How do we know when other people need to know what we know? On scientific purity, useful fictions and evidence in the classroom

In a great series of recent threads (started here, and subsequent discussions here, here and here, amongst others), Matt Slocombe, from Birkbeck University has been arguing against the ‘politicisation’ of cognitive science evidence in recent education reforms, and looking to build a consensus of researchers who might seek to influence the direction of future decision-making. Specifically, Matt objects to the central importance given to Cognitive Load Theory (CLT), and the consequent absence of evidence from other sources within cognitive or developmental psychology, in new teaching frameworks, such as the initial teacher training and early career frameworks (you can read more from Matt and colleagues on this here).

I know Matt a bit and am always impressed by his passion, drive and breadth of knowledge. Also, in the way that you can only have really interesting discussions with people with whom you generally concur, I agree with a number of his criticisms. I do, however, have some reservations regarding some of the implications of this push to influence policy.

 

Scientific criticisms of a model are not de facto criticisms of its applied value

In his tweets, Matt lays out some of the major criticisms against CLT as a model of cognition. Others have also been a consistent voice pointing out weaknesses with some of the underlying theoretical assumptions (and research methodology) of CLT, such as Christian Bokhove. I’ve also pointed some of these out in the past, for example being quoted with my PhD supervisor in this TES article that was generally critical of CLT (though I should say that I didn’t agree with the conspiratorial tone that the author eventually took or his interpretation of some of our comments). I’m certainly not arguing, therefore that the CLT model is entirely accurate, or even that it isn’t severely deficient in some respects. The question is, how much does this matter?

Given that no-one involved in this debate would presumably expect teachers to be as informed as researchers in this area, we’d all then accept that teachers’ understanding of cognition will always be based on models – imperfect approximations of the truth. The question, then, is how harmful is it to have a model that is slightly too ‘rough’, (or conversely, how much does it really benefit teacher to have a model that is slightly more accurate)? It is likely that diminishing returns will be at play here – ever increasing scientific knowledge will not lead to ever increasing skill at implementation. Indeed, having just finished a four year PhD and now preparing to go back into the classroom in January, I can attest that there can be a certain level of ‘evidence paralysis’ that can set into planning when you are too aware of all of the drawbacks of every theory going! So where is the line, which will improve outcomes for students in the classroom, without burdening the teacher with too much in the way of niche scientific debate? We don’t know. We don’t have studies on the effects of a widescale adoption of practices informed by a slightly simplistic model of CLT in classrooms. We might in ten years, given the new ITE frameworks, but at present we’re pretty much clueless about the ideal level of sophistication for teachers’ understanding of the science of learning.

As a result of this, while I am perfectly happy for people to criticise aspects of the underlying theory of CLT, and to argue for a slightly more nuanced or broader evidence base to be covered in initial teacher training, I think it is less easy to state with any certainty that the current model of CLT is likely to be harmful for teachers or students, and that it therefore should be changed. We simply don’t know this. It’s also not clear how closely the scientific criticisms affect many of the parts of the theory that teachers actually use. For example Damian Benny, in a reply to Matt here, acknowledges some of the criticisms of the theory, but argues that the valuable applications to the classroom are not really affected by them. Others have also commented that, even if a ‘strong’ reading of the CLT model might be interpreted as suggesting that content knowledge is prioritised entirely above any other skills such as creativity or problem solving (an interpretation that I don’t fully buy into), this isn’t how the theory is being used in classrooms anyway.

CLT, as with all models, is a ‘useful fiction’. How useful, and how fictional, remain to be seen. I can’t therefore accept that it can be dismissed with any certainty.

 

Developmental research is not de facto relevant to education

A related second point concerns the idea that current frameworks are insufficient, because they are limited in the evidence that they cover. In Matt’s original thread he refers to “a rich body of relevant cognitive and developmental evidence being ignored”. Of course, developmental psychology is historic and rich subject, which has revealed many fascinating aspects of human development. I would contend, however, that there is no automatic reason to suppose that developmental findings will be relevant to educational practice, and in fact there is very little evidence which we can currently state with certainty is something that ‘teachers need to know’. Remember here that the debate is not about what we would ideally like the most enthusiastic, informed teacher to know, but rather what it should be necessary for teachers to know, as mandated by training frameworks and government policy. In order to qualify for that high bar (especially given the limited bandwidth available for this in amongst all the other things trainee teachers need to learn), research needs to do two things:

  • Tell teachers something that they are not already aware of from other sources.
  • Provide teachers with clear guidelines for how their practice might be adapted

If these criteria are met, then learning in more detail about the underlying mechanisms and developmental processes is justified. For perfectly good reasons, however, very few (in fact I’d say almost no) developmental studies reach this bar. That’s simply not what the scientific studies are designed to do.

Let’s take an example, based on my own PhD research looking at distraction in adolescents in the classroom. Three years ago I blogged about what we know about attention during adolescence, pulling out two key messages, that adolescent attention is inconsistent, and that it is disproportionately biased during these ages towards interference from motivational or emotional stimuli, and by things such as their peers. There is a lot of ingenious and hugely impressive work by researchers that has explained many of the neuroscientific processes behind these conclusions… but what does this mean for teachers? Do teachers not already know that teenagers are inconsistent, easily distracted (especially by their peers) and emotionally volatile? Hardly. Do teachers need telling that reducing distractions in classrooms is probably a good idea? Hardly. I’d therefore argue that developmental work on adolescent attention fails both of the criteria above. In another nice example of this recently, Nick Rose examined attachment theory – a psychological theory examining relationships between children and their parents or caregivers – and its relationship to teaching practice. He concluded that the main implications of attachment theory for teachers – that they should aim for good relationships with students and that they should pass on safeguarding concerns to people trained to deal with them – is exactly what teachers already do. As a result, the theory is of limited use in terms of its applications for teachers.

Matt has put together a site of cognitive and developmental science research which he thinks might to inform educators and education policy. I think that this is a fabulous idea and have suggested some contributions, but the papers included did rather reinforce my view that there is very little research around there at the moment with direct, clear classroom applications which do not simply confirm what teachers already do. To my mind, this is less to do with politics, and simply a reflection of the chronic lack of translation studies, which take cognitive findings and examine how they can best be applied in real classrooms. In the absence of such findings, we have no way of knowing which research teachers need to know, and which they don’t. If we aren’t happy with the use of CLT in policy, then it’s unfortunately incumbent on the researchers to demonstrate that other ideas are as valuable, if not better. We can’t do this at present.

At the heart of this is a philosophical debate about how ‘good’ evidence has to be for scientists to be happy to put it forward to influence policy and practice. Clearly no piece of evidence is ever perfect, nor any theory beyond reproach, so if we prevaricate for too long searching for absolute certainty then we’ll never influence anything. My personal feeling, however, is that if we as researchers are to start applying our knowledge across disciplines, such as into education, then we need to be pretty damn sure that our suggestions actually work in that new environment, and that they can be easily and successfully implemented into classrooms. I just don’t think we’re there yet.

Which is probably why I’ll never end up influencing anything.

How do we know when other people need to know what we know? On scientific purity, useful fictions and evidence in the classroom

Constructivism is a theory of learning, not a theory of pedagogy. Neuroscience explains why this is important

Unsurprisingly, teachers are very interested in the brain, and excited about the potential for psychology and neuroscience evidence to be used in improving what they do (see for example the responses to this survey). It’s a shame, then, that so little time in many teacher training routes has traditionally been given over to psychology and the science of learning. Especially if you trained five or more years ago (as I did), it seems to have been very rare to have received more than a cursory introduction to the subject… apart from one notable exception. However little psychology gets covered otherwise, one mandatory inclusion on many teacher training courses is constructivism — the idea that every individual constructs their own understanding and knowledge of the world, based on their own unique experiences. As a result, constructivist ideas are undoubtedly the most widely held ‘folk-psychological’ belief about learning amongst current teachers (See Torff, 1999, and Partick & Pintrich, 2001 for more on the content of teacher training and changes in trainee teachers’ beliefs about learning).

In this post I am going to argue that this state of affairs is dangerous. Importantly, though, the problem is not that constructivism is wrong. Constructivism works well as a theory of learning. The problem comes, however, when it is assumed that the theory of learning implies a particular pedagogical approach (‘educational constructivism’, or sometimes the closely related approach of ‘constructionism‘). Using evidence from neuroscience, I will try to show that there is a great deal of support for constructivism the learning theory, and a good deal less for constructivism the pedagogical approach.

Some of the misapplications of constructivism to pedagogy have been well documented before. Phillips (1995) described ‘The good, the bad, and the ugly‘ of constructivist pedagogy, and Hyslop-Margison & Strobel (2007) put a slightly more nuanced slant on the same topic. Much of their focus is on the potential for more radical constructivist views to lead to a relativistic approach to knowledge in which, because knowledge is individually constructed, there is ‘your knowledge’ and ‘my knowledge’ and little scope for any external validation.

My focus here, however, is on the extent to which the operations of the brain, especially during development, can help us to see why constructivism the learning theory does not entail the pedagogy. I’ll use the evidence underpinning the theory of ‘Neuroconstructivism’ (Mareschal et al., 2007), one of the most popular theories of developmental cognitive neuroscience, to help me with this.

Partial representations, and context-specificity

A central feature of neuroconstructivism is that the development of our brains and the storage of information in them is hugely context-dependent. At any one moment, activity in the brain is a reflection of the context that the organism finds itself in. Mareschal et al., provide four different levels where the activity of the brain is constrained by the context in which it occurs: neural, network, bodily and social. We’ll look at examples of context-specificity for each of these in turn in a minute. The important point, though, is that brain activity reflects the precise state of the organism at the time of the event, so any new piece of ‘learning’ will be encoded completely within the context of the learning experience, rather than reflecting any general underlying feature of it. In the terminology of the theory, we are only ever able to create ‘partial representations’; representations of the world which capture some, but not all of it. Partial representations are, by their very nature, completely context-dependent; that is, they reflect the features of the world (and of the brain) which were the case when the information was originally stored. Let’s then look at the four levels which can lead to the creation of such ‘partial’ representations. This section is heavily paraphrased from a previous description of the theory, but I reproduce it here for clarity.

  1. Neural context – ‘encellment

The cellular neighbours of a neuron exert a large influence over its eventual function as a processor of information. The characteristics of its response and the way in which it connects and influences other neurons is in turn dependent on the type and amount of activity that the neuron itself receives. One classic example is that “cells that wire together fire together” — the more that cells communicate with each other, the more that their connections are strengthened, and the greater influence that a preceding cell exerts over the activity of subsequent cell.

However, the context-dependence of neural activity is not limited to the simple co-operative strengthening of connections. They can also compete. Hubel and Wiesel’s Nobel Prize-winning work on vision in cats involved looking at what happened if a part of the brain called the visual cortex was understimulated (because they had covered up the eye which sent information to it). They found that after 2-3 months, the nerve cells in the understimulated area began to switch functions and process information from the uncovered eye instead.

So what?

The activity of any one neuron is context-specific. This context is constantly changing due to a number of different factors: the ever-changing strengths of connections to potentially thousands of other inputs, competition (or co-operation) between neighbouring cells, or a progressive specialisation of the cell’s function. Therefore a signal from a neuron can only be interpreted as representing that cell’s response to a particular set of circumstances at that specific time; the neural context.

Another consequence of the reliance of each neuron’s activity on so many of its neighbours is that this means that any information that is encoded by the neuron is likely to be done so in a distributed fashion, across large groups of neurons. Such ‘distributed representations’, whilst more robust on the face of damage and brain changes, are also far more likely to be ‘partial representations’, relying as they do on numerous small contributions from different neural sources. They will never capture a concept or an idea in its entirety. Instead, they record a blurred snapshot of some of the key details approximating the concept, a partial representation.

