The Idiot Who Talked About Discovery Learning at Interview

Time to bring Moderately Guided Instruction in from the cold

A few years ago, I had a job interview.

I’d left a centuries-old, top-performing private school. It had no behaviour policy. Nobody told the teachers what to do in their lesson. The corridors were a boisterous jumble of bodies and there were so many different sports happening at breaktime that walking across the playground was a health risk.

Although I didn’t know it at the time, the school I was interviewing at was an imitator of the Michaela Community School in London, dubbed ‘the strictest school in the country’. All I noticed then was the eerie silence as students moved between lessons and toilets without doors – to make sure nothing untoward was happening inside.

At the interview, the deputy head asked me to describe my ideal lesson.

I talked about an A level physics lesson in which I used a worksheet to lead students through a complex derivation. Most of the students were able to get there on their own, using the hints on the sheet – a considerable achievement in my book, I told him.

He looked up at me. Then asked me to discuss my approach to teaching.

I talked about exciting children’s curiosity. Using what they knew already as a bridge to discovering something new.

‘Would you say you endorse discovery learning?’ he asked, scribbling away on the sheet in front of him.

The phone call I got later that evening wasn’t a surprise: ‘we don’t do discovery learning here,’ he told me.

I later found out from a friend who was doing a teacher training placement at the school that I’d become infamous. ‘So you were the idiot who talked about discovery learning at the interview!’ she told me.

Is enquiry learning malpractice?

Greg Ashman recently published a piece arguing that there are times enquiry learning can be malpractice. In the comments section, a reader criticised an academic paper Ashman hadn’t referred to. Ashman defended the paper’s argument: if adding guidance to enquiry learning strengthens it, why not go all the way to full guidance – direct or explicit instruction?

Four reasons, I’m going to argue.

One: we can do direct instruction. Doing ‘moderately guided instruction’ (more on what that actually is later) does not preclude actually explaining things to students together with everything else that goes alongside DI.

Two: direct instruction isn’t as far as we can go. There’s pressure on teachers to go further still and to strip out the judgement and flexibility that good DI depends on. I’ll call that end point ‘total instruction’.

Three: there’s no decisive evidence that direct instruction beats moderately guided instruction.

Four: it’s an odd question to ask anyway. Imagine researchers asking which technique the most effective parents use to bring up their children – direct instruction or discovery learning? The fact we’re asking it tells us something about how we see the role of the teacher.

Before that, though, what was that paper the reader referred to?

Meet the academic paper that cost me a job

That interview was my first contact with the phrase ‘discovery learning’. The fact it happened in a Michaela-inspired school is no coincidence.

The Michaela Community School was ‘surely’ motivated by the 2006 work of Kirschner, Sweller and Clark, claimed schools minister Nick Gibb in a speech.

It seems an obscure thing to mention. The Michaela Community School is still going strong, but Nick Gibb is long gone and this academic paper is now 20 years old.

But the legacy of both the paper and schools minister remain. Kirschner, Sweller and Clark (2006) – KSC – is one of the most cited education papers of the century. It’s been repeatedly cited by government ministers, too. Education secretary Michael Gove claimed:

The most impressive scientific evidence on how children learn - from experts like Paul Kirschner, Richard E Clark and John Sweller - all points towards the importance of direct instruction. Their work on ‘why minimally guided teaching techniques do not work’ is hugely powerful.

More anecdotally, it’s the only academic paper I’ve seen referred to directly during school professional development – and it’s come up more than once.

A steep learning curve

Education researchers have gone over the KSC paper with a fine-tooth comb. There was a whole issue of Educational Psychologist dedicated to rebuttals, in fact (KSC replies are here).

I’m not going to do that here. I’m going to acknowledge the positive impact, then point out what we’ve lost as a result.

First, the good. When I started teaching fifteen years ago, I thought that group work and tasks with little guidance were signs of a forward-thinking teacher. I thought regular social interaction would guide my students through Vygotsky’s Zone of Proximal Development. I left students to play around with practical equipment, then got frustrated when they were confused or made little progress.

I don’t do that now, obviously, because experience has shown me it doesn’t work. More importantly, trainee teachers don’t do it either. Students learn more and teachers have a better experience as a result.

But we’ve lost something along the way.

Two things, in fact.

First, those tasks I’d dreamed up might have worked, with better planning, guidance and management. This is ‘moderately guided instruction’, and we’ll return to it below.

Second, teachers are losing the freedom and desire even to experiment with different methods of instruction. If students are not used to having responsibility and are more comfortable being passive, as a teacher there’s a lot less friction in simply embracing this.

I attended a masterclass by Tom Sherrington at researchED on Saturday about how to run paired talk.¹ Since covid, he said, student talk is dying out in classrooms. Teachers struggle to manage class discussion. It’s easier not to bother.

