cognitive science: the wrong end of the stick

A few years ago, some teachers began advocating the application of findings from cognitive science to education. There seemed to be something not quite right about what they were advocating but I couldn’t put my finger on exactly what it was. Their focus was on the limitations of working memory and differences between experts and novices. Nothing wrong with that per se, but working memory and expertise aren’t isolated matters.

Cognitive science is now a vast field; encompassing sensory processing, perception, cognition, memory, learning, and aspects of neuroscience. A decent textbook would provide an overview, but decent textbooks didn’t appear to have been consulted much. Key researchers (e.g. Baddeley & Hitch, Alloway, Gathercole), fields of research (e.g. limitations of long-term memory, neurology), and long-standing contentious issues (e.g. nature vs nurture) rarely got a mention even when highly relevant.

At first I assumed the significant absences were due to the size of the field to be explored, but as time went by that seemed less and less likely.  There was an increasing occurrence of teacher-B’s-understanding-of-teacher-A’s-understanding-of-Daniel-Willingham’s-simplified-model-of-working-memory, with some teachers getting hold of the wrong end of some of the sticks. I couldn’t understand why, given the emphasis on expertise, teachers didn’t seem to be looking further.

The penny dropped last week when I read an interview with John Sweller, the originator of Cognitive Load Theory (CLT), by Ollie Lovell, a maths teacher in Melbourne. Ollie has helpfully divided the interview into topics in a transcript on his website. The interview clarifies several aspects of cognitive load theory. In this post, I comment on some points that came up in the interview, and explain the dropped penny.

1.  worked examples

The interview begins with the 1982 experiment that led to Sweller’s discovery of the worked example effect. Ollie refers to the ‘political environment of education at the time’ being ‘heavily in favour of problem solving’. John thinks that however he’d presented the worked example effect, he’d be pessimistic about the response because ‘the entire research environment in those days was absolutely committed to problem solving’.

The implication that the education system had rejected worked examples was puzzling. During my education (1960s and 70s) you couldn’t move for worked examples. They permeated training courses I attended in the 80s, my children’s education in the 90s and noughties, and still pop up frequently in reviews and reports. True, they’re not always described as a ‘worked example’ but instead might be a ‘for instance’ or ‘here’s an example’ or ‘imagine…’. So where weren’t they? I’d be grateful for any pointers.

2 & 3. goal-free effect

Essentially students told to ‘find out as much as you can’ about a problem, performed better than those given specific instructions about what to find out. But only in relation to problems with a small number of possible solutions – in this case physics problems. The effect wasn’t found for problems with a large number of possible solutions.   But you wouldn’t know that if you’d only read teachers criticising ‘discovery learning’.

4. biologically primary and secondary skills

What’s determined by biology or by the environment has been a hugely contentious issue in cognitive science for decades. Basically, we don’t yet know the extent to which learning is biologically or environmentally determined.  But the contentiousness isn’t mentioned in the interview, is marginalised by David Geary the originator of the biologically primary and secondary concept, and John appears to simply assume Geary’s theory is correct, presumably because it’s plausible.

John says it’s ‘absurd’ to provide someone with explicit instruction about what to do with their tongue, lips or breath when learning English. Ollie points out that’s exactly what he had to do when he learned Chinese. John claims that language acquisition by immersion is biologically primary for children but not for adults. This flies in the face of everything we know about language acquisition.

Adults can and do become very fluent in languages acquired via immersion, just as children do. Explicit instruction can speed up the process and help with problematic speech sounds, but can’t make adults speak like a native. That’s because the adults have to override very robust neural pathways laid down in childhood in response to the sounds the children hear day-in, day-out (e.g. Patricia Kuhl’s ‘Cracking the speech code‘). The evidence suggests that differences between adult and child language acquisition are a frequency of exposure issue, not a type-of-skill issue. As Ollie says: “It’s funny isn’t it?  How it can switch category. It’s just amazing.”  Quite.

5. motivation

The discussion was summed up in John’s comment: “I don’t think you can turn Cognitive Load Theory into a theory of motivation which in no way suggests that you can’t use a theory of motivation and use it in conjunction with cognitive load theory.

 6. expertise reversal effect

John says: “As expertise goes up, the advantage of worked examples go down, and as expertise continues to go up, eventually the relative effectiveness of worked examples and problems reverses and the problems are more helpful than worked examples”.

7. measures of cognitive load

John: “I routinely use self-report and I use self-report because it’s sensitive”. Other measures – secondary tasks, physiological markers – are problematic.

8. collective working memory effect

John: “In problem solving, you may need information and the only place you can get it from is somebody else.” He doesn’t think you can teach somebody to act collaboratively because he thinks social interaction is biologically primary knowledge. See 4 above.

9. The final section of the interview highlighted, for me, two features that emerge from much of the discourse about applying cognitive science to education:

  • The importance of the biological mechanisms and the weaknesses of analogy.
  • The frame of reference used in the discourse.

biological mechanisms

In the final part of the interview John asks an important question: Is the capacity of working memory fixed? He says: “If you’ve been using your working memory, especially in a particular area, heavily for a while, after a while, and you would have experienced this yourself, your working memory keeps getting narrower and narrower and narrower and after a while it just about disappears.”

An explanation for the apparent ‘narrowing’ of working memory is habituation, where the response of neurons to a particular stimulus diminishes if the stimulus is repeated. The best account I’ve read of the biological mechanisms in working memory is in a 2004 paper by Wagner, Bunge & Badre.  If I’ve understood their findings correctly, signals representing sensory information coming into the prefrontal area of the brain are maintained for a few seconds until they degrade or are overridden by further incoming information. This is exactly what was predicted by Baddeley & Hitch’s phonological loop and visual-spatial sketchpad. (Wagner, Bunge and Badre’s findings also indicate there might be more components to working memory than Baddley & Hitch’s model suggests.)

John was using a figure of speech, but I fear it will only be a matter of time before teachers start referring to the ‘narrowing’ of working memory. This illustrates why it’s important to be aware of the biological mechanisms that underpin cognitive functions. Working memory is determined by the behaviour of neurons, not by the behaviour of analogous computer components.

frame of reference

John and Ollie were talking about cognitive load theory in education, so that’s what the interview focussed on, obviously.  But every focus has a context, and John and Ollie’s frame of reference seemed rather narrow. Ollie opens by talking about ‘the political environment of education at the time [1982]’ being ‘heavily in favour of problem solving’. I don’t think he actually means the ‘political environment of education at the time’ as such. Similarly John comments ‘the entire research environment in those days was absolutely committed to problem solving’. I don’t think he means ‘the entire research environment’ as such either.

John also observes: “It’s only been very recently that people started taking notice of Cognitive Load Theory. For decades I put papers out there and it was like putting them into outer-space, you know, they disappeared into the ether!” I first heard about Cognitive Load Theory in the late 80s, soon after Sweller first proposed it, via a colleague working in artificial intelligence. I had no idea, until recently, that Sweller was an educational psychologist. People have been taking notice of CLT, but maybe not in education.

Then there’s the biologically primary/secondary model. It’s ironic how little it refers to biology. We know a fair amount about the biological mechanisms involved in learning, and I’ve not yet seen any evidence suggesting two distinct mechanisms. The model appears to be based on the surface features of how people appear to learn, not on the deep structure of how learning happens.

Lastly, the example of language acquisition. The differences between adults and children learning languages can be explained by frequency of exposure and how neurons work; there’s no need to introduce a speculative evolutionary model.

Not only is cognitive load theory the focus of the interview, it also appears to be its frame of reference; political issues and knowledge domains other than education don’t get much of a look in.

the penny that dropped

Ever since I first heard about teachers applying cognitive science to education, I’ve been puzzled by their focus on the limitations of working memory and the characteristics of experts and novices. It suddenly dawned on me, reading Ollie’s interview with John, that what the teachers are actually applying isn’t so much cognitive science, as cognitive load theory. CLT, the limitations of working memory and the characteristics of experts and novices are important, but constitute only a small area of cognitive science. But you wouldn’t know that from this interview or most of the teachers advocating the application of cognitive science.  There’s a real risk, if CLT isn’t set in context, of teachers getting hold of the wrong stick entirely.

references

Geary, D. (2007).  Educating the evolved mind: Conceptual foundations for an evolutionary educational psychology, in Educating the evolved mind: Conceptual foundations for an evolutionary educational psychology, JS Carlson & JR Levin (Eds). Information Age Publishing.