 

2. Network context – ‘embrainment’

Just as individual neurons can be affected by the context in which they find themselves, so entire brain areas can co-operate, compete and change function as a result of their context within the brain as a whole. On a larger scale than that noticed by Hubel and Wiesel, Cohen et al (1997) found that in people who have been blind from an early age, ‘visual’ cortex begins to take on other functions entirely, such as touch when reading braille. Similarly, if you re-route visual information into a ferret auditory cortex (an area that would normally deal with processing sounds), the area will begin to respond to visual patterns instead (Sur and Leamey, 2001)! In less drastic fashion, maturation in the brain involves the progressive specialisation of many different brain areas, which gradually take over sole control of functions that previously relied on wider networks. Again this process can be categorised by competition, with one area gradually coming to exert a dominant influence over a particular kind of processing. Good examples of these sorts of processes have been found in the pre-frontal cortex (PFC) during adolescence, with different sections of that brain area becoming ‘responsible’ for particular functions.  (see Dumontheil, 2016 for some examples).

So what?

Most formal education is taking place during periods of rapid brain development and maturation. Brain areas are progressively specialising and refining their functions, dependent on their relationship to other brain areas and input from the outside world. In this context, the distributed and partial representations that we build of the world are likely to be highly context-dependent, not only on the particular pattern of inputs, but also on the time and stage of development in which the information was learned.

 

3. Bodily context – ‘embodiment

The brain does not sit in glorious isolation from the rest of the body. Some hard-wired nervous behaviours, such as reflexes, can in fact form the basis for the beginnings of brain development. Infants make spontaneous reaching movements from an early age and even new-born infants will move their limbs to block a light beam (Van der Meer et al, 1995). These kinds of behaviour initiate the beginnings of feedback mechanisms between the visual and motor areas of the brain and eventually allow for the development of complex visually-guided behaviour. Of equal importance, the design of some parts of the body can constrain brain development by ensuring that it does not need to develop certain skills; cricket ears are designed to respond preferentially to male phonotaxis (a sound made by rubbing one wing against the other). The cricket brain has no such specialisation for making this distinction, because the job has already been done (Thelen et al., 1996). In human cognition, examples of ‘embodiment’ might include state-dependent memory; the finding that we recall information more successfully in a similar ‘state’ to when we learned it, for example after exercise (Miles and Hardman, 1984) or even when drunk (Goodwin et al., 1969).

So what?

The development of our brains is constrained and uniquely differentiated by our nervous systems and by the body in which we find ourselves. Again, this is not just the case between individuals but also within individuals as they develop over time, and as they pass through the myriad different internal states which characterise our existence. The representations that we have of the world will reflect these changing embodiments, and will be ‘partial representations’ in that they are formed, and linked to, this embodied context. This therefore provides further scope for learning to be constrained by the situation (in the widest possible sense) in which the information was initially encountered. The Goodwin et al. paper is particularly relevant here, as it tested two outcomes; recognition and transfer. They found that, whilst recognition memory was not significantly affected by the a change in states between learning and recall, the ability to transfer the information was. Transfer, as a more complicated cognitive procedure than simple recall, is as a result even more susceptible to being restricted by the context in which it occurs.

 

4. Social context – ‘ensocialment

The concept of ensocialment, the idea that the social context for any act of learning is crucial to shaping the learning that takes place, will be the most familiar of these four levels of analysis to educators. Vygotsky’s social constructivist theories are probably the most famous educational application of this sort of idea. People learn from others with more skills than them; with the more knowledgeable mentors using language and guidance to ‘scaffold’ the learner’s interactions with the world in the most productive manner. The concept of scaffolding; a supportive structure which is gradually removed as the learner gains in ability, is used to one degree or another by almost every major educational approach.

So what?

The type of scaffolding that is used may become inextricably linked to the solution that is produced, to the point where the ‘partial representation’ that we have of the solution is not accessed when the problem is framed differently. In one famous example, children working in Brazilian markets were able to demonstrate mathematical strategies on their stalls that they could not do in the classroom. Knowing how to do something in one context is no guarantee of being able to demonstrate the same skill in another.

 

Neuroconstructivism and constructivist learning

This brief survey has aimed to show how our experiences in the world can only ever lead to context-specific, partial representations of these events in our brains. It should hopefully be immediately clear that this evidence fits very well with many of the main ideas of constructivism as a theory of learning. The constructivist idea that meaning and knowledge are created by the individual in response to their specific experiences and ideas clearly fits very comfortably with the notion of partial representations. Given that each individual will have their own unique neural, network, bodily and social context at any one moment, even the same environmental stimulus is likely to lead to different partial representations of that stimulus in different people.

The neuroscientific evidence also complements the ‘constructive’ aspect of learning very nicely: the building up of ever more complex schemas through the gradual formation of multiple, overlapping partial representations of the world. It also appears to explain the focus of at least some constructivist theories (e.g. Piaget and Dewey) on the individual, highlighting the unique individual context in which any act of learning takes place. Indeed, the individual context is so important that, as we have seen, learning will often not transfer to contexts (even seemingly very similar tasks and environments) which do not share enough of the original features.

 

Neuroconstructivism and constructivist pedagogy

So far so good. Unfortunately, it has been common practice in education to go a step further, and to assume that the theory of learning implies/favours a particular pedagogical approach. Because constructivism emphasises the unique individual context in which learning takes place (and the individual’s role in the construction of that knowledge), so constructivist pedagogy places the onus on the individual student to construct their own understanding of the world. Teachers are therefore encouraged to design learning environments through which students are able to learn for themselves, sometimes facilitating the learning, but generally providing limited explicit guidance. Because constructivism emphasises the ‘active construction’ of knowledge, so constructivist pedagogy often places increased value on hands-on, ‘active’ or experiential learning (such as by experiment, project or solving real-world problems). This will, naturally, be entirely familiar to any teachers reading this post. If your teacher training was anything like mine, it will have been exactly how you were taught to teach.

