Why forgetting is so crucial to remembering

One of the key things to know from over 100 years of research on learning is the weird paradox that an important part of learning anything is actually forgetting it. A brief post on how to harness this principle for effective learning

In 1914, Edward Thorndike outlined his law of effect (based on experiments with cats) which became the blueprint for the behaviourist idea that positive experiences are reinforced and negative ones are weakened. But part of this broader theory was his 'theory of disuse' which establishes another fairly simple idea: the less you use something the more you forget it. The key idea is that unless you rehearse learned information and skills, it will fade or 'decay' as he put it, over time (use it or lose it). Now on the surface this seems like something everybody knows, but like so much about learning, it's really not that simple at all.

The 'decay' part of Thorndike's theory of forgetting was the problem and robustly debunked by John McGeoch in 1932. He advanced the idea that it isn't so much the "disuse" of learned material over time which causes it to be forgotten but rather the conditions under which it is to be recalled. In other words, the stuff is still there, you just can't remember it. (I like this paper because it contains one of the spiciest putdowns in early 20th Century psychology, at a time when everyone was unfailingly polite. He says Thorndike is chatting nonsense "because the principle of passive decay has no analogue anywhere else in science, and is illogical.")

The big idea that McGeoch advanced is the notion that memory is associative in nature, in the sense that certain stimuli will trigger remembering. In other words, you remember stuff through other stuff. But this can have negative consequences. He coined the term 'Reproductive Inhibition' which deals with the challenges faced in recalling new information in the face of previous learning. Put simply, the idea is that the ability to remember learned material can be reduced or enhanced by subsequent learning.

When new information is introduced, it can interfere with the memory of earlier material, making it more difficult to retrieve what was originally learned. This is also referred to as a "self-limiting process" (Roediger, 1978). This idea emphasizes that the process of forgetting is not just about the decay of memories over time, but also about how new learning can actively suppress or interfere with the recall of old information. The key point here which starts to emerge is that learning is not just about remembering, but also about forgetting.

In 1950, William Estes published a paper titled "Toward a Statistical Theory of Learning, " in which he argued that learning is a probabilistic process which is possible to analyse using statistical models. However, he claimed that learning is also a very unstable, uncertain process – things which work one day often don’t work the next day even with the same conditions. In other words, he predicted that learning is unpredictable.

What I personally like about Estes is that he suggested that our understanding about how learning happens across time was all wrong - he shifted us away from thinking about learning as a single episode to a more cumulative process which happens across many different episodes. This is a hugely important idea - more of that later...

In the 80s, the great John Anderson, outlined his model of learning and memory with his Adaptive Control of Thought (ACT) theory in which he basically argued that because of the limitations and constraints of human cognition, we will seek to optimise how we work to suit our goals. Because it's not efficient to remember everything, we will seek to narrow down what we remember by discarding a lot of stuff, despite the fact that it's in our best interest to remember it. In other words our brains work against us: we actively want to forget stuff, despite the fact that we probably need that stuff. How many times have you forgotten something you have carefully read or studied recently yet can still recall every character from your favourite movie as a 10 year old? ('Return of the Jedi' in my case)

Then in 1992, Robert and Elizabeth Bjork published 'A New Theory of Disuse' (a response to Thorndike) which really crystalised many of these ideas about forgetting and remembering and more importantly, framed them in a way that is especially useful for educators. There are two important theoretical ideas about forgetting in this paper. They define any item in memory has having one of two 'strengths':

1. Storage strength: This is how deeply encoded or well-entrenched a memory is. This is largely to do with how well it is understood or its meaningfulness.

2. Retrieval strength: This is about how easy it is to recall that memory.

This would lead them to form the concept of "desirable difficulties" which states that by varying how we learn, we can strengthen the ability to remember stuff. In other words, short term pain = long term gain.

If you want to remember stuff a year from now then you need to do something which can often feel uncomfortable in the present.

As Bjork/Bjork neatly illustrate, we are good at getting information into our memory but very bad at getting information out of our memory. This will be familiar to any teacher who teaches what they feel is an effective lesson to which all the students give a smiley thumbs up and appear to understand everything only for the teacher to encounter a sea of blank faces next week when asking what they remember from that lesson. Again, it's not that it's not there as Thorndike argued, it's that they just can't remember it.

