Mixed Ability, Sets, and Streams – a teacher’s perspective – Part 3

I’ve taught sets, mixed ability, and streams.

What follows isn’t any rigorous analysis, or appeal to research.  From what I’m aware of the research, the conclusions aren’t exactly conclusive: lower attainers benefit from mixed groupings, higher attainers suffer.  Mark McCourt reiterates this point, and takes it further by pointing out that those conclusions aren’t necessarily subject or key stage specific, while Lucy Rycroft-Smith suggested there was broadly no impact either way (MrBartonMaths Interview, roughly 52 min in.  Research Espresso.)

So, this has nothing to do with research, just my thoughts and feelings having had some experience of all three.

I’ve split it into four parts:

  1. Setting
  2. Mixed Ability
  3. Streaming
  4. Conclusion


This is Part 3 – Streaming



The whole thing left me looking forward desperately to…


Once in Year 10, the classes that pupils at KSA have been in for three years are finally shaken up.

They are placed into new ‘streams,’ with roughly 30 in the top stream, 20 in the middle stream, and a 14 in the bottom stream.

These streams operate largely like the mixed ability classes from their earlier year – this is their form group, and they are in these classes for all of their core subjects.

That means: maths, science, English, and MFL.

Optional subjects like art, P.E., music and the humanities, were taught in different groups.

The streams still have names; they are not numbered, and they are never referred to as ‘top’ or ‘bottom’ set by any of the teachers.


This, for me, was the best option, in the best of all possible worlds.


Most all of the issues with setting didn’t manifest.  True, in a school so small the pupils were savvy enough to realise that there was a ‘top/middle/bottom’ group, and figure out which one they were in; but it wasn’t emphasised and all-encompassing in the way that it was in my first school.


Pupils were placed in these sets based on their performance in maths, English and science exams at the end of Year 9, but upon taking the exams, this was never mentioned by teachers – there was never any threat that pupils had better revise hard ‘lest they find themselves in the dreadful bottom set!’

The party line was always that groups were chosen based on teacher judgement about what would work best for each individual pupil, taking into account many factors; and this was true.  The groupings were heavily informed by the exam results, but not determined by them alone, and two of these groups, representing 75% of the year, followed identical curricula.


There were still never any ‘set changes,’ and while some arrogance could creep into the top stream, and some disillusionment in the bottom, by and large the cultural effects were far weaker than I’d seen with sets.  I often saw pupils who were very low attaining, in the bottom stream, work as hard as anyone in any other group – they were as determined as anyone to be successful.  They didn’t see themselves as ‘dumb,’ and ‘not very smart.’


Posted in Uncategorized | 2 Comments

Mixed Ability, Sets, and Streams – a teacher’s perspective – Part 2

I’ve taught sets, mixed ability, and streams.

What follows isn’t any rigorous analysis, or appeal to research.  From what I’m aware of the research, the conclusions aren’t exactly conclusive: lower attainers benefit from mixed groupings, higher attainers suffer.  Mark McCourt reiterates this point, and takes it further by pointing out that those conclusions aren’t necessarily subject or key stage specific, while Lucy Rycroft-Smith suggested there was broadly no impact either way (MrBartonMaths Interview, roughly 52 min in.  Research Espresso.)

So, this has nothing to do with research, just my thoughts and feelings having had some experience of all three.

I’ve split it into four parts:

  1. Setting
  2. Mixed Ability
  3. Streaming
  4. Conclusion


This is Part 2 – Mixed Ability



…so, yeah, sets sucked in a lot of ways.

Mixed Ability

But fortunately KSA had mixed ability groupings for all subjects in Years 7 to 9.  Children joined the school, were put into their form of 20-22 pupils, and took every single lesson with that group.

I taught Year 9, all of Year 9 – three groups of ~20 children each.

At first, I adored mixed ability teaching.

Literally every problem I listed above had evaporated.


There were absolutely no ‘set’ changes… indeed, it would hardly make sense.

You couldn’t even move classes to get away from a pupil, or vice versa.  If I teach every group in Year 9, then moving classes isn’t going to make a difference!  And I can think of one or two cases where I certainly would have fallen into the trap of begging for the pupil to move to another class, had I been in my previous school.