Unfortunately, the leap from learning theory to pedagogy is not justified. Constructivism, as conceived purely as a theory of learning by Piaget, was not designed to be associated with any specific pedagogical approach. More importantly, the raft of neuroscientific evidence supporting the theory of ‘neuroconstructivism’ actually, in my view, provides strong evidence to suggest the opposite, that constructivist pedagogies are unlikely to be the most effective approaches to learning, at least until schemas are well developed.

 

Partial representations, and constructivist pedagogy 

Let us return to ‘partial representations’, those context-dependent neural traces which constitute how the brain changes in response to experience. As we have seen, these traces are:

  • partial – in that they reflect not the underlying structure of the knowledge but the whole neural, network, bodily and social context in which the knowledge was formed.
  • distributed – in that they are made up of numerous small contributions from neurons distributed across brain areas. No one piece of information therefore resides in any one place, and any reactivation of that knowledge will be an approximate reconstruction, rather than a video-tape playback.
  • context-dependent – as the knowledge contained in our partial representations often does not transfer to situations.

What relevance does this have for pedagogy? Well, let us take a constructivist-inspired pedagogical approach, which involves minimal explicit guidance being given to a student. Imagine a child over the course of a lesson discovering – through some trial and error and some careful teacher-facilitation in a thoughtfully structured learning environment – the formula for Pythagoras’ theorem. It sounds like a lovely educational experience for both student and teacher. Unfortunately, at the end of this process we have (time-consumingly) produced but a single partial representation, which is likely to be highly susceptible to context-dependency.

What the neuroscience of learning tells us is that in order to increase the likelihood of them being more easily accessed subsequently, students require multiple, overlapping partial representations, which are strengthened through repeated access. In his book ‘The Hidden Lives of Learners‘, Graham Nuthall wrote that students need to encounter information three or four times to learn it, and with the idea of multiple, overlapping partial representations we can see why such repeated exposure to information is so important. From this perspective, it is not the discovery of the strategy which is important for subsequent success, but the repeated exposure to it, and practice at accessing it, multiple times and in multiple different ways. There is therefore nothing wrong with the student learning about Pythagoras in the way described above, provided that they are subsequently afforded repeated opportunities to revisit and practise their new knowledge in different contexts. The problem is that, all too often, demonstration of a skill in the learned context is taken as indicating mastery of the skill in general, so lessons move on after a limited number of demonstrations of any new idea. Neuroconstructivism clearly shows that under such circumstances, we are unlikely to form enough overlapping partial distributions to be able to transfer our knowledge to a new context.

 

‘Neuroconstructivist pedagogy’: Multiple, overlapping, partial representations

What I have tried to show above is that our representations of the world, by their very nature, are only ever ‘partial’ representations. Given that, it makes sense for educators to work to create as many of them, and to strengthen them, wherever possible. So is there a particular pedagogical approach that is specifically favoured by the neuroscientific evidence? No. I should be clear I do not think that neuroscience findings can ever be used to directly conclude that any one pedagogical method is better than others. I do, however, think that the evidence underpinning neuroconstructivism provides some important considerations which different pedagogical approaches can all benefit from taking into account. These considerations are encapsulated in the statement:

Students require multiple, overlapping, partial representations of knowledge in order to transfer it to new contexts.

Given that having knowledge that we are able to transfer to a new context is pretty much the point of education itself, I think this is an important message. Again, however, I am not arguing here in favour any one pedagogical approach. I have my personal preferences, but I want to keep this separate from what I think the neuroscience evidence objectively tells us. The creation of multiple, overlapping, partial representations could conceivably be achieved through numerous pedagogical approaches, as long as students have the opportunity for repeated exposure to information and multiple chances to apply new knowledge to different contexts. It could be achieved through more explicit or direct instruction methods, through dialogic approaches, through well-structured co-operative or group strategies, even (if you have enough time spare in your curriculum) through repeated overlapping inquiry-style investigations such as the one described earlier. What teachers need to consider is, in their own context, which (combination of) teaching methods is most likely to produce repeated exposure to knowledge, and practice at applying it to different contexts. What this looks like in each classroom is a decision for each individual teacher.

So there we go. Some people are sceptical that neuroscience has anything to offer education. I agree that the neuroscientific evidence doesn’t provide support for any one concrete strategy of instruction (indeed, I don’t think it ever could). I think it does, however, create a simple and powerful question that I continue to ask myself as I evaluate how I teach. It is also a question that I wish I had been asked right at the start of my teaching career. So… how are you going to ensure that you form multiple, overlapping, partial representations?

 

 

Constructivism is a theory of learning, not a theory of pedagogy. Neuroscience explains why this is important

Bridge of Sighs. My thoughts on reading yet another paper on the relationship of neuroscience and education

When John Bruer wrote his seminal 1997 paper ‘Education and the brain: A bridge too far’, arguing that the potential of neuroscience to directly influence education was limited, I imagine that he would not have anticipated two aspects of the subsequent debate over 20 years later. The first surprise would be that the substantive points of the debate have hardly moved on at all over that time. A great deal of ink has been spilled by those who agree with Bruer’s scepticism (e.g. here, here, here, here and here) and those who are more optimistic (here, here, here, here, here, here and here), but for the most part they have tended to talk past one another, relying on different characterisations of what a ‘neuroscientific application to education’ actually entails (prime example of this to come). The second surprising consequence of Bruer’s article has been to spawn a whole academic sub-genre of papers about education and neuroscience with ‘bridge’ related titles. We’ve had calls to ‘build bridges’ (e.g. here and here), ‘envisioned bridges’, ‘boundaries as bridges’ and (most enjoyably), ‘bridges over troubled waters’.

https---upload.wikimedia.org-wikipedia-commons-thumb-b-b2-Bridge_Astore.jpg-1024px-Bridge_Astore

A paper just published in Current Directions in Psychological Science is the latest addition to both exhibits. In ‘Neuroscience and Education: A Bridge Astray’, Michael Dougherty and Alison Robey argue that,

Although we acknowledge the value of neuroscience for understanding brain mechanisms, we argue that it is largely unnecessary for the development of effective learning interventions. We demonstrate how neuroscience findings have failed to generalize to classroom contexts by highlighting the recent popularity and failed results from brain-training research

As this summary suggests, the paper contrasts brain training, an apparent example of an ineffective educational intervention from neuroscience, with more effective educational interventions from cognitive science, for example memory strategies such as spaced retrieval. Unfortunately, this comparison does not stand up to a great deal of deeper scrutiny. I’ll address first my specific concerns with the paper, before trying to show how I found many of the points in the paper symptomatic of my frustrations with much of the previous literature, on both sides of the argument.

