Wednesday, February 21, 2007

The Science of Learning

As Clive Shepherd writes, "cognitive neuroscientist Dr Itiel Dror of Southampton University. Itiel is becoming a bit of a celebrity amongst the e-learning community in the UK as someone who avoids the grand theories of learning and concentrates instead on practical tips based on what we know about the brain and how it works (assuming we really do and this I must place on trust)."

Of course, the sceptical side of me says that this is something like saying that so-and-so can tell us best how to win an auto race because he's a mechanic. There is, indeed, a distinction between knowing how something works and knowing how best to use it.

Reading through the points in thsi summary, they seem sort of right, but not exactly right. Let me clarify them.

  • The brain is a machine with limited resources for processing the enormous quantity of information received by the senses. As a result, attention is extremely selective and the brain must rely on all sorts of shortcuts if it is to cope effectively.

My response: no criticism of this; it seems to be about right.
  • Teachers/designers can adopt two strategies to reduce the risk of learners experiencing cognitive overload: provide less information (quantitative approach) or take much more care about how this information is communicated (qualitative approach).
Well, you see now, this approaches the problem not from the nature of the brain but rather from the nature of the information. And when we look at the information as nothing more than a pile of stuff to be processed by the brain, then sure, these are the ways to deal with it.

But the other way to look at it is to, as promised, look at it from the perspective of neuroscience. What does the brain do in cases of cognitive overload? This is important because, if we know how the brain will adapt, we know how to shape our information (if at all).

This is the subject of the next few points, so I'll continue.
  • It is easier for a person to focus their intention on the desired point if there is minimal noise (other information) surrounding it. Reducing noise also reduces context, so a balance needs to be struck.
I assume he meant 'attention' and not 'intention'. In any case, I'm sure there are all kinds of tests proving this, but I will point out that the nature of the subject is a much more significant variable.

People are able to focus on things even in the most extreme of circumstances if they are sufficiently interested. That's how you can have kids playing video games even while the house is burning down around them (I guess that's the sort of 'context' that would be important). By contrast, if you aren't really interested in what you are doing, the least amount of noise distracts you.

We know this because (as was just stated above) we know that the brain is extremely selective and filters out stuff that isn't important.

Perspective matters. From the teacher's point of view, the content (lessons or curriculum) is constant, while the level of background noise is the variable. From the point of view of the learner, however, the content is also variable. That's why you get two very different interpretations of the same phenomena.

  • Overload can be reduced by grouping items/steps (what Itiel calls 'chunking'). Grouping can be accomplished by placing people/objects/events into categories, or by compressing a number of procedural steps into one, automatic action. Visually you may separate items by space, size or colour. Learners will naturally employ grouping as a strategy, although they may do this inappropriately and the process requires effort. Better for the designer/teacher to present material ready grouped.
This is a good strategy and one I have recommended elsewhere to help people write academic essays easily and proficiently (and without notes, but I digress). I find it interesting, though, that he used the DE design term 'chunking'. Maybe reading something other than neuroscience?

Yes, there are different types of groups. Groups that make sense conceptually, especially if linked to a larger framework, are better (I would add that colour is rarely, if ever, a part of such a framework).

But is it better for the teacher to present the material already grouped? How does that follow? If the intent is to have the student learn the information (ugh, bad terminology) then we must ask, is it the groups that aide remembering and understanding, or the process of grouping that does this? If it's the latter, then presenting the information already grouped may help the teacher remember, but will do nothing for the student.

Because, as I noted above, it is better if the groups align with a pre-existing conceptual framework, it is better then if the student does the grouping, because that way the process allows the student to connect, in an organized way, new knowledge with existing knowledge.
  • A side effect of grouping is that once the action is completely familiar (that old 'unconscious incompetence' phase), the individual finds it hard to explain how they do it; they lose control over the process because it has become automatic (so old hands may not always be the best teachers?). Grouping is essential to our functioning, but there are obvious dangers, i.e. unhelpful stereotyping.
Here there seems to be a confusion between grouping, as in the sense of classifying different perceptual entities into types, and grouping, as in the sense of combining several activities into one. Now this isn't necessarily bad (I have said elsewhere that learning to read is similar to learning to ride a bicycle) but can be very misleading unless carefully explained.

I think it would have been better to present them separately.

There are mental processes that can become automatic. Add 1+1 for example. One of these processes is 'categorization'. You look at a bunch of things and automatically associate some with the others, based on habitually formed patterns of association. In some cases, such as grouping people by colour, this sort of automatic association can be inappropriate.

There are also mental processes that constitute sequences of steps. The steps involved in a logical derivation, for example. So a process that actually involves multiple steps may be performed by an experienced logician as though it were only one step (I called this 'skipping steps' in logic class and complained bitterly about it. "It's obvious," said the professor. "Whaaaa?" I responded).

These are very different phenomena that are essentially the result of the same neural process but which instantiate very differently and need to be approached very differently. Kind of like the way the steering used to recover from a spinout may be exactly the same as the steering required to navigate a hairpin curve. Sure, it's the same motion. But you would describe the two events very differently.
  1. Individuals use top-down processing to reduce overload. This draws automatically on their past experience of the particular context, existing knowledge and intelligence and avoids them having to evaluate all new information from the bottom up. An example would be how people can easily read a sentence in which the letters in each word are jumbled up.
Yes. But...

This is not 'top down' processing as traditionally understood.

There is a very large difference between inferring something on the basis of similarity to a prototype (that is, apttern recognition), and inferring something based on a general principle or rule. By 'top down' we typically mean the latter. But when describing character recognition, as in the example, we are describing the former.

I would also be wary of building the (Darwinian?) intent into the process. People use pattern recognition. It reduces information overload. But it is not necessarily true that people use pattern recognition in order to reduce information overload. People use pattern recognition because that's how neural networks work. Perhaps evolution directed us in this way, perhaps it did not. Either way, our use of pattern recognition in a particular circumstance is not caused by some such intent. It occurs naturally, as though by habit.
  • Designers/teachers need to take account of the way in which the information is likely to be encoded and processed - it's not 'what you teach' but 'what is learned'.
Except... it is very misleading to say that it's 'encoded'. Otherwise, yes, there is a large distinction between what the teacher teaches and what the learner learns (which is why information-theoretic and transmission-theoretic theories of learning are wrong).
  • Different parts of the brain specialise in different tasks. Individuals can engage in more than one task at the same time, as long as each uses a different part of the brain.
Of course, these parts of the brain form dynamically and according to experience and circumstance, so there's no telling in advance, or in general, what processes occupy the 'same' part of the brain and what processes do not.

That's why I can read and write while listening to loud music, as I'm doing now, while my father couldn't.
  • It's a myth that we only use 5-10% of the brain - we use it all.
  • The brain continues to change throughout our lives, even though we stop adding new brain cells in our early 20s. Some parts of the brain are relatively hard-wired (through nature or nurture), some very plastic. It makes sense to concentrate in recruitment on finding those people with hard wiring which suits the job, because no amount of training will sort the problem out later. (Itiel did not go into detail about those capabilities which tend to be hard-wired and those which are more plastic - this is clearly important.)
The main part - that neural nets are plastic - is true and important.

It is also true that some parts are pretty much hard-wired -- good thing, too, or our hearts wouldn't beat and our eyelids wouldn't blink.

But as to how much this carries over into learning or into life skills - this is very controversial. I can certainly agree that there are people with currently existing wiring that may be more or less suited to the job. That's no more controversial than saying people learn different things. But to say that these capabilities are hard-wired is much more questionable.
  1. As you grow older the hard-wired capabilities persist - the most learnable capabilities go first.
This is demonstrably false. For otherwise there would be no incontinence in old people.
  1. Language is more than just a means for expressing thought - in many ways it is thought. If a person is not exposed to any language in early years, then by the age of seven they are incapable of learning it.
I doubt that this is uniquely true of language - it is probably true for any pattern set. Can a person become fluent in mathematics despite never having been exposed to numbers. Can a person become a musician having never been exposed to tone and melody?

People who have not specialized in the nature of language typically take language as a given - some sort of folk-psychological representation of Chomskyian generative grammar. And then suppose that this then must be the nature of thought.

Even if language is thought - which i would not grant for a second - we still know nothing about the nature of thought if we do not agree on the nature of language. Which we probably most emphatically do not.
  • The two sides of the brain really do have different functions (I thought this was just pop psychology). The left brain concentrates on language and analytical skills; the right has the spacial abilities. The left side of the brain controls the right side of the body and vice versa. The left and right sides of the brain do not interact physically.
They do interact, through something called the 'corpus callosum'. But yes, the two sides of the brain do specialize, as observed. That said, to my knowledge, this specialization is not hard-wired.
  • The size of a person's brain is not an indicator of intelligence.
Within the normal variation of human brains, that is. My poor cat, with her cat-sized brain, will never reach human intelligence. But she's still very cute.
  • 20% of your blood is in the brain.
Which stresses the importance of nutrition to brain function.
  • You never lose anything from long-term memory, just the ability to retrieve it. Retrieval is a function of how you encode memories / the number of links you provide.
Well, yeah, in the sense that the connections that constitute the long-term memory are basically permanent, more or less. But you don't 'retrieve' memory the way you retrieve a book from the bookshelf (even though it feels that way).

'Retrieval' (properly-so-called) is a case of pattern recognition - and the less salient a pattern becomes in the mind, the less likely it is to be associated with a current perception.

  • Working memory consists of 7+/-2 items (again I thought this was pop psychology).
Yes. And there's that 'grouping' again, in a third guise. We can remember things that are more than 7 words long by recognizing them as coherent patterns. That's why I can remember something like 'Turn right at the third light and then left at the second stop sign, then go four blocks' even though it consists of 18 words (and 70 letters). or 1-800-857-2020 even though it's 11 digits long.

What we put into working memory first depends on pattern recognition.
  • To reduce cognitive overload, take out every word or picture that is not necessary or relevant to your learning goals. Even then, don't deliver more than the learner can handle (presumably by modularising the learning).
This is effective only at the very gross level, and not particularly useful as you get into finer details (and more precise definitions of 'necessary').

I have seen studies, for example, showing that a slide show with bullet points is more easily remembered than a slide show with the same bullet points and animated graphics.

I expect, though, that the placement of the 'NRC' logo in the corner of the same slides would not have an impact either way.

I also expect that the removal of 'unnecessary' letters in the bullet points would actually hinder memory. Fr exmpl, rmving mst f th vwls. Thy r nt necsry in the snse that we cn stl rd the sntnce, bt they hlp wth the grping.

So - better advice would be something like - present material that accords (perhaps with some cognition) with patterns that will already be familiar to the learner.
  • Provide the learning when it is needed, not before.
Sure. But why? This isn't determined by whether it is necessary or not, but rather, by whether it is salient or not.
  • Be consistent in the manner of your presentation, e.g. the interface.
This can actually be distracting, taken to extremes. That's why documentaries switch from having a person talk to showing some nature scenes with a voiceover to interviewing some other person.
  • Be consistent in the level of your presentation, i.e. not too complex, not too simple. Try to work with homogeneous groups; better still personalise the learning.
Yes, but again, why? I would argue that this facilitates pattern matching .
  1. Engage the learner by grabbing their attention, allowing them to determine their progress, providing constructive feedback, introducing an element of excitement/surprise.
Again, this doesn't follow from the presentation above. What has happened here is that some hackneyed (and vague) pedagogical tips have been attached to some discussion of neural function, without a clear linkage between them.
  • Be careful of allowing the learner too much control over the learning process if they don't have the metacognitive skills, i.e. they don't know what they know and what they don't know, nor how best to bridge the gap. Ideally help learners to increase their metacognitive skills, i.e. learning how to learn.
This has utterly nothing to do with the brain theory discussed above.

If the content has more to do with attention than, say, distraction, then taking control away from learners, even in areas where they do not have skills, may cause more harm than good.

And what are the metacognitive skills. What is 'learning how to learn', for example?

  • Providing the learner with control over pace and allowing them to go back and repeat any step is important.
Same point.
  • The learning benefits by being challenging. Performance targets, rewards and competition can increase the degree of challenge, perhaps through the use of games.
And again, same observation. This doesn't follow from what has been stated above. Sure, it's good advice (what Seymour Papert and James Paul Gee call 'hard fun'). But why is it good advice. What else can we learn about this piece of advice. What kind of games, for example (see Aldrich on this).

Anyhow, those are my thoughts based on this reading of Shepherd's article. I also read a bunch of Dror's publications online and certainly have no quibble with his neuroscience. I just think that the study of teaching and learning involves more than just neuroscience, and that there are areas of complexity and potential confusion Dror may not have considered in his work.


  1. As noted, he's avoiding the grand language of theory... whereas you're analysing his approach based on his language.

    The only way to really tell if he's got the right idea is to observe his actions and note if he's doing the right thing... regardless of what language he uses to explain his approach.

    Language is for communication, and is related to, but doesn't directly correspond with, internal modelling of systems.

  2. How one learns a task is associated with balance. Noise can be considered an inabalance in the structure of the task which is to be a measurement for retention.

    The receipe for learning often reflects groups of items and events that promote the short term memory * i believe* to become long term memory when associated with an event of emotion. Catagories that would be considered sterotyping... are they possibly called a learning curve?

    How events or learning is encoded upon the brain depends upon several key factors that allow for rentention and depending on noise , which can detract from that encoding.

    How you learn is how students should be approaced when putting together the way that they can respond. Knowing your audience is tre importante.

    For example: When asked a yes or no question in a court room. The elements of what was learned from an experience is expected to be answered with a yes or no. Due to the way the question is posed, although the answer seems obvious to others, the learner who has experienced the task and the ability to retentively regurge what was learned, is grouped with an emotion, flash back of what is seen in the minds eye, felt with the physical body, etc... which adds to the distraction of how to communicate this in a simple NO or Yes answer. Conflict developes within those that have cognitive retention or communication element that affect and effect how they are to proceed to present the whole picture of what was learned.

    This would be a measurement in itself, while frustrating others who expect yes or no, or if not allowed to complete the task of cognitively processing what was learned would then cause failure to have the truth prevail of the event and what was learned.

    I agree with Itiel on these key issues because I have observed this response.

    When a questioned is posed under stress. An event which triggered the learning is brought back to mind with triggers mentioned above.

    I know encoding a response is disrupted by surprize and that encoding has a chink in the link. Those that have learning disabilites, or physical disabilties that impact the visualization or tactile infrenences often are substituted by other means of going around the the disruption if the event was a short term encoding event.

    Humor often can replace that minds eye disruption in learning that can be associated with cognitive overload. Letting the learner pick out a series of context of tasks that are available they will gravitate to the strongest and most effective elements where what was communicated can be transformed into a quantive effect with qualatative application.


I welcome your comments - I'm really sorry about the moderation, but Google's filters are basically ineffective.