Cognitive Load Theory: a life of its own

In 1971, in Palo Alto, California, the Stanford Prison Experiment took place.  You’ve probably heard of it. In case not, here’s a quick rundown.

The SPE was led by Philip Zimbardo, of Stanford University’s psychology department.  It involved about 20 students, all volunteers.  One morning, half were “arrested” for theft or armed robbery.  They hadn’t really committed the crimes, but the arrests looked, and doubtless felt, real enough.   The prisoners were taken away in police cars, stripped naked, given numbers and prison garb, and locked up three to a cell.  

The other volunteers took on the role of prison guards.  They were to keep order.

Within hours, the inmates rebelled.  The guards responded with fire extinguishers, physical punishment, sleep deprivation and refusals to let the prisoners empty their sanitation buckets.  The researchers concluded that these otherwise upstanding students had acted in such appalling ways as a result of their environment.  Put otherwise good people into bad situations, and their behaviour will deteriorate.

However, according to Rutger Bregman’s book Human Kind, the study was “a hoax.”  The scientist in charge had told the guards how to act, rather than letting their behaviour develop unencumbered.  Zimbardo denied it for years.  As a result, according to Bregman,

“In the decades since the experiment, millions of people have fallen for Philip Zimbardo’s staged farce.”

And as for Zimbardo himself, asked in 2018 how he felt about his research, he said:

“It’s the most famous study in the history of psychology at this point.  There’s no study that people talk about fifty years later.  It’s got a life of its own now.”

He’s not wrong.  Today (18 September 2022) I googled “Stanford Prison Experiment.”  The very first hit – apart, of course, from Wikipedia – took me to SimplyPsychology.org.  It describes the experiment and its conclusions in some depth.  It doesn’t say anything about it being a hoax.  A life of its own indeed.

———————————–

I feel like this about cognitive load theory (CLT).  I’m not saying it’s a hoax, nor impugning the good auspices of its proponents.  I’m also not saying that I don’t buy it.  What I am saying is that it’s got a life of its own, and that may not be a universal good.

What is CLT?

If you’re not totally familiar with CLT, an admirably succinct definition was given by Oliver Lovell, author of Sweller’s Cognitive Load Theory in Action, in this webchat with Tom Sherrington: “to reduce extraneous load and optimize intrinsic load.”

What this means is that our students’ working memory, which as we know is limited to around 4-7 items at any one time, can be filled by:

  • intrinsic cognitive load.  This refers to the innate difficulty of whatever it is you are teaching, be it fractions or the backhand or erosion; or,
  • extraneous cognitive load.  These are the distractions provided by the environment: what’s on the walls, the flickering of the light that doesn’t quite work and, most pertinently, the amusing yet irrelevant cartoons on your worksheets, the crazy animations in your powerpoints, the jokes in your explanations – anything that stops your class thinking about what you want them to concentrate on.

There’s also often held to be a third form of cognitive load: germane.  This is the good stuff: the transfer of knowledge from working into long-term memory, the “Ah! I get it now!” moment of clarity. In other words, the actual, y’know, learning.  CLT often posits that we need to reduce extraneous load, so that we can manage the intrinsic and leave space for germane.

This leads me to another key point about CLT: the loads are usually thought to be cumulative.  This well watched (40,000+ views at time of writing, probably 41,000 by the time you read this), admirably clear introductory video to CLT describes (below) and shows (right) it like this: “These [the different types of load] add up in order, meaning that if the demands of the first two are too great, there’s less room for germane load before cognitive overload occurs.”

Simple, right?  This tells us, nice and succinctly, that our job as teachers is to ensure our students think about the right things (the intrinsic load), with minimal distraction (the extraneous load) so that there’s brain space left over for learning (the germane load). You can read a much fuller and better explanation by the redoubtable Adam Boxer, here.

Evidently, a belief in CLT will affect the way you teach.  According to the New South Wales Department of Education, which produces excellently accessible and actionable documents, you should:

In practice, this means take account of what your pupils already know; strip back your explanations, powerpoints and worksheets to what’s essential; make sure you put important things together, rather than putting X on one slide and Y, which relates directly to X, three slides later (because then your class will be trying to remember back to the first slide, which uses up working memory); scaffold, then gradually remove the struts.  Sensible stuff.

A life of its own?

Most certainly.  How do I know?  Here are four examples.

  1. This tweet. It’s gone down in edu-history.

(Sweller is the academic most closely associated with the birth of CLT.  We’ll return to him later.)

2. This result from a quick google:

I confess that I’ve not actually read this blog, but it’s just one example of what you get if you google CLT.  People everywhere are extolling its benefits.

3. This, from Daniel Muijs while still at Ofsted: Developing the education inspection framework: how we used cognitive load theory.  The blog is clear that CLT was not the sole basis for the framework, and links to some criticisms of the theory.  But still, OFSTED using CLT?  Not a bad namecheck.

4. Books like Tom Sherrington’s Rosenshine’s Principles in Action.  The black and red are the racing silks of the evidence informed teaching movement, of which I am a supporter, by the way.  This one is not specifically about CLT (though the Lovell one I mentioned earlier, which comes similarly attired, is), but Rosenshine’s principles are, more or less, the implementation manual.  It’s a good book.  Lots of helpful strategies.  I recommend it.

So what’s the problem?  If all these edu-gurus love CLT, why do I worry about CLT’s popularity?  Here are five reasons.

  • We can’t readily measure cognitive load.​
  • Worked examples, the delivery method most closely associated with CLT, may not work.​
  • Even if they do work, you may not be doing them properly.​
  • The CLT world is ever changing.
  • You may like it for the wrong reasons

I’ll take each in turn.

We can’t readily measure cognitive load

If we are to base our teaching on the amount of cognitive load we are imposing, we need to be able to measure that load. Stands to reason. The problem is, we aren’t very good at doing that.

In this article, Professor Ton de Jong of the University of Twente explains how cognitive load is commonly measured. Daniel Muijs also links to the article in his Ofsted blog, above. De Jong writes:

“The most frequently used self-report scale in educational science was introduced by Paas (1992). The questionnaire consists of one item in which learners indicate their “perceived amount of mental effort” on a nine point rating scale [see below] (Paas 1992, p. 430). In research that uses this measure, reported effort it seen as an index of cognitive load.”

So, students measure their own perceived mental effort in a task.  In Paas’ original scale it ran from “Very, very low” to “Very, very high”, although others prefer a five point scale, others ten, others 100, and still others use different terms.  Some studies into CLT ask this question at the end of the task, others during.  From this perceived mental effort we deduce cognitive load: lots of effort = lots of load.  

I’m no scientist or psychologist, and maybe this represents the best we can do at the moment.  If so, great.  But it’s still a bit less robust than I had imagined.  More to the point, trying to distinguish between “very low” and “very very low”, or “rather high” and “high” is, shall we say, tricky.  It reminds me of mark schemes in History (my subject) which require me to differentiate between “clear” and “very clear” explanations.  Basically impossible, and totally subjective. 

And not only that.  Here, Professor Stewart Martin, Emeritus Professor of Education at Hull University, tells us this:

“A small amount of mental effort could be taken to mean that the learning task generated a small cognitive load (it was an easy task), or that the task was difficult but the learner possessed high expertise (the learner found the task easy because of their high degree of competence), or it could equally indicate that the cognitive load demanded was so high that the learner gave up trying to understand or complete the task.”

I concur.  I can quite see that the reason why a pupil in my class might not expend a great deal of mental effort on my stunning worksheets.  They might be hungry, or desperately fancy the person next to them, or have a football match later, or have noticed that one of my shirt buttons has been accidentally left undone.  Distractions happen.

Even if we can manage those distractions (I check my buttons, and zip, before leaving for lessons, without fail), questionnaires about cognitive load don’t always distinguish between the three types.  That’s a shame, because we need to know how of much there is of each, so we know how to rebalance.  Some surveys have tried, using formulations like this:

  • How difficult was the learning content for you?​ (Trying to get at intrinsic load.)
  • How difficult was it for you to learn with the material?​ (Extraneous.)
  • How much did you concentrate during learning? (Germane.)

I think these are tricky.  Even though I know what they are driving at, I reckon I’d be hard pushed accurately to distinguish between difficulty imposed by what I’m being taught, and how I’m being taught it.

Others have tried to use secondary or additional tests as measures of CL, the idea being that the better people do in those, the better managed the CL has been in the original tests.  Still others have used physiological methods, such as pupillary tracking.  Apparently the latter show most promise, but they’s not exactly accessible to most of us. Further, lest you think I’ve hand picked my academics from some sort of anti-CLT underground, complete with closed Facebook group and secret handshake, how about this from a 2019 article, whose authors include Sweller and Paas themselves: 

“It is assumed that low or high performance on the secondary task are indicative of high and low cognitive load imposed by the primary task [i.e. the higher the load in the primary task, the worse people will do in the secondary].  However, secondary task techniques have been criticised for their intrusiveness (i.e. imposing an extra cognitive load that may interfere with the primary task; Paas et al. 2003) and inability to differentiate between different types of cognitive load.”

And that’s not all.  Here’s Ton de Jong again:

“The literature shows a wide variety of scores on the one-time questionnaire, with a specific score sometimes associated with ‘good’ and sometimes with ‘poor’ performance.  Therefore, it seems as if there is no consistency in what can be called a high (let alone a too high) cognitive load score or a low score.” 

Dammit.  Got any better news for us, Stewart Martin?

“Attempts to obtain direct objective measures of the theory’s central theoretical construct – cognitive load – have proved elusive. This obstacle represents the most significant outstanding challenge for successfully embedding the theoretical and experimental work on cognitive load in empirical data from authentic learning situations.”

Oh.  But let’s leave the final words of this segment to the late Roxana Moreno, of the University of New Mexico:

“The conclusion is clear: there are no standard, reliable, and valid measures for the main constructs of the theory.”

Ka-boom.

Worked examples may not work

Worked examples are often touted as a good way of putting CLT principles into classroom practice.  The idea is that by showing your pupils exactly what to do, you reduce the CL.  Unlike discovery learning, in which pupils try to work out answers with less guidance and which, so the theory goes, imposes greater demands on working memory, worked examples allow laser focus on the right way to do things.  Then you gradually reduce the scaffolding until your class is happily working away with minimal guidance.

You’ve almost certainly used worked examples yourself, any time you’ve demonstrated something on the board, via a visualiser or on a worksheet, or by deploying the “I do, we do, you do” technique.

Sadly, worked examples may not be as effective as we’d hope or even expect, since it does seem to make a lot of intuitive sense.  Moreno (2006) pointed out that the research is somewhat inconclusive:

“Reisslein et al.’s (2006) results contradict the findings of Renkl et al. (2002, 2004)…​ Catrambone and Yuasa’s (2006) results fail to replicate Atkinson, Renkl & Merrill, 2003; Chi et al., 1989; Renkl et al., 1998)…​Finally, Gerjets et al.’s (2006) research contradicts two CLT hypotheses.”

She posits that if we are going to say, with conviction, that worked examples reduce CL and therefore help learning, we need to be able to show this.  But we can’t, because we can’t measure CL very well.

It gets worse.  2018 research at the University of California found that “The transfer of learning…is weakest to problems involving worked examples.”  This means that pupils’ ability to apply a newly learned principle to a range of problems, beyond the precise ones they have studied, is not much enhanced by teaching using worked examples.  Worked examples might make you very proficient at solving the kind of problems you have already encountered, but they may not help you apply that understanding to similar but different situations.  As Moreno says,

“Despite their promise, there is strong evidence that worked examples don’t always work and yet, the cognitive load field is unable to produce reliable explanations for why this is the case.”

That’s a problem.  If we can’t definitively show that worked examples reduce CL, nor that they help pupils do very much more than solve the types of problems they’ve previously worked on, then worked examples may not work very well at all, however logically they flow from CLT.

Even if worked examples do work, you may not be doing them properly.​

In that webchat between Lovell and Tom Sherrington, Lovell explains that while every teacher uses worked examples in some way (see above), most teachers model a couple of examples on the board, check for understanding, ask for questions then give some additional practice problems.  However, that’s not what CLT requires of worked examples.  Lovell says:

“The key difference between [CLT’s requirements] and what we usually do is the level of the persistence of the instruction and of the scaffolding and of the support of the students…We often assume that after the modelling [i.e. showing them how to do it] the students have got it, but really that’s just the start.  They still need a lot more scaffolding, and persistence in that scaffolding, before they get to the truly independent problem solving or question answering.”

So you can’t just show them a couple of times and say, “Great, I’ve done worked examples, so according to CLT, all is well.” It’s not.  You need to do loads, loads more.  Not “I do, we do, you do,” so much as “I do, you do, I do, you do, I do, you do,” and repeat to fade.

The CLT world is ever changing

“Germane cognitive load does not constitute an independent source of cognitive load. It merely refers to the working memory resources available to deal with the element interactivity associated with intrinsic cognitive load.”

Not my words.  Not Ton de Jong, Stewart Martin or even Roxana Moreno.  These are the words of John Sweller himself, the OG of CLT, the man whose papers provided the impetus for the whole damn lot.  And this isn’t recent: the quotation is from a 2010 paper.

Hold the phone!  Here we have Sweller, da man, telling us that germane isn’t germane. 

It’s taken others a while to cotton on, but in 2019 Paul Kirschner, an eminent fellow traveller, tweeted thus:

(John is, I assume, Sweller.  Kalyuga is another academic in the field.)

Unfortunately, most people don’t seem to have kept up. In 2017 the aforementioned NSW Centre for Education Statistics and Education went into print about CLT.  I’ve annotated part of their document.

That’s not all.  Remember the video I mentioned, the one with 40,000+ views?   Here are its first 15 words, with my comments.

“Cognitive load comes in three types: intrinsic, extraneous and germane. [No it doesn’t.]  These add up in order. [We don’t know that, and even if they do, as germane isn’t a load, presumably it can’t be included].” 

Admittedly both these were pre-Kirschner’s tweet, but well after Sweller’s article.  Even so, it’s not clear that even these luminaries are correct.  Here’s edublogger and writer Michael Pershan:

“Even as Sweller has moved away from germane load, many other researchers operating within the CLT framework continue to use the concept…Whatever problems Sweller now sees with the notion of germane load, others prominent within the field do not share his concerns.”

So even if you are keeping up – in fact, particularly if you are keeping up – how do you know what CLT actually says any more?  Of course, in one sense this is just evidence of healthy debate, a theory being usefully tested and hardened.  But it would seem to be a work in progress. Maybe – and I confess to speculating here – this is why Dylan Wiliam appears to have moderated his position:

Still an endorsement, and not necessarily a contradiction his earlier view that CLT is “the single most important thing for teachers to know,” but perhaps rather less unequivocal.

So you need to stay up to date, otherwise you’ll still be banging on about germane cognitive load.  Although maybe you should be.  I don’t know. 

You may like CLT for the wrong reasons

Way back in 2008 the great Daniel Willingham, of whom I am an enormous fan (ask any of my classes), made a video called Learning Styles Don’t Exist.  I’m sure that somewhere he once apologised for its “garage band quality”, which was rather endearing, but the information is good and, as it happens, I buy it.  Of particular interest here is that Willingham himself says that the learning styles theory, specifically the visual/auditory/kinaesthetic variety, “seems to make a lot of sense.”  He even asks, “Why does it seem so right?” He gives three reasons:

  1. Because everyone believes it;
  2. Because something close to the theory is right;
  3. Because of confirmation bias (though he doesn’t use that term).

We can apply much of this to CLT. 

  • Everyone believes it?  I refer you to the “Life of its own” section, above.  All those red and black books.  Wiliam, Sherrington, Lovell and many other heavy edu-hitters.  That’s not to say it’s wrong – just that lots of people, if not exactly everyone, believes it.
  • Something close to the theory is right?  Here we get a little more subjective.  It seems to me that some of CLT’s proposals, or at least the teaching strategies based on them, probably are right.  Overcrowded powerpoint slides, oh-so-hilarious worksheets, teacher as entertainer – all of these would seem to make concentrating on learning more difficult.  It may even be that they are sensible because they derive from CLT.  But, as we have seen, it may not. We just don’t know.  For what it’s worth, I reckon there’s something in it.  Enough to power my whole pedagogical approach, though?  Not sure.
  • Confirmation bias?  Oh yes.  Your students wrote rubbish essays?  You probably overloaded their working memory with new information.  They can’t factorise?  I expect you didn’t do enough worked examples.  They forgot to bring their swimming kit?  Well, that was one of four messages you gave out in form time yesterday.  

Alternatively, perhaps you explained the essay really badly.  Maybe, during your factorising lesson, they were all thinking about hockey later on.  Perhaps they were hungry in form time, or focused on the latest gossip.  The CLT explanation might feel right, but is it?  How can you know?

Willingham’s video makes a similar point.  He imagines a teacher trying to explain the structure of the atom, but “it’s not really clicking.  Finally, you say, ‘Picture the solar system. The nucleus of the atom is like the sun, and the electrons are like the planets spinning round it.’ The student understands and you think, ‘Aha!  The student must be a visual learner.’  But maybe that was just a good analogy that would have helped any student, or maybe the student needed just one more example for the idea to click.  Why the student understood at that point is actually ambiguous.” (My emphasis.)

The same could be said of CLT.  You’ve read about it, maybe been to talk or a webinar.  So you strip down your explanations, use worked examples, rid your slides of diversions and – ta-dah! – your students ace the test.  Cause and effect?  Maybe.  Probably your teaching, free of frills and frippery, was clearer and sharper, and thus more likely to produce the desired results.  But can you be sure that was due to your adherence to CLT?  Or was it just that you thought harder about how you were going to explain things?  As Willingham says, “if you already believe, ambiguous situations are interpreted as consistent.” 

Conclusion

Let me be clear.  I am not out to get CLT.  I like Rosenshine, my powerpoints are clean, my teaching proceeds in small steps.  If I do tell the odd joke it’s because I’m hilarious, and I understand the need to get back to the serious stuff, sharpish. 

What I am trying to do, though, is suggest that the foundations of CLT may not be completely sound, and that the way it’s implemented may not be completely in line with its recommendations.  That, I think, is important to know if you plan to base your teaching on CLT. 

If CLT is the way you want to go, don’t let my ramblings put you off.  But the above may give you some pause for thought.  Not too much, though.  You might suffer some cognitive overload. 

2 thoughts on “Cognitive Load Theory: a life of its own

Leave a Reply

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

WordPress.com Logo

You are commenting using your WordPress.com 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