the optimum time for this gap to be 3-7 days), then either a warm reactivation event alone or warm reactivation and new knowledge together to start the architectural reformulation necessary for expression of the arc gene in the next warm reactivation event. In addition, less retrieval practice is necessary. This is because we currently use retrieval practice (interleaved or not) to condition a pupil into producing knowledge in response to a question – known as the testing effect. By making the original formation of memory stronger through the priming process (expression of the arc gene) remembering will be stronger. You are supercharging the formation of memory through creating more gene expression at cell level. We have worked up further thinking here on what the implications would be for interleaved retrieval pratice, Rosenshine’s Principles of Instruction, OFSTED inspections, NPQs and even SEND. All of these areas could be affected by this concept of supercharging memory formation.
There is lot of theorising there and I’m sure those with good knowledge of both science and education will be able to add more thoughts. This paper does not at any point inform us how to teach and it’s important to emphasise that point. There is a major caveat also which is that taking science and turning it into educational practice has lots of limitations in terms of ecological validity. However, it is an interesting paper and it does suggest a different, and scientific rather than theoretical, model of how memory is formed and how a priming event could be better than retrieval for the start of a lesson.
We have launched a two phase project to investigate the concept of priming and enhancing the formation of memory using these ideas. The active part of this project will run from September 2021 to July 2022 and we have completed Phase 1. Our thanks to everyone who participated. We are now working on the tool kit for Phase 2. We recruited 40 schools who are currently participating in Phase 2 of this research. If you are interested in being part of this research (we still have space for further schools in Phase 2) or any other research projects then drop me an email at email@example.com or you can find me on Twitter at @englishspecial.
Dr James Shea, Principal Lecturer in Teacher Education