He wasn’t suggesting a return to discovery learning. He wasn’t saying we need to let students discover scientific laws themselves. He was saying that teachers are afraid now of using well-controlled turn and talk, something that can be a central part of properly executed direct instruction.

But, to me, it’s a sign of how far the influence of KSC has spread as much as it is the effects of covid.

We’ve created a new form of teaching: total instruction.

The KSC dilemma

In the paper, Kirschner, Sweller and Clark present teaching as if it’s this:

Some of the rebuttals point out we should see it more like this:

But then, as Ashman pointed out, if you think going from the left to the middle is a positive step, why not go all the way?

At this point, I’m going to move into conjecture.

Good explicit instruction can involve peer discussion, open questioning and student independence on tasks once they have the requisite knowledge to tackle these with a reasonable expectation of success.

But I’m willing to make a bet – go into classrooms and see how often these things occur. They all involve danger, risk, teacher judgement. Much easier to cut them back and stick with the false stereotype of direct instruction – that it’s simply the teacher talking and students following instructions.

This is not the fault of KSC and it is not what they argue for, but I see it as a sign that their argument has been twisted to suit a particular position: teachers have knowledge and must pass this on didactically by talking; teacher judgement should be removed and students are passive recipients of information. This degrades DI which, at its most effective, relies on exactly the judgement being removed.

This isn’t an enquiry vs DI argument. You can be the firmest believer in explicit teaching and still think the teacher should be trusted to decide how to use it. Total instruction has no defenders – only a system that finds teacher judgement inconvenient.

Haili Hughes, also speaking at researchED, discussed this kind of move – stripping out what makes an approach effective to leave something impoverished in its place. Following Ed Haertel, she termed it a ‘lethal mutation.’

What are all these things?

First, a couple of definitions. Part of the problem with the KSC paper is they don’t give that many, so I’ve tracked down some given elsewhere:

Constructivism: The idea that everyone constructs ideas, theories and ‘mental schema’ for themselves. You can’t hand someone knowledge like you hand them a restaurant bill. This is how the brain works – and nobody in the debate disputes it.

Direct/explicit instruction: The teacher gives ‘full’ guidance (can be via demonstration, video, simulations etc); can include class discussions and activities. For problems and tasks, step-by-step instructions are given, and the guidance withdrawn slowly as students become more competent.

Unguided/minimal guidance: Students are expected to ‘discover’ some or all of the key ideas they are supposed to learn themselves, e.g. ‘students receiving partial instructional guidance may be given a new type of problem and asked to brainstorm possible solutions in small groups without prompts or hints’.

Moderately guided instruction: ‘The student receives problems to solve but the teacher also provides hints, direction, coaching, feedback, and/or modeling to keep the student on track’. Also called, in the literature, guided discovery, enhanced or assisted discovery, and scaffolded enquiry.

The key isn’t the activity but the leap: the student is scaffolded toward something they haven’t been walked through first, and this is what separates it from well-conducted explicit teaching.

The student needs enough prior knowledge for the leap to be reasonable, and that’s where a teacher’s skill, experience and judgement come in. The novice/expert distinction is fine, but we shouldn’t use it as a proxy for ‘not old enough yet’. A seven-year-old can be ready for a cognitive jump in one context, while, in another, a graduate might not be.

What does the evidence say?

Let’s start with the older evidence, that was around when KSC were writing. Mayer writes ‘there is little reason to believe that pure discovery will somehow work today’ then summarises his view of the evidence:

In many ways, guided discovery appears to offer the best method for promoting constructivist learning. The challenge of teaching by guided discovery is to know how much and what kind of guidance to provide and to know how to specify the desired outcome of learning. In some cases, direct instruction can promote the cognitive processing needed for constructivist learning, but in others, some mixture of guidance and exploration is needed.

Mayer concedes that ‘in some cases’ direct instruction is needed but in others ‘some mixture of guidance and exploration’ is needed.

Why draw on Mayer? Because it’s a key piece of evidence KSC use to argue that discovery learning doesn’t work.

Let’s move onto more recent studies.

Meta-analyses gather together a group of studies. Inevitably, some of the studies will be stronger or more relevant than others – this is always a risk with meta-analyses. Overall, they’re a useful way of summarising a large body of evidence. A 2011 meta-analysis of 164 studies first compared explicit instruction to unassisted discovery. In EEF terms, the finding was 5 months’ more progress in favour of explicit instruction. Second, though, they compared enhanced and/or assisted discovery to other forms of instruction including – but not limited to – explicit instruction. ‘Assisted discovery’ – a form of moderately guided instruction – came out on top, giving 4 months’ extra progress.

Another meta-analysis found that guidance reliably improves enquiry learning, but there’s no single correct ‘dose’, and ‘too much guidance inevitably challenges the inherent nature of the inquiry process’.

Other evidence is mixed. Large-scale international data (PISA) has been used to argue for or against highly teacher-led instruction. It also rests on students’ own reports of their lessons, so it’s hard to use it as definitive support for a particular method of instruction.