In one of those cases, we went on to have a hugely positive relationship after three years, and I’m deeply proud of that pupil’s achievements.

The other case is less of a resounding success story, but we struggled through, we got along as best we could, things did improve, and we certainly did better than we would have in a school where we could have ‘gotten away from each other.’


Each form has a name, not a number denoting their ability.  Gone the self-concept based on the set you’re in.


Gone, the horrible language of low-expectation.




After a few months, the darker side of mixed ability teaching started to rear up.

I often felt trapped.  I had super-smart wannabe mathmos in a group with pupils who couldn’t add and subtract negative numbers accurately.

The top end asked deeply insightful and interesting questions, but responding to them meant that 2/3 or more were shut out of the class for the next few minutes.

By my standards, the bottom end failed, failed, and failed almost every lesson, and I couldn’t go slowly and deliberately enough to cater to them without boring, to the point of insurrection, a different 2/3 of the class.


Across a year, I was deeply confused, and if my preferences were like a needle in a metre, pointing to either ‘sets’ or ‘mixed,’ it was swinging wildly back and forth, month after month.


Sets, or Mixed Ability?


Even though my teaching, and the outcomes, were undeniably better than they had been in the previous two years – thanks to the excellent discipline systems at KSA, its outstanding culture of high aspirations, and being a teacher in my third year – by the end of the year I was deeply dissatisfied with the experience.

I felt that I couldn’t teach the top end to the limits of their potential, and I was consistently failing the bottom end, while the middle sorta chugged along relatively unnoticed.

And these were classes of 20… I could hardly imagine what it would be like with traditional classes of 30!


The whole thing left me looking forward desperately to…


Posted in Uncategorized | 2 Comments

Atomisation – Compound Shapes

Recently I used the word coined by Bruno Reddy, atomisation, to describe the process of breaking ‘solve two simultaneous equations’ into 13 different sub-tasks.


Even more recently. Ben Gordon used a similar approach to turn ‘find the area of a compound shape’ into 28 different sub-tasks!

I’m struggling to find anything missing… the obvious one that gets overlooked is ‘find unknown lengths between parallel lines,’ but it’s in there.


So, can anyone find anything that’s missing?  Can we split the atom any further, and turn 28 sub-tasks into even more?  (Assuming calculation / arithmetic is already secure.)



Posted in Uncategorized | 1 Comment

Mixed Ability, Sets, and Streams – a teacher’s perspective – Part 1

I’ve taught sets, mixed ability, and streams.

What follows isn’t any rigorous analysis, or appeal to research.  From what I’m aware of the research, the conclusions aren’t exactly conclusive: lower attainers benefit from mixed groupings, higher attainers suffer.  Mark McCourt reiterates this point, and takes it further by pointing out that those conclusions aren’t necessarily subject or key stage specific, while Lucy Rycroft-Smith suggested there was broadly no impact either way (MrBartonMaths Interview, roughly 52 min in.  Research Espresso.)

So, this has nothing to do with research, just my thoughts and feelings having had some experience of all three.

I’ve split it into four parts:

  1. Setting
  2. Mixed Ability
  3. Streaming
  4. Conclusion


This is Part 1 – Setting



In my first school, where I taught for two years, pupils were set, as seems to be most common.  They were in up to 8 ability groups.  Each child knew with group they were in, and classes were named based on that number e.g. 9.7 for Year 9 Set 7.

There were many things I disliked about this system.


First, language.

I found all teachers used the language of sets when talking to pupils.  For example:


“I expect better behaviour from a top set!”

Thus implying that we don’t expect better behaviour from every other class…

“Maybe if you work hard, you’ll be able to move up a set…”

…up a set where you’ll finally learn something, because no-one learns in this class.

I was not above this.  In clueless moments of desperation I have uttered these words and hated myself while saying them.  I think the language that I’ve seen sets engender in teachers probably sums up all their worst features. But then…


Then, set changes.

Every time a child moves sets, information is destroyed.

I found I could teach the pupils I knew best, best, since I knew what they knew, knew what we’d discussed, and could draw on historic experiences as prompts, or build relationships between knowledge.

 I deeply disliked it when a new pupil joined my class, and wasn’t thrilled when one left.  This wasn’t just caused by human bias against change, and the emotional effort of forming new relationships, it was also, arguably mostly, because I wasn’t confident that I could be a better teacher to them than the previous person who knew them better.

If these changes happened rarely, I might consider them manageable.  In my experience in this school, with the exception of the ‘protected’ top set, the churn was incessant, and the consequences dire.  Considering our relatively poor ability to accurately measure performance, never mind learning, its doubtful that these set changes were truly meaningful, or helpful.

See this slide by Dylan Wiliam for a glorious example of what I mean.


Test reliability 0.9, predictive validity 0.7, 100 students, 50% in ‘correct’ set

FYI – 0.9 reliability is like, stupidly high compared to what you can expect from most school/teacher set assessments (which I think from memory is typically closer to 0.7, but shout out if you know better and I have that wrong.)


The protected top set.

There was an undercurrent of belief throughout my school that the top set were the only people who would truly learn.  Everyone else was grist was the C grade mill.  Once they banked it, they would be chucked out of the target groups and into the second set, where they were expected to more or less languish… maybe pick up a B if they were lucky, but you know, no biggie.


Child trading.

A sub-set of set changes – if a pupil and teacher don’t get along, it was generally understood that one or the other could petition the head of department to have the pupil moved up or down a set, so they didn’t have to be together.

Again, I was not immune to this.  I would be more than happy to advocate that maybe a belligerent should move up a set, if I dreaded seeing them each day.


Finally, obviously, self-concept and stereotype threat.

Best demonstrated by this 9 second clip from Tough Young Teachers


“Bottom set, what does that mean to you?”

“Dumb!  We’re dumb.”

“Not very smart.”

“So, is that what you think of yourselves…?”



Pretty typical of students in the bottom set.  But then actually, pretty typical of almost all students who aren’t in the top set, I found!

Then you get the kids in the bottom of the top set, who are obviously pretty high achievers, but think they’re the worst, because in their little sphere of experience, they are.

Then you get the kids who do well, think they’re really smart, think everyone else thinks they’re smart, and now don’t want to try for fear of failure and losing that impression – it’s okay to fail if you weren’t even trying, not okay to try and fail and signal that you’re not that smart after all.


So, yeah, sets sucked in a lot of ways.


Posted in Uncategorized | 4 Comments

The world’s most effective learning experience

Create and provide the world’s most effective learning experience, accessible to all.

This is the promise of Up Learn, a growing educational start up driven by the belief that every child can learn whatever we have to teach them, if only we could get the teaching right.  It reminds me of a point I made at the bottom of this post.


Cognitive science forms the bedrock of its instructional approach, and machine learning adaptive algorithms personalise the experience to each student.

I know, everyone’s promising computers will change everything, but I’m writing about this, in part, because having met and supported more edu-entrepreneurs than I can now recall, two things about this one stood out.

The first was just how far its co-founders had come in terms of reviewing material from the literature of cognitive science, applying it to their service, and writing it up into an internal summary that would give Craig Barton a run for his money.

The second was their promise of an A or A* to every student who used their service, or they would return the money of anyone who paid; they were backing themselves to get this right in a way I haven’t seen from others.


For now, Up Learn funds its research by providing an alternative to expensive one-one tuition for students studying A Levels.  Video explanations are scripted in full, and lessons are driven by a careful selection of questions before, during and after, based on principles from cognitive science and other research into educational video design.  Explanations are then recorded and animated by a professional team.

In its first year of operation, 95% of students who took its pilot Economics course achieved an A or A* in their final exam.  Despite this, the team somehow decided the videos weren’t good enough (?!) and so redesigned the course from scratch.  Last year, 100% achieved an A or A*.

To date, over 20,000 students have signed up to the platform, and this forms the other part of the reason I’m writing about this: a couple of weeks ago I joined Up Learn as their Director of Education, and we’re trying to find really, really, really good lesson designers to create content for new A Level subjects, and it’s not easy.  I’m hoping you might know someone who’ll be interested, and point them our way.