 

Building bridges, or building silos?

One of the key assumptions of arguments on the sceptical end of the educational neuroscience debate is that it is possible to very clearly demarcate between research disciplines, e.g. that there is a clear boundary between neuroscience and cognitive psychology. This is obviously essential if you are planning to claim that neuroscience cannot influence education, but psychology can. The problem is that this demarcation is increasingly difficult to perform. The ever-expanding field of cognitive neuroscience explicitly blurs the boundaries between the cognitive and the neural, with each (in theory) reciprocally informing the other. Dougherty and Robey are therefore left with a bit of a dilemma in terms of how to separate ‘neuroscientific’ interventions, from ‘cognitive’ interventions. Their answer is to latch onto some foundational neuroscientific ideas (brain plasticity and synaptogenesis), and to argue that these concepts have not been translated into any useful educational interventions. This is already somewhat shaky ground as pretty much every brain process relies on brain plasticity, including of course the memory changes caused by the interventions based on cognitive theory which they are using as a contrast. However Dougherty and Robey seem want to differentiate interventions which have been initially inspired by neuroscience evidence, as opposed to those initially inspired by cognitive/behavioural evidence. It’s still a difficult and slightly artificial balancing act, but we will accept it for the time being.

 

Their demonstration of an intervention ‘inspired’ by neuroscientific evidence (and also, therefore, their example of the failure of neuroscience to be able to influence education) is the case of ‘brain training’. Brain training interventions (i.e. training on one or more cognitive tasks with the aim that this generalises, or “transfers,” to improved performance on other cognitive tasks and to daily life) has generally been found to be ineffective (see e.g. here, here, here, here and here). With this I am in absolute agreement with the authors. They then write, however:

Brain training is emblematic of the gulf between basic neuroscience and education, wherein seemingly groundbreaking neuroscience findings (e.g., brain plasticity, synaptogenesis, pruning) simply do not scale up to practical education interventions

Is this casual alignment of ‘brain training’ and neuroscience merited? For starters, as the authors themselves point out in the article, ‘brain training’ interventions are just as often called ‘cognitive training’ or ‘working memory training’! This seems somewhat odd if the origin of such training programs can (as the authors seem to claim) be so clearly linked back to purely neuroscientific findings, rather than cognitive psychology, or (whisper it quietly), a mixture of the two. It’s an example of the need for some opponents of educational neuroscience to silo off research disciplines into mutually exclusive territories in order to contrast them; a view which is becoming increasingly difficult to sustain in the face of interdisciplinary research fields such as developmental cognitive neuroscience. Their claim also, however, relies on two key assumptions, which merit dealing with in some more detail.

 

  1. Is ‘brain training’ “emblematic” of educational neuroscience?

The answer to this is a very clear no. There is a huge range of work being done at the moment under the broad banner of ‘neuroscience and education’, and brain training is, frankly, but a tiny fraction of this work. Some of these projects I am very excited about, others cause me concern. For example, I think that far too many ‘educational neuroscience’ projects are designed with little understanding (or interest) in the realities of everyday educational environments. I have taken issue with the unwarranted application of neuroscience into areas of education where it has no business before (e.g. here), and argued that the immediate role of neuroscience in education is not to create brand new, flashy pedagogies or tools, but to help us understand and critique the ideas that are already in use in the classroom (e.g. here and here). In summary I am no evangelist for the idea that ‘educational neuroscience’ research will always produce educational benefit, but I can recognize a misrepresentation of a broad research area when I see one. Treating brain training as “emblematic” of attempts to use neuroscience to improve education is exactly that.

 

  1. Is ‘brain training’ really a neuroscientific idea?

A central part of the argument of Dougherty and Robey is that the interest in ‘brain training’ can be traced directly from basic neuroscientific research findings. I question this assumption. As we have seen already, the fact that brain training is also just as often called cognitive training or working memory training, already raises some doubts as to how purely ‘neuroscientific’ this idea is. Beyond this, however, I would argue that the hypotheses behind brain training are not actually supported by basic neuroscience research and theory, and therefore that the claim that it is a ‘direct neuroscientific intervention to education’ is false. At best, perhaps, it is a ‘direct intervention to education based on a misrepresentation of neuroscience evidence’, but that is rather less catchy. In actual fact, the idea that brain training would be an effective intervention is incompatible with a good deal of what we know about how the brain operates.

Brain training programs rely on the idea of ‘cognitive transfer’, the idea that knowledge or skills learned in one context can be applied to another context (for example, that someone who has improved their ability to do working memory tasks will, as a result, also be better able to do mental arithmetic). Unfortunately, a very large amount of evidence attests to the fact this this is not, in fact, how the brain works, and that brain activity is often stubbornly, frustratingly, context specific. In the influential developmental theory of ‘neuroconstructivism’, for example, context specificity is taken as a central feature of how the brain operates (I have described neuroconstructivism in more detail previously). This means that at all levels, from the neural to the environmental, the activity of the brain is dependent on the context that it finds itself in. Individual neurons compete and constrain one another’s activity. Brain networks do the same, The state of our bodies and the environmental context at the time all also determine the activity of our brain in response to any specific event. In such a system, any neural trace is merely a ‘partial representation’, a representation of the world which captures some, but not all of it. Partial representations are by definition hugely context specific, they record the precise state of the organism at the time of the event, rather than any general underlying feature of it. This means that they don’t generalise very well to other contexts. In one great example (from Karni et al., 1995: note, well before the existence of any ‘brain training’ programs), learning a sequence of movements in one order was found not to transfer to improved learning of a sequence of the same movements in a different order. Given results like this, we should hardly be surprised when brain training programs fail to show transfer to much more distant cognitive abilities.

Dougherty and Robey seem to suggest that brain training programs are a logical extension of neuroscientific findings. I would argue that, in actual fact, a raft of neuroscientific evidence supports the opposite conclusion, that changes in the brain are often context-dependent, and as a result form only partial representations of the world which do not effectively transfer to new contexts.

 

The journal’s gain is teaching’s loss

A lot of water has flowed under the bridge since Bruer first erected it, but I feel that Dougherty and Robey’s article illustrates many of the wider reasons why little progress has been made. A straw man of a ‘neuroscience intervention’, unnatural and forced demarcations between subject areas, strange definitions of what the ‘success’ of a neuroscience intervention would look like (here, successful cognitive transfer, in other cases ‘brand new pedagogical ideas’). Both sides are responsible for these, for example I have been critical (as indeed are the authors of the article), of attempts to co-opt findings which were clearly initially the result of cognitive or educational psychology as ‘educational neuroscience’, based on the fact that we have subsequently done a few brain scans which back up the original findings1. Perhaps the name ‘educational neuroscience’ itself is inadvertently creating some of this tension, suggesting a primacy of neuroscience over psychology. In truth this is not the view of anyone working in the field that I am aware of, but the perception may remain. It is partly for this reason that American researchers have tended to use the more accurate (though less pithy) title ‘Mind, Brain and Education Science’ when referring to the same field.

The sadness in this logical and semantic squabbling is that is has real world consequences. Teachers are genuinely interested in, and keen to learn about, findings from neuroscience and psychology which may be relevant to their practice (Simmonds, 2014). In the absence of any coherent messages from academia (and whilst journals happily lap up publication fees for articles from either side) other voices slide into the void offering ‘brain-based’ solutions to teachers which may be at best ineffective, and at worst downright harmful.

http---www.chinatoday.com-entertain-china.funny.pictures-funny_picture_2012_13-943-broken_bridge
The bridge which REALLY needs attention is the one to education

The real priority: the collapsed bridge to education

In truth I don’t think it actually matters at all which academic silo/s your data was created in. The debates above are spending time arguing about the order of the traffic, without noticing that the only bridge that is genuinely important here, the one between academia and education, is in a terrible state of repair. The pressing issue, for all parties, is that we still have no coherent framework for translating research findings into educational applications, and hence no clear system for actually finding out which research actually isbeneficial to education, and which is not. Translating research to the classroom is a challenging and complex task, and just as neuroscience findings shouldn’t be applied direct to classrooms, neither should cognitive psychology ones. Amongst other things, any promising research result will need to be adapted, ideally in collaboration with teachers, to create an intervention which is practical (for use in different classrooms and by different teachers), beneficial (i.e. producing outcomes that are of direct relevance and interest to a teacher) and valid (retaining scientific rigour and the ability to be evaluated for effectiveness).  At present, however, there exists almost no formal guidance or research on how this process can best be achieved to the mutual benefit of all parties. This is, for me, a far more important issue than to debate the philosophical implications of bridges between different subject areas.

Only with a proper framework for translating research into practice will we be truly able to see which neuroscientific or psychological findings will translate into valuable educational interventions and which will not. Results will be inconsistent; such is the nature of trying to apply research to the messy real world. I’d love to see it happen, though I wouldn’t like to predict which ideas will prove the most impactful. We’ll just have to cross that bridge when we come to it.

 

Footnote:

1A prime example might be the spacing effect in memory (the idea that information is more effectively learned in a number of spaced sessions than in one go, first reported by Herman Ebbinghaus in 1885), whichis currently under investigation in schools as a project funded by an ‘Education and Neuroscience’ research grant, despite the findings almost pre-dating psychology, let alone neuroscience.

Bridge of Sighs. My thoughts on reading yet another paper on the relationship of neuroscience and education

Another picture to give adolescent researchers nightmares!

My last post looked at a picture I’d made of a timeline of adolescent cognitive development, and the somewhat pessimistic conclusions that I drew from it regarding the way we conduct developmental research. Discussions continue, and I agree that larger scale online (possibly game-based) research tools may be a way of addressing some of the concerns I raised.

In discussing the timeline when I first sent it out on Twitter, Lucia Magis Weinberg, a developmental cognitive neuroscientist at UC Berkeley, sent me another picture that is similarly, perhaps even more, haunting!

It’s taken from this review paper by van Duijvenvoorde et al (2016) on adolescent responses to rewards. It simply shows how different studies have chosen to define ‘adolescence’ (and well as ‘children’ and ‘adults’ in some cases). Light blue represents children (as defined in the studies). Blue represents adolescence (as defined in the studies), and dark blue represents adults (as defined in the studies).

image.png

The picture caption concludes:

The graph shows that there are large differences between studies in how adolescence is defined in ages in years. Moreover, there is overlap between studies in the ages that refer to as children and adolescents. What can be concluded from this figure is that here is a great need for more detailed measurement of adolescence in terms of age in years, but also in terms of pubertal development

In less measured terms, how on earth can we have any idea what ‘adolescents’ can and can’t do if no one can agree on how to define the period in the first place?

Another picture to give adolescent researchers nightmares!

I made a picture, and it made me worry about the state of academic research with adolescents

Recently I made the picture below (larger version can be viewed and downloaded here).

Development of EF functions timeline_adjusted

It’s a timeline of the maturation of adolescent cognitive control, in other words when different studies have found that adolescents reach ‘adult’ performance levels on various tasks that require some sort of effortful control of behaviour (such as resisting distraction, inhibiting responses etc.) Above the arrow are ‘hot’ tasks (with an emotional/affective component). Below the arrow are ‘cool’ tasks (with no affective component). Blue arrows between studies = same test/task used in between different studies. Red arrow = different tasks used by the same study/sample (or an arrow to a picture of what the task looks like).

Caveats as follows:

  • I only included studies which gave some comparison to ‘normal’ adult performance. Unless otherwise stated, the age given is the age at which adult performance is reached…
  • … any other performance level is described, e.g. “still developing” “more than adults” etc
  • I doing this I ignored lots of research comparing different developmental age groups to each other (but not to adults)
  • I will also have undoubtedly missed many other useful studies. If you think I have then please get in touch and I will be only too happy to collaborate on an updated version.
  • Being at an ‘adult level’ by a particular age could of course mean that the stage has been reached earlier – the ability to draw these conclusions is dependent on the age groups chosen by each study
  • Some of these studies, especially some of the older ones, have fairly small sample sizes (though I did try to police this, and other methodological issues, to an extent). If you know of any specific problems with cited studies then please let me know.
  • I have definitely made mistakes

I made this picture as I hoped that by creating a visual depiction of adolescent cognitive development, I would find it easier to synthesise the research into some general themes or points of agreement. In actual fact, however, I think the picture tells us less about adolescent development, and more about how we do research (both specifically research on adolescents, but also research in general). Specifically, I think it sheds some light on why it is often so difficult to draw any firm conclusions from parts of the academic literature. To summarise them into two primary concerns:

We have too many, construct-led experimental tasks and procedures.

Too many different tasks/paradigms

The huge variety in experimental tasks used in developmental research makes comparison very hard. I have no idea what half of the tasks in these papers actually involve without looking them up. Often I have even less idea about how performance on one of them can help us understand or predict performance on a different task. This is partly the result of a novelty bias in research where people are encouraged to create ‘novel’ research tests and designs, (which may merely muddy the water rather than extend our knowledge of how skills develop). The relatively small number of blue arrows on the picture (showing the same task being used across studies) bears testament to this, and sometimes these are just the same research group reusing the task that they invented! Of course, this is part of a larger issue which can be tracked all the way back to the replication crisis in Psychology and the perverse academic incentives which encourage bad science. Many others have written about this far better than I can. However, the confusion is not entirely due to this. It is also partially the result of…

Too much construct-led research

To make matters worse, our tests are often construct-led, which means that they are designed to investigate a specific construct, however poorly-defined and arbitrarily demarcated. For example, we have a huge variety of overlapping terms and constructs describing various aspects of how we control our thoughts and behaviour. Inhibitory control is sometimes used to describe just inhibiting specific movements or behaviours (also sometimes called response-inhibition, or just self-control). Sometimes, however, ‘inhibitory control’ is used to also encompass inhibiting attention (variously called selective attention, attention control, executive attention or distraction control), and inhibiting thoughts or memories (also sometimes called cognitive inhibition). I have also seen cognitive and attentional inhibition grouped as ‘interference control’ as distinct from response inhibition.

These constructs also dictate the design of the tasks that we use in experiments, and the conclusions that we draw from them. For example, the ‘Flanker Task’ (below) is generally used as a test of executive/selective attention, but is also sometimes described as a test of general inhibitory control.

flankers
In the flanker task, the participant has to report the direction of the middle arrow, ignoring the direction of the ‘flanking’ stimuli

Similarly, the famous ‘Stroop’ task is sometimes defined as a test of purely response competition, but I have also seen it used as a ‘test of’ general inhibitory control and even of just attention control.

stroopsample
The Stroop test

This multiplicity of tests, domains and definitions creates a very muddled picture1 (see the very muddled picture at the top of the page for a prime example!) How useful is this? What does this allow us to really conclude about when the average adolescent can display a certain skills? More importantly, how much does allowing constructs to guide our research aims actually tell us about the real world? Does the distinction between, for example, cognitive inhibition and selective attention actually help us to predict real behaviour in adolescents? Can we dissociate them and their effects on everyday behaviour? Remarkably little research has been done which comes close to answering these sorts of questions.

I would suggest that research on adolescents (and in other areas) would be greatly enhanced by moving more from being a construct-led process, to being a function-led process. By this I mean that research aims and hypotheses should be initially guided by real world abilities and behaviours produced by adolescents, and the known outcomes of these2. Examples of these might be reports of distraction in schools classrooms, peer influence on driving safety, sleep patterns, drug taking, social media use, depression and so on3.

If we know, for example, that distracted children do worse at school (they do), then we can start by trying to design a test which might mimic the ability to control attention in a real classroom, see if this test can quantify and then predict the levels of distraction of that child in the classroom. We can use this same test repeatedly to see how this ability develops within individuals over time and then relate this to classroom reports again to see if the same patterns of development are present in behaviour. Finally we can see if our test (or other measures derived from the test) can be of use in reducing distraction in students, for example by testing the effectiveness of a number of simple environmental adjustments to reduce external distractions. This research program (which, as I have unfortunately discovered, is rather closer to a career’s worth than a PhD’s worth!) makes no reference to a pre-existing research construct regarding the control of behaviour. It would be an example of applied, function-led cognitive neuroscience research and, I think, it would tell us much more about what real world skills adolescents are capable of, and when.

Looking at the muddle of my timeline, I rather hope it happens soon!

 

P.S. If any readers draw any other interpretations from the timeline, or read any patterns in the tea leaves that I can’t spot, then I’d love to hear from you! Please do get in touch.

 

  1. … and I haven’t even touched here on research which looks at one of these constructs in conjunction with another one, for example inhibitory control in emotional settings or under working memory load!
  2. Construct-led, basic scientific research on developmental abilities is still important and should continue (see for example this interesting recent paper which dissociated brain activity associated with various types of inhibition), I would just like to see the balance shifted to, at the very least, parity between projects which are primarily guided by real world applications and those which are guided by more theoretical concerns.
  3. This is not to say that there are no programs researching these areas. In some of these examples, such as peer influence on driving safety, the link between the lab and real world behaviour has been made very successfully and convincingly. They are in a minority though.
I made a picture, and it made me worry about the state of academic research with adolescents

It is more useful for teachers to see attention as an effect not a cause

We frequently urge our students (and ourselves!) to “pay attention”, but what do we really mean by this phrase? The notion of ‘paying attention to’ an object creates an impression of attention as a resource which we have at our disposal, ready to be deployed (or not) at our command. This is reinforced by the numerous metaphors we have for the concept (a filter, a spotlight, a zoom lens, even a glue). In all of these metaphors attention is cast in the role of a ‘tool’ for us to use. Going a level deeper, such metaphors all take for granted a reified concept of attention, i.e. that attention is a real, measurable ‘thing’. Increasingly in the academic study of attention, however, there is some opposition to these traditional notions of what attention is. These can generally be summarised as an effort to recast attention as an effect rather than a cause (e.g. here, here and here, amongst others).