There’s also the EEF’s RCT on project-based learning published in 2016 involving 4,000 students across 24 schools. The drop-out rate was so high that the EEF themselves gave it a ‘security’ (reliability) rating of only one padlock out of five. Still, they concluded that while the project had no clear impact on literacy (and may have been detrimental):

We found that PBL was considered to be worthwhile and may enhance pupils’ skills including oracy, communication, teamwork, and self-directed study skills.

At best, a responsible commentator would conclude three things:

First, ‘pure discovery learning’ doesn’t work. But then, nobody is defending pure discovery learning. Decades ago, there were genuine calls for minimal instruction; there aren’t now. I’ve spent fifteen years in schools and half a dozen in university departments of education, and I’ve never heard anyone suggest we let kids loose to educate themselves. Even professors of education have met enough children to know they wouldn’t all race to the library the moment you shouted ‘school’s cancelled!’

What we should be debating is how much guidance enquiry needs – not whether to abandon all guidance.

Second, robustly comparing moderately guided instruction with direct instruction is hard – almost impossibly so, given the degree of teacher judgement both involve. An RCT comparing two groups’ grades also can’t capture most of what we want from education: long-term retention, transfer to novel problems, motivation, oracy, self-regulation, etc.

Third, and most fundamentally, the question assumes there’s a single best method waiting to be found. There isn’t – there are as many ways of teaching as there are teachers. The real question isn’t ‘which method?’ but how much leeway we give teachers to decide.

To see how odd the original question is, look at another place children learn: the home.

The point of parenting

Nobody is asking which instructional methods the most effective parents are using at home so that we can get less effective parents to follow their lead.

Let’s take two scenarios: a) teaching about traffic safety; b) how to use a climbing frame. Both involve physical safety, but a parent could teach their child very effectively in both situations and be justified in going to each end of the spectrum – allowing no autonomy and complete explanation about traffic, whereas allowing full autonomy on the climbing frame:

Even when we think about a knowledge-focused task like supporting children with their homework, a parent might choose a point along the spectrum depending on any number of factors like the child, their mood, the importance of the task, or how much they’re struggling with it.

Ultimately, the goal of parenting is to equip children for life. They need increasingly challenging situations. We need to understand their capabilities and build up their autonomy. We need to let them know there’s a safety net. But not to the extent they immediately let go and fall back into it without putting in the effort first.

Which parents fill their kid’s day with highly structured activities, charts, timetables and rules?

It’s not the best parents but the most worried.

And at the moment, it isn’t the teachers who are worried – education systems are worried on their behalf.

Why settle for anything less than total instruction?

I’m not arguing that direct instruction causes learned helplessness. I don’t have the data, and I don’t believe it anyway. What I do believe (but still don’t have the data for) is that only using direct instruction in every lesson is unlikely to equip students well for ambiguous situations, ones in which they have more freedom but less clear instructions or outcomes.

Coincidentally, this is also the point curiosity comes in. To be curious – to pursue curiosity – you need to be comfortable with uncertainty. You need to sit with the unknown and acknowledge your own ignorance.

That’s a discussion for another day.

To return to the central question: which is better – direct instruction or moderately guided instruction? The evidence is, at best, mixed, probably because there is no single way of teaching that is universally effective for all the outcomes we want for our students.

As with parenting, the obvious answer is that teachers should do a mixture of both. But this depends on teacher judgement and that scares some people.

As I found out when I began teaching, moderately guided instruction can be difficult to do. It requires agency, skill, understanding of the needs of the class and the individuals within it.

We could train teachers in this. We could give them scenarios and let them watch skilled practitioners at work. We could let them try it – and maybe fail, as I did – then give them feedback.

But it’s easier to tell them any form of instruction that isn’t maximally guided doesn’t work.

So the question has become the inverse of Ashman’s challenge.

Why settle for anything less than maximally guided ‘total instruction’?

First, because we value teacher judgement. Teachers should have a say in how much guidance is the right amount of guidance, not just policymakers, especially in the absence of decisive evidence to the contrary.

Second, the case for guidance is strong; the case for stripping out judgement is non-existent. Moderately guided instruction can be effective. Direct instruction can be effective. Total instruction is a lethal mutation – the end point of a system that has stopped trusting its teachers.

There’s an alternative: help teachers to develop the skills needed to judge when to use each type of instruction.

If we don’t train teachers in this, is it a surprise they do it badly? It’s a vicious cycle. We should be asking why we aren’t trying to break it.

England’s Department for Education is about to begin an AI tutoring trial, to complement rather than replace teachers.

Let’s see if they’re still saying that when the results of the pilot roll in.

Of all the times in the history of education, this is the most dangerous to be systematically deskilling teachers.

1. Sherrington makes the fuller case that teacher-led instruction and student-centred learning aren't opposites here. Recognising this, he argues, can help strengthen the curriculum and improve students’ learning experiences.