My usual definition of learning is ‘a change in long term memory,’ (after Kirschner et. al. 2006). Some might argue for a different definition to mine if they are prepared to accept a definition at all. Others might comment that reproducing the movements required to ride a bicycle and recalling the events leading up to the first world war are quite different and require different models. Fair enough, but ultimately both kinds of knowledge are represented as patterns in the brain that persist over time.
Is learning permanent?
I have not stated that learning is a ‘permanent’ change in long term memory. We gradually lose knowledge and studies map this predictably. On the other hand, some kind of residue is left behind.
I took my last chemistry class in the first year of university in 1995. Since then, I have taught basic chemistry up to Grade 10 and I also spent a brief period teaching physical chemistry to Grade 11 in the early 2000s. So it had been almost 20 years since I’d completed a titration when I was called upon to teach Chemistry at Grade 12 in 2014. I could not remember much of this unprompted, particularly electrochemistry, organic chemistry and those titrations. Yet learning it again was not like learning it for the first time. Not at all. I quickly filled-in the gaps and made mostly correct intuitive leaps.
Dan Willingham discusses this issue in a recent article for American Educator.
Learning is latent
The reason why we cannot definitively answer questions about the permanence of learning hints at another key issue: We cannot measure learning directly. This is not because it is an abstract or reified concept; learning does physically exist. If we were to dissect the brain of a person who had learnt something and we knew what we were looking for then we would perhaps be able to observe this learning in terms of the connections between particular neurons.
Yet the way that we tend to measure learning is by proxy. We give students some sort of task to complete and infer learning from this. If the goal of training is simply to be able to complete such a task then we don’t need anything more. Think about learning to drive; if you can drive then you’ve learnt how to drive.
Many teachers of academic subjects would not be satisfied by a student simply being able to complete a particular task. We are more interested in whether a student has grasped underlying principles. The tasks that we give to students do not always represent something of value in their own right, or even if they do they are used as indicators of other qualities.
And such ideas are not just the preserve of progressive educators. Teachers will cross the divide and all condemn the notion of ‘teaching to the test’. At its worst, this is about giving students short-cuts that make them appear to have an understanding that they don’t actually possess. Yet teaching to the test can also have in iterative value; learning what to do can lead to a greater understanding of why to do it.
For instance, a good test of a student’s understanding of connected bodies in physics is to ask them a question involving objects connected by pulleys, one of which is falling. However, this has become something of a common item on VCE physics exams. As a physics teacher, I therefore rehearse many examples, explicitly highlighting certain features and common errors. If a student subsequently answers such a question correctly then we cannot be sure whether they actually have a deep understanding of connected bodies or whether they are relying on the teaching that was specific to that problem type. Yet we might also expect explicit instruction on this problem type to lead to a deeper understanding of connected bodies. So it is not just about somehow gaming the exam.
There are those that think that we can overcome this problem by designing different kinds of tasks. My view is that, no matter how well-designed, a task will only ever be a proxy for learning. For instance, some people suggest that maths students should write explanations for their answers rather than simply solve problems. But we could still end-up teaching the explanations just as readily as we find ourselves teaching procedures. “Remember folks,” we might say, “If they ask you about… then you need to write…”
The most common solution is to suggest that we need to demonstrate transfer: if we can take what we know and apply it in a novel situation then this is evidence of learning. Indeed, in their highly influential guide to unit planning, Wiggins and McTighe suggest that the ability to transfer is understanding.
There are two major problems with this approach. Firstly, it is hard to control what students are exposed to ‘for the first time’. A novel problem to one student may look very similar to an example that a different student went through in class with her teacher or whilst researching at home. If teachers are aware of what is on a test then the temptation, conscious or unconscious, is to tailor teaching towards this.
The second problem is that transfer is a very high bar to set for evidence of learning. Transfer is a property of expert performance but even experts don’t need it much of the time. A plumber solving a plumbing problem is likely to relate it to plumbing problems experienced in the past. Truly novel solutions will be rare. If I learn to ride my bike down the street then are we really to conclude that I have learnt nothing if I cannot ride your bike along the local trail? No, I have certainly learn something, even if I cannot transfer this learning. In fact, much of our learning is inflexible in this way early in training. The danger with focusing solely on the end-products of complex learning is that it gives us little information and guidance along the path to such expertise.
We must certainly not equate assessment evidence with learning. However, if we cannot infer too much from a single measurement then a good strategy is to ensure that we have lots of measurements. Rather that focusing only on end-products such as essays, it also seems sensible to try to break down the learning that is inferred through such tasks into more discrete units; can a student write a paragraph? does the student know the text? Multiple measures like this will make our inferences about learning stronger.
If we want to avoid teaching to the test then the obvious logic is to use unseen tests. If we don’t know exactly what is on an assessment then we cannot teach towards it. You can do this in a school by asking teachers who are outside the teaching team, but who possess the relevant knowledge, to write the assessment. It is particularly odd to note that many people who are against teaching to the test also favour internal assessment over standardised tests. And yet standardised tests provide a wealth of norm-references information and usually use questions previously unseen by the teacher.