Another issue is the problem of memory regression. this occurs when we don't access updated knowledge such as a person's newly married surname or how to drive a new car or a newly learned golf swing. Without using that new knowledge, we regress back to our old knowledge and forget the newly learned material or skill. BUT there is a big difference between experts and novices here - the regression can be bad for novices but actually positive for experts.

If I had to sum up the crucial idea of learning as a cumulative process of forgetting and remembering and in one sentence, it would be this: "the act of retrieval is itself a potent learning event." (Bjork, Bjork 1992) Retrieval practice isn't about testing whether someone has learned something, it's actually a vital part of the learning. In other words, learning doesn't happen in a single lesson or episode where the teacher has "covered" the content. It happens over repeated episodes where we encounter, forget, retrieve, associate and consolidate that knowledge. This process doesn't fit neatly into the boundary of a lesson unit which is why asking teachers to 'show learning in a lesson' is so misguided. What you're seeing is merely the briefest glimmer of learning which might lead to actual learning but which depends greatly on what happens next.

So what are the implications of over 100 years of research on forgetting for educators?

Principles to know:

• Our capacity to store stuff in memory is relatively limitless but our capacity to retrieve it is not. In other words, just because you know something, it doesn't mean you can remember it. This process happens in classrooms every day without us realising it - so much effort (planning/teaching/student work) is wasted because of ineffective instructional design.
• Our brains are quite efficient at some things but also really inefficient at other things which is why learning how to do quadratic equations is so hard for teenagers. It's also why for the vast majority of people, trying to discover stuff on your own is such a bad way to learn anything.
• We're very good at learning stuff we want to learn but incredibly bad at learning stuff we don't want to learn. (and most kids don't want to learn quadratic equations.) Storage strength and retrieval strength are crucial concepts to long term learning - without regular retrieval practice/questioning/generative activities we will not be able to recall learned material - it's as if our brain throws stuff in a warehouse without cataloguing/indexing any of it. Retrieval practice is the way to make that catalogue/index.
• This process of long term learning is largely dependent on the context in which students try to remember stuff and the good news is that teachers can do something about that (curriculum + instruction).
• You don't learn something when you encounter it, you learn it when you forget it-remember it-forget it -remember-link it to something else you know and so on... Think of learning as something which happens over 6-12 months not in a single lesson.

What to do:

• Because of these limitations of how learning happens, the first thing to consider is curriculum. But curriculum doesn't just mean what you're teaching (content), it also means how you teach it (instruction).
• Don't just ask students questions that have 'answers' - those answers really need to be connected to stuff they already know and this needs careful planning. For example, knowing the year of the storming of the Bastille is likely to be of very little use unless it's linked to deeper understanding of how this event relates to the broader trajectory of the French Revolution.
• When planning curriculum and building in retrieval practice, consider when they will forget the material and where they will remember it. As in, actually plan for them forgetting it. If you're teaching them something in September and then giving them a quiz on it in May and thinking that's retrieval practice then you're doing it wrong. Think really carefully about where they will re-encounter what you've taught them.
• Do you want students to remember what you're teaching them now or in a year? If it's the latter then you have to use desirable difficulties otherwise they will probably only remember it by chance not design. (See Bjork on this and also Fiorella/Mayer on generative learning.)
• Lastly, If you're leading teaching and learning in a school, stop asking teachers to 'show student learning' in a lesson. This is a waste of time and in most cases, gets in the way of actual learning.

Works cited:

Bjork, R. A., & Bjork, E. L. (1992). A new theory of disuse and an old theory of stimulus fluctuation. In A. Healy, S. Kosslyn, & R. Shiffrin (Eds.), From Learning Processes to Cognitive Processes: Essays in Honor of William K. Estes (Vol. 2, pp. 35-67). Hillsdale, NJ: Erlbaum.

Estes, W. K. (1950). Toward a statistical theory of learning. Psychological Review, 57(2), 94–107. https://doi.org/10.1037/h0058559

McGeoch, J. A. (1932). Forgetting and the law of disuse. Psychological Review, 39(4), 352–370. https://doi.org/10.1037/h0069819

Thorndike, E. L. (1913). The psychology of learning. Teachers College, Columbia University.