Kuhl, P. (2004). Early language acquisition: Cracking the speech code. Nature Reviews Neuroscience 5, 831-843.

Wagner, A.D., Bunge, S.A. & Badre, D. (2004). Cognitive control, semantic memory          and priming: Contributions from prefrontal cortex. In M. S. Gazzaniga (Ed.) The Cognitive Neurosciences (3rd edn.). Cambridge, MA: MIT Press.

 

 

 

 

 

 

 

 

 

 

Advertisements

evolved minds and education: evolved minds

At the recent Australian College of Educators conference in Melbourne, John Sweller summarised his talk as follows:  “Biologically primary, generic-cognitive skills do not need explicit instruction.  Biologically secondary, domain-specific skills do need explicit instruction.”

sweller.png

Biologically primary and biologically secondary cognitive skills

This distinction was proposed by David Geary, a cognitive developmental and evolutionary psychologist at the University of Missouri. In a recent blogpost, Greg Ashman refers to a chapter by Geary that sets out his theory in detail.

If I’ve understood it correctly, here’s the idea at the heart of Geary’s model:

*****

The cognitive processes we use by default have evolved over millennia to deal with information (e.g. about predators, food sources) that has remained stable for much of that time. Geary calls these biologically primary knowledge and abilities. The processes involved are fast, frugal, simple and implicit.

But we also have to deal with novel information, including knowledge we’ve learned from previous generations, so we’ve evolved flexible mechanisms for processing what Geary terms biologically secondary knowledge and abilities. The flexible mechanisms are slow, effortful, complex and explicit/conscious.

Biologically secondary processes are influenced by an underlying factor we call general intelligence, or g, related to the accuracy and speed of processing novel information. We use biologically primary processes by default, so they tend to hinder the acquisition of the biologically secondary knowledge taught in schools. Geary concludes the best way for students to acquire the latter is through direct, explicit instruction.

*****

On the face of it, Geary’s model is a convincing one.   The errors and biases associated with the cognitive processes we use by default do make it difficult for us to think logically and rationally. Children are not going to automatically absorb the body of human knowledge accumulated over the centuries, and will need to be taught it actively. Geary’s model is also coherent; its components make sense when put together. And the evidence he marshals in support is formidable; there are 21 pages of references.

However, on closer inspection the distinction between biologically primary and secondary knowledge and abilities begins to look a little blurred. It rests on some assumptions that are the subject of what Geary terms ‘vigorous debate’. Geary does note the debate, but because he plumps for one view, doesn’t evaluate the supporting evidence, and doesn’t go into detail about competing theories, teachers unfamiliar with the domains in question could easily remain unaware of possible flaws in his model. In addition, Geary adopts a particular cultural frame of reference; essentially that of a developed, industrialised society that places high value on intellectual and academic skills. There are good reasons for adopting that perspective; and equally good reasons for not doing so. In a series of three posts, I plan to examine two concepts that have prompted vigorous debate – modularity and intelligence – and to look at Geary’s cultural frame of reference.

Modularity

The concept of modularity – that particular parts of the brain are dedicated to particular functions – is fundamental to Geary’s model.   Physicians have known for centuries that some parts of the brain specialise in processing specific information. Some stroke patients for example, have been reported as being able to write but no longer able to read (alexia without agraphia), to be able to read symbols but not words (pure alexia), or to be unable to recall some types of words (anomia). Language isn’t the only ability involving specialised modules; different areas of the brain are dedicated to processing the visual features of, for example, faces, places and tools.

One question that has long perplexed researchers is how modular the brain actually is. Some functions clearly occur in particular locations and in those locations only; others appear to be more distributed. In the early 1980s, Jerry Fodor tackled this conundrum head-on in his book The modularity of mind. What he concluded is that at the perceptual and linguistic level functions are largely modular, i.e. specialised and stable, but at the higher levels of association and ‘thought’ they are distributed and unstable.  This makes sense; you’d want stability in what you perceive, but flexibility in what you do with those perceptions.

Geary refers to the ‘vigorous debate’ (p.12) between those who lean towards specialised brain functions being evolved and modular, and those who see specialised brain functions as emerging from interactions between lower-level stable mechanisms. Although he acknowledges the importance of interaction and emergence during development (pp. 14,18) you wouldn’t know that from Fig 1.2, showing his ‘evolved cognitive modules’.