I think that this seemingly niche academic debate actually has some interesting implications for educators. Thinking of attention as an effect rather than a cause can throw a new light on the understudied problem of attention in schools, and what teachers can do about it.

What’s the problem?

There are two main problems with the metaphors which reify attention as a resource, one logical and one practical. The logical one (which is of less relevance perhaps to educators, but still useful for fans of logical validity), is that the evidence for these models of memory often rest on circular arguments. For example, if we take the metaphor of attention as a flashlight, imagine the following dialogue, taken from this paper by Vince Di Lollo 

Person A: a stimulus flashed at a location just ahead of a moving object is perceived more promptly and more accurately. 

Person B: why is that? 

A: because the attentional spotlight is deployed to that location, and stimuli presented at an attended location are processed more promptly and more accurately. 

B: and how do we know that attention has been deployed to that location? 

A: we know it because stimuli presented at that location are perceived more promptly and more accurately. 

The practical problem is that metaphors such as this place the burden for the control of attention firmly on the student themselves, to deploy as their preferences or abilities allow. Now I am in no way arguing that students are not capable of exercising control over their attention, nor that teachers should be held responsible when a student’s attention wanes; indeed I would strongly repudiate this. I do think, however, that a model which casts attention as a resource of the student is unhelpful to teachers. Teachers looking to improve student performance using this model are left with few options, other than perhaps brain training (so far unimpressive) or vague appeals to a student’s better nature (“direct your attention towards this pleeeease”).

Attention as an effect not a cause

Far more productive for educators would be the discussions arising from seeing attention as an effect, rather than a cause. This reconceptualisation naturally invites the consideration of

  1. which conditions are most likely to engender the effect of focused attention?

and importantly…

  1. which information should we create these conditions for?

Again, this suggestion in no way denies that students are causal agents in their own behaviour, merely that teachers will be more empowered by a focus on the conditions that they can create whereby attention emerges as an effect.

I will look at each of these two questions in turn

Which conditions are most likely to engender the effect of focused attention?

Decades of careful psychological work in dark laboratories has helped to confirm a lot of common sense notions about how attention can be captured (e.g. by bright colours, unexpected shapes or other features which stand out, stimuli which move, or loom, motivation, meaning, reward and so on.) If these features are present in the task, we are generally more able to focus on the task. If they are present in a stimulus which is not part of the task, then we are more likely to be distracted. 

So far so good; as teachers we need to try to make our stimuli as salient as possible and reduce other distractions. However this apparent simplicity leads us on to the second, less commonly considered implication of considering attention as an effect; if we can create the conditions to direct student attention, what information exactly do we want to create these conditions for? In other words (returning to the old metaphor for simplicity’s sake), what do we want student attention to be focused on?

Which information should we create these conditions for? Selecting the target of focused attention

Where attention is discussed in education, it tends to be focused on the ideas above, in terms of strategies for capturing attention. Indeed, the goal of my teacher training on this topic was entirely this, the creation of an attentive, engaged class. What exactly they should be engaged by seemed less of a concern. If attention could be attracted by the teacher, then learning was assumed to be an inevitability.

Sadly this position is mistaken. I have written before of the limited capacity ‘bottleneck’ of attention. This limited capacity means that only a tiny fraction of information arriving into our perceptual systems will ever be processed to a meaningful degree. Therefore whilst an attentive class can clearly be a step in the right direction, they will only be learning efficiently if we as educators ensure that their attention is focused precisely on the stimuli that we want it to be. Just as becoming distracted by low level disruption from other students may inhibit learning, so will engaging with superfluous material presented by the teacher. If attention is the result of creating the right conditions, then we need to be very clear about exactly what is worth creating those conditions for.

Take the case of powerpoint slides. I spent many hours early in my teaching career crafting aesthetically pleasing powerpoint slides. My backgrounds were salient, full of nice bright colours. Some of the words zoomed in from the side of the screen. Text was usually accompanied with a picture (or even a gif if I was feeling particularly creative), usually humorous and tangentially related to the main information. Looking back, especially through the lens of considering attention as an effect and questioning where I was encouraging that effect to occur, is sobering. My salient features were either surface level features (the movement of the text rather than its content) or entirely irrelevant to what I wanted the students to know (the background and the pictures). I was actively inviting attention to be directed away from that which I thought was most important. This is why important and potentially impactful strategies such as dual coding, which combine visual and verbal materials (and which has been valuably popularised recently by figures such as Oliver Caviglioli), need to be treated with caution.  Bad dual coding is not just ineffective, it leads to split attention or outright distraction.

Deciding on the right targets for our students’ attention is a challenge that cuts right across education, from the pedagogical issue of how best to deliver information to the curricular decisions required to identify precisely what it is that we want students to know in the first place. I have been delighted to see an increased focus on curriculum amongst teacher networks as a result (e.g. here, here and many great posts here for example, though to my knowledge the specific link between the importance of curriculum design and attention has not be explored either in blogs or research).

A new way to view attention in the classroom

Viewing attention as an effect makes us value it (and the contribution that we can make to it) more. It makes us consider more carefully how to attract attention, but crucially also what we want to attract attention to. Capturing attention is not in itself the aim. The goal is to provide the optimal conditions so that attention is captured by the exact stimuli that we have identified as most valuable. I have tried to argue here that this process may be assisted if we define attention less as a cause of student behaviour and more as an effect of the conditions that we put in place. 

It is more useful for teachers to see attention as an effect not a cause