You don’t get the day to day hustle and bustle of a lively school environment, but what you do get is the chance to sit for hours on end, uninterrupted, super-deep in thought about the best possible way to sequence and communicate your subject’s content; then, best of all, you get to roll it out to thousands of students and have intelligent deep-learning analytics tell you what’s working, and what isn’t, so you know exactly what needs to be improved.  I love it, but it’s probably not for everyone; this one’s a job that will probably best suit someone you know who massively loves their subject, and spends forever learning more, and thinking about the minutia of how to teach it.  They may even live as a bit of a tortured soul where they see students not getting the top grades as resulting from their failure to teach well enough; if only they had more time to plan…

From all this work, we’re hoping to eventually derive some general principles of instruction that can be applied by teachers of all subjects, and all ages, in all classrooms, up and down the country, so it will probably also suit someone who values scale of impact over proximity to impact.

We’re prioritising chemistry, biology and psychology for now, but would still be keen to hear from teachers and tutors of other subjects.

Please, if you know any great teachers who you think would love to do this, ask them to take a look and get in touch.

Posted in Uncategorized | 4 Comments

Knowledge Organisation

There was all kinds of fuss and frustration expressed by individuals earlier in the year, when several teachers started extolling the virtues of Knowledge Organisers.

A quick Google for this will reveal several examples of these on Google Images.

Most of the criticism was asinine, in some cases seeming to go out of its way to be obtuse.  For example, one criticism I can recall was ‘But pupils need to learn more than this!’ at a time that precisely no-one had claimed otherwise, and most had expressed how they were using knowledge organisers as a tool to develop schema forming (e.g. through self-quizzing key factual knowledge outside of lesson time, so the teacher could focus on fleshing out further knowledge, relationships, interpretations etc. during lessons.)

This is unfortunate, because it is possible to levy real criticism at knowledge organisers.

Another way of putting it would be to question whether knowledge organisers are just the first step in a greater journey of expanding our understanding of knowledge organisation, more broadly.

Organising Structures

Frederick Reif defines three kinds of knowledge organisation:

  1. Associative network
  2. List
  3. Hierarchy

The first is how our mind works.  Concepts are associated with other concepts through some relationship.


Lists are exactly what they sound like.  A list has a heading, and then its sub-points simply continue in length.  If more than one list is presented, the assumption is that there is no real relationship between them:


A hierarchy can be thought of as a series of lists connected by grouping or categorisation.  E.g. Susan is one example of ‘human beings,’ which are collectively one example of ‘mammals,’ which are collectively one example of ‘life forms’…


Reif argues that hierarchies are the most desirable structure – associative networks are difficult to mentally navigate, and lists have too little organising structure, being little more than a single grouping.

These are not the only ways of organising knowledge, a quick glance through the work of Nancy Duarte will quickly reveal that.


Or even just taking a look at MS Smart Art:


Siegfried Engelmann also has a fascinating series of principles around how knowledge should be organised and presented to learners, but that’s a story for another day.


The reasonable criticism that I think can be levelled at Knowledge Organisers, as they currently stand, is that they all seem to be making use of the list structure, and only the list structure.

This doesn’t mean they will be ineffective, not at all, but it does mean they might be less effective than they could be.

By ‘less effective,’ I mean it might take a learner more time to commit the facts to long-term memory, that they might do so with lesser storage strength, and that the relationships between the facts will be weaker, than if an alternative structure were adopted.

Note: Before I go any further, it’s worth noting that I can also see very good reasons for constructing knowledge organisers the way that everyone is constructing them, and I’ll speak to that at the end.


For now, I’m going to use a single example to talk through this, characters in The Tempest.

I finally saw this for the first time last month, and set about learning the plot and characters so that I could follow what was happening on stage.  SparkNotes presents character information as a list, similar to how it would be presented in a conventional Knowledge Organiser:


While reviewing this, I found I had to do a lot of mental work to formulate a mental image of who was whom, and how they related to one another.  I voluntarily undertook that work, but it would be easy for this to be presented to a child who didn’t automatically undertake that effortful work, or didn’t know how to, or struggled to hold and process everything in working memory simultaneously.

In that instance, the pupil would only see a list of unrelated loosely connected names, and wouldn’t develop the mental schema of how they relate to one another – it would resemble the so-called ‘rote learning’ that many teachers viscerally fear.

An alternative would be to present the information like this (click for larger image):


Here, relationships between characters are made explicit.  Size of box and colour show relative importance of the character in the play, and hint at who to read about first.  Ultimately, this is the mental schema we would want pupils to construct in their own mind; by laying it out transparently we guarantee success for everyone.

Note: I had wanted to try to follow Reif’s advice and construct a hierarchy, but for something like this there didn’t seem to be a useful way of grouping the relationships into ‘higher’ and ‘lower’ levels.  Anything I might have chosen – e.g. relationship to one another, importance in play, proximity on stage – all fell apart when trying to bring the groups together.

Pros and Cons

The associative network I drew, above, has some drawbacks.

If presented altogether, it is probably more overwhelming – more likely to induce cognitive overload – than the list; where do you start reading?  What do you read next?  etc.  (I ran into this problem a few times when trying to present several different overviews of all of mathematics to pupils – it was always too much information shared at once, and I didn’t see that they couldn’t possibly navigate it the way that I did.)

This can be overcome by introducing it to pupils in stages – something that Engelmann would do e.g. Start by showing only Prospero’s box, then the Orange Boxes…  However, this restricts its utility as a self-quizzing tool, something that pupils can make use of independently of teachers (indeed, Engelmann’s chapter on knowledge organisation assumes that the information is being presented to pupils by a teacher.)  It also limits its ability to serve as a single sheet of paper, given to pupils at the start of a new sequence of lessons, that presents all the most important facts in a given unit.

It’s a lot more difficult to construct: prone to perfectionism, and consumes much more time than a simple list.

Finally the benefit of the spatial layout is offset by its consumption of space – you can fit far less information into the same space.


As I said right from the start, there are good reasons that Knowledge Organisers are being constructed the way they are, and their utility and ease of construction means that probably shouldn’t be abandoned.

There are alternative structures, though, that could increase the probability of a given child’s speed in committing facts to memory, the storage strength and enduring retrieval strength of those facts, and their knowledge of the relationships between the facts.

It would be interesting to see the results if teachers started to experiment with these structures, in addition to the list.

Posted in Uncategorized | 11 Comments

My best planning. Part 4

Craig Barton interviewed me recently, during which I discussed a series of lessons I planned and taught on solving simultaneous equations.

I could be wrong, but I think this was the best planning and teaching I ever did.

Several people have asked if I would share examples of what I described during the interview, so I’m adding that here. It’s a bit lengthy, but hopefully provides the detail many people were asking for, as well as some insight into how Siegfried Engelmann’s Theory of Instruction can be applied to the classroom.

I’m splitting the post into four parts:

  1. Specification of content
  2. Sequencing of content
  3. Pedagogy / Instructional Approach
  4. Limitations of Atomisation


This is Part 4 – Limitations of Atomisation


The Limits of Atomisation

I promised some commentary on the question of how far we should break concepts into a greater number of smaller and smaller pieces.  I’m borrowing a word that Bruno Reddy suggested to me years ago to describe this: atomisation.

The benefits of atomisation are simple: it increases the probability of success for each child.

Picture one of the classes you teach.  Now for each of those children, picture them with a probability hovering over them: for any given thing you might strive to teach them, selected at random, in any particular lesson, this is the probability that they will succeed in learning it.

Note: The way I’m using the word ‘learning‘ here, is incorrect, but speaks to our intuition about what’s happening in the classroom.  Learning is a long-term effect resulting in changes in long-term memory.  What I really mean here is the probability that pupils will respond successfully to predetermined questions, for that lesson only, which I would argue is a necessary but not sufficient condition for learning to eventually take place.

Seating Plan 1

The way in which we teach can affect these probabilities, but the general distribution will likely remain – I suspect it is probably not true that ‘some methods work better for some children and less well for others,’ and there is a deeply sinister and insidious consequence of this line of thinking, as well, hinted at towards the end.

