In defence of imperfect retrieval

If educational blogging is dead then no one seems to have passed that memo on to the memory department. There has been a wealth of superb writing about memory recently, drawing out the educational implications of research on the formation, encoding and retrieval of memories. These posts have been scholarly, accessible and persuasive.

For example, focusing just on memory retrieval, Sarah Cottingham has written two brilliant blogs on the mechanisms of why retrieval leads to stronger memories, and the classroom lessons for this. The first explains how interference (different memories interfering with each other) can be reduced by varying the retrieval context and questions, and by gradually increasing the effort required (by reducing the depth of the cues given) over time (whilst still balancing this effort with ensuring a decent success rate). The second talks about how we can use contextual cues to activate schemas for successful retrieval (again balancing this against the need for retrieval to be effortful), and do this on multiple occasions, spaced over time. Taking this a neurobiological step deeper, Gareth Bates’ blogs on ‘priming’ show how at a genetic level, the expression of genes related to memory formation is strongest on being re-exposed to newly learned material within a period of 3-7 days, with subsequent retrieval tasks then being preceded by a ‘warm reactivation’ of the schema before full retrieval.

Putting all these together, we have a nuanced and multi-step process of memory formation and the role that retrieval plays in it. We should be initially allowing students experiences of relearning and ‘warm reactivation’ (perhaps using relatively extensive contextual cues), before carefully managing the balance between making their subsequent retrieval effortful and successful as we remove the scaffolding. I’m going to refer to this sort of sequence as ‘warm’ retrieval practice from now on. The idea of ‘warm’ retrieval also fits very nicely with the work of Rosenshine, who talked about retrieval activities as a means of activating the prior knowledge that students would need for the coming lesson, in order to both facilitate the original memory and the new learning. Such nuance, however, is not always conducive to the writing of whole school teaching and learning policies, and as Tom Sherrington has this week observed, ‘retrieval practice’ in schools can often become something pretty far removed from the responsive process above. Inner Drive and Adam Boxer have also recently blogged about ‘how to avoid getting retrieval practice wrong’. Whilst Tom especially is criticising the use of whole school strategies which are not responsive to actual learning, he is also critical of ‘cold’ retrieval activities, e.g. questions from across an entire specification, which may not relate to the topic of the coming lesson, and which may not have been primed or reactivated in advance. He writes:

“Rather than it being a consolidatory reinforcement of their learning, plugging a few gaps, students feel they are being quizzed on the thin air of vaguely-encountered wisps of disconnected factoids from a dim past.”

Sherrington calls retrieval practice in some of the guises he describes a ‘lethal mutation’ of the research, and it is a shame to see retrieval practice being so obviously misused in places. I wanted, however, to offer a slight counterpoint to all of the exceptional work above. Importantly, this is in no way to dispute the research or the conclusions that others with more expertise than me in the area have drawn from it. If I were able to follow their prescriptions all the time then I absolutely would do so. But I’m not sure I am. Here’s why.

  1. Not all schemas can be adequately warmed

I teach a content heavy subject (A-Level Psychology) in which the vast majority of my students have not done the subject at GCSE. They are therefore learning the majority of the material entirely afresh (given that most of it also doesn’t really overlap with any other GCSE subjects). In addition to the amount of content, the course takes in a number of separate units (11 in total) most of which examine different applied in topics in psychology. In some of these we can very easily follow the process above, as we can look at how different fundamental ideas about psychology have been used to explain behaviour in different areas. We can use warm reactivation of the schemas for the basic ideas, integrate this meaningfully into the knowledge of the applied area, and then gradually withdraw the support and expect students to be able to recall all of this together. I can also then subsequently draw on these schemas when teaching other related content further down the line. Unfortunately, this isn’t always the case. Some of the applied units are much less clearly linked to each other, or to the foundational ideas of the subject. They are just as important to the students’ success, and fluent recall of them is just as essential, but the schemas that students will have for these units will be far less integrated into a coherent whole, and therefore any retrieval that we do on those areas is always more likely to be ‘cold’ (I tend to use a retrieval activity at the start of lessons that has a few ‘rewarming’ questions on recent and related content, and a few older questions from the rest of the course). If I had unlimited time then I could of course ensure that these areas are ones where I reteach concepts or rewarm schemas progressively over the course of a week or so, before revisiting them a few months later, but I don’t. I have other material to cover and limited time in which to cover it. My choice is therefore to include it cold, or to not include it at all. This is an open question I think. Perhaps the opportunity cost of the time I lost doing colder retrieval questions would be better spent doing more teaching. I’m not convinced by this, however, for the two further reasons below:

2. How cold is cold?

Ideas of ‘warm’ and ‘cold’ retrieval, as described above, appear to relate to the activities of a teacher. But of course we are only one (less important) half of the learning equation. As students progress through to the higher age groups the focus on independent learning increases, as does the requirement that they are developing their own habits of regular review and self-testing. If this is happening, then actually retrieval cues that might appear to be cold could be ideally placed, following some independent rewarming of the relevant schemas. In a similar way, the gap between ‘effortful’ retrieval (which is good), and unsuccessful retrieval (which is bad) is going to be very woolly, and subject to all sorts of individual differences. Taking it a step further along the learning time line, would an initially unsuccessful retrieval experience, with some adequate feedback, not then subsequently serve as a reactivation of the relevant schema, which can then be capitalised on by further independent review? That would seem to make sense, given that we know that retrieval failure followed by corrective feedback leads to the same learning gains as initial retrieval success. As a result I’m not convinced that we can label retrieval experiences ‘warm’ or ‘cold’ in a way that describes the experience of a whole class reliably, especially over time.

3. Student self-efficacy

Building on the above, one of the major benefits that I have seen retrieval have for most (though I admit not all) my students is in encouraging their self-efficacy as learners. Yes they probably get some questions cold, but if the right conditions are in place (see final section) then this can still form a very valuable learning experience. A belief in your ability to effectively take charge of your own learning would be an effect that operates, if not entirely independently of the actual memory effects (as it is more likely to be fostered by successful recall), then at least partially independent of them.

4. Final performance vs acquisition

I also wonder if it is realistic, early in the learning sequence, to expect schemas to be deeply enough interconnected for perfect ‘warm’ retrieval. It is increasingly accepted in schools that performance in the acquisition stage of a skill may not resemble the final performance of it at the end of a learning sequence. Consequently, assessments increasingly distinguish between assessment methods during learning from the methods used to measure final performance. Won’t the same be true for schemas? The end goal for the knowledge of our students is that they can flexibly transfer it to a range of situations, and see the big picture connections and overarching narratives of our subject. One of the great joys of teaching Y13s is seeing this develop. However, it often only comes about pretty late on in the course, sometimes as late as the final weeks or days. Why? Because it’s only then that the depth of their knowledge has become sufficient for these connections, the ley lines of the subject, to magically appear. I remember a very similar process happening in the week before my finals exams at university, when I’d finally learnt enough for seemingly disparate topic areas to coalesce into a coherent whole. It was exhilarating. Of course, this isn’t an argument for ‘cold’ retrieval specifically, it’s merely an observation that if we accept that early schema development may not be sufficient to assess progress using measures of final performance, then we should also not always expect it to be developed enough to allow us to do perfect ‘warm’ retrieval practice.

Making imperfect retrieval non-lethal

As I said above, were I always able to fulfil the criteria for ideal retrieval in my lessons I would. Having said that, and with no claims about this being in any way an ideal system, here is what I do which I think makes a reasonable attempt of an imperfect situation.

  • Have clearly specified questions, which are available to students at any time. We use a version of a core questions curriculum, with retrieval questions drawn from these. This means that (IN THEORY!) no questions on past material should be entirely cold, given previous exposure.
  • Constantly emphasise the process of learning outlined above, and the student’s own proactive role in it. E.g. taking control of the ‘re-exposure’ to learned material in the week after initial learning. Using self-testing to reactivate schemas regularly. Providing resources to aid this, e.g. all students have the spreadsheet of core questions, and it is also available on Carousel learning so they can easily self-quiz electronically., as well as regularly setting retrieval homeworks.
  • Create accountability in retrieval. Tom Sherrington talks in his blog about ‘green-penning’ answers with annotations which then don’t go anywhere. Two things I use to try to avoid this practice are:
    • If a student gives a wrong it incomplete answer upon cold-calling after the retrieval quiz, then I note the number of the question and the name of the student, and make sure that the same question is asked to the same student next lesson. With just one entertainingly defiant exception this year the student has always been able to give me the right answer. This provides a lovely chance to celebrate success, even for students who might initially struggle with recall, and also hopefully contributes to a ‘culture of error’ where we understand as a group that learning doesn’t always follow a smooth path.
    • Dedicated homework tasks to ‘follow-up on feedback’, where students are required to show how they have responded to any aspect of the previous week’s lessons which identified a gap in their understanding. Often this might include following up on some of the ‘green-penned’ annotations from the starter questions, for example by making flashcards or looking at practice questions on that topic.

Does this system make more of an impact to my students than if I did less retrieval, but with more time spend on warm reactivation of schemas, even for topics where that might not fit naturally into the ongoing sequence of learning? I don’t know. As is often the case with questions like this, it would be great to have some classroom-based research which actually investigated such specific practice-focused questions. I’m also very happy to be persuaded that my approach is misguided, in which case look out for my retraction blog next week! For at least the rest of this week, though, I’m going to go on trying not to make perfect the enemy of decent.

In defence of imperfect retrieval

One thought on “In defence of imperfect retrieval

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s