At first glance, Geary’s distinction between stable biologically primary functions and flexible biologically secondary functions appears to be the same as Fodor’s stable/unstable distinction. But it isn’t.  Fodor’s modules are low-level perceptual ones; some of Geary’s modules in Fig. 1.2 (e.g. theory of mind, language, non-verbal behaviour) engage frontal brain areas used for the flexible processing of higher-level information.

Novices and experts; novelty and automation

Later in his chapter, Geary refers to research involving these frontal brain areas. Two findings are particularly relevant to his modular theory. The first is that frontal areas of the brain are initially engaged whilst people are learning a complex task, but as the task becomes increasingly automated, frontal area involvement decreases (p.59). Second, research comparing experts’ and novices’ perceptions of physical phenomena (p.69) showed that if there is a conflict between what people see and their current schemas, frontal areas of their brains are engaged to resolve the conflict. So, when physics novices are shown a scientifically accurate explanation, or when physics experts are shown a ‘folk’ explanation, both groups experience conflict.

In other words, what’s processed quickly, automatically and pre-consciously is familiar, overlearned information. If that familiar and overlearned information consists of incomplete and partially understood bits and pieces that people have picked up as they’ve gone along, errors in their ‘folk’ psychology, biology and physics concepts (p.13) are unsurprising. But it doesn’t follow that there must be dedicated modules in the brain that have evolved to produce those concepts.

If the familiar overlearned information is, in contrast, extensive and scientifically accurate, the ‘folk’ concepts get overridden and the scientific concepts become the ones that are accessed quickly, automatically and pre-consciously. In other words, the line between biologically primary and secondary knowledge and abilities might not be as clear as Geary’s model implies.  Here’s an example; the ability to draw what you see.

The eye of the beholder

Most of us are able to recognise, immediately and without error, the face of an old friend, the front of our own house, or the family car. However, if asked to draw an accurate representation of those items, even if they were in front of us at the time, most of us would struggle. That’s because the processes involved in visual recognition are fast, frugal, simple and implicit; they appear to be evolved, modular systems. But there are people can draw accurately what they see in front of them; some can do so ‘naturally’, others train themselves to do so, and still others are taught to do so via direct instruction.  It looks as if the ability to draw accurately straddles Geary’s biologically primary and secondary divide.  The extent to which modules are actually modular is further called into question by recent research involving the fusiform face area (FFA).

Fusiform face area

The FFA is one of the visual processing areas of the brain. It specialises in processing information about faces. What wasn’t initially clear to researchers was whether it processed information about faces only, or whether faces were simply a special case of the type of information it processes. There was considerable debate about this until a series of experiments found that various experts used their FFA for differentiating subtle visual differences within classes of items as diverse as birds, cars, chess configurations, x-ray images, Pokémon, and objects named ‘greebles’ invented by researchers.

What these experiments tell us is that an area of the brain apparently dedicated to processing information about faces, is also used to process information about modern artifacts with features that require fine-grained differentiation in order to tell them apart. They also tell us that modules in the brain don’t seem to draw a clear line between biologically primary information such as faces (no explicit instruction required), and biologically secondary information such as x-ray images or fictitious creatures (where initial explicit instruction is required).

What the experiments don’t tell us is whether the FFA evolved to process information about faces and is being co-opted to process other visually similar information, or whether it evolved to process fine-grained visual distinctions, of which faces happen to be the most frequent example most people encounter.

We know that brain mechanisms have evolved and that has resulted in some modular processing. What isn’t yet clear is exactly how modular the modules are, or whether there is actually a clear divide between biologically primary and biologically secondary abilities. Another component of Geary’s model about which there has been considerable debate is intelligence – the subject of the next post.

Incidentally, it would be interesting to know how Sweller developed his summary because it doesn’t quite map on to a concept of modularity in which the cognitive skills are anything but generic.

References

Fodor, J (1983).  The modularity of mind.  MIT Press.

Geary, D (2007).  Educating the evolved mind: Conceptual foundations for an evolutionary educational psychology, in Educating the evolved mind: Conceptual foundations for an evolutionary educational psychology, JS Carlson & JR Levin (Eds). Information Age Publishing.

Acknowledgements

I thought the image was from @greg_ashman’s Twitter timeline but can’t now find it.  Happy to acknowledge correctly if notified.