If, as Daniel Willingham says, we are more alike in how we learn than we are different, then changing the instructional method likely increases or decreases everyone’s probability of learning successfully, while more or less maintaining the distribution, the landscape of probabilities.

Important Note: Sometimes we think that ‘one way of teaching is better for some children than others’ because we switch to a different explanation, analogy, instructional method, and find that, when we do, a given child ‘finally gets it.’  We conclude that we had finally hit upon ‘the successful method’ for that particular child, but Willingham argues that it is more likely that the child simply needed more time, and trying different things gave them the time they needed to process the idea, or that they may have simply needed more examples / analogies, in other words, the eventual success was not caused by the particular example we gave them last, but by the cumulative effect of having given them three examples; whichever example we started with would still have resulted in failure.  This is important because it’s this kind of reasoning that led us into the traps of ‘VAK learning styles’ and ‘left-brain / right-brain dominance,’ ideas that seek to categorise and therefore limit what we believe people to be capable of.

In this model, I am suggesting that atomisation raises the probability of each child being successful:

Seating Plan 2

Suddenly, a class that seemed to have just a few ‘super smart’ kids in it now looks as though it has a whole bunch of them, with only a narrow gap in ‘ability’ for most.

This increased success likely results for several reasons that cognitive science can explain; I won’t go into them all here, but a simple one would involve the way atomisation helps us to avoid overloading Working Memory.

But there is no such thing as a free lunch.

The limitation: With increased atomisation, comes an increase in time needed to cover the content.

There were people in the class who previously had a very high chance of learning the content before atomisation, so what happens to them, do they lose out as more time is spent on the same topic?

What happens if we take this to an extreme, and break things down into the smallest components possible; what was once treated as a single idea, is now treated as a hundred – in this case is it possible that the time needed to cover everything in such minute detail would result in a diminished return?

The solution

…is not a simple one, but does exist.

As with most things in life, it’s a balancing act.

First, we’re generally so bad at this (speaking for myself,) and our standard textbooks tend to be equally bad at it – in other words, at the moment, atomising more will probably lead huge gains for most children, in most circumstances, so I would judge that there’s little risk in you striving to apply it.

Second, yes, it leads to an increase in time to teach initially, but it also results in the guaranteed initial apprehension of concepts that otherwise would have had a very low probability of being communicated successfully.  This means the increase in time spent at the start is rewarded by increased probability of learning future content, thus spending time now, in order to save time in the future (a return on investment.)

Time for Mastery

Third, atomisation reveals concepts that are otherwise implicit, and overlooked in the curriculum.  For example, we played with adding and subtracting three and four equations, and looked at adding equations without any intention of eliminating a variable, ideas that are often overlooked if the ‘process of solving simultaneous equations’ is simply taught from beginning to end.  As a result, even the ‘higher attainers’ are learning more than they would otherwise (I spoke to this in the podcast.)

Finally, there is still a balance to be had.   Too much of anything is bad, by definition, and too much atomisation is probably possible.  I wonder whether the appropriate balance shifts from pupil to pupil, and this in turn leads me to wonder what role streams (a potentially better version of setting) and differentiating by time might play.  Perhaps a top stream would experience less atomisation compared with a lower stream, but the lower stream would be gifted more time with their teacher to mitigate against the time cost.


The problem with traditional teaching

Engelmann has a way with words.  Much as I find labelling unhelpful, what most people refer to as ‘progressive’ teaching, for him, is ‘traditional.’  To him, everything that has come before his ideas is ‘traditional.’

It is traditional, because in his mind the traditional position in education is that some kids can learn very well, and others can’t.

In stark contrast with this, Engelmann believes that if you get the teaching right, all children will be successful.  The diagrams with the percentages hopefully speak to the image this conjures in my mind.

Consider the following three classes, with their respective probabilities of success in any given lesson:

Seating Plan 1


Seating Plan 2


Seating Plan 3

Engelmann would argue that the traditional teacher sees three classes, with different pupils.

He sees three classes of the same pupils, with different teaching.

Posted in Uncategorized | 5 Comments