Science / Fiction
HOW LEARNING STYLES BECAME A MYTH
By Carol Black
In case you missed it, research has now shown that (spoiler alert) LEARNING STYLES DON’T EXIST!!! Most people believe they exist, of course –– silly people! –– because, based on their own direct experience, most people find that either they really need to see a thing to understand it, or that actually it's much better if you just explain it to them, or that they wish you would please be quiet so they can read the instructions and go through it step by step, or that they’d rather get a root canal than read the instructions and they just have to fool around with it until they figure it out.
But this common sense perception on the part of the great unwashed population, according to education researchers Paul A. Kirschner and Jeroen J. G. van Merriënboer, is not merely “unsupported by data;” it is an “urban legend” on an intellectual par with the belief that there are alligators in our sewers.
Multiple frameworks for understanding learning styles have been proposed, from the popular Visual-Auditory-Kinesthetic framework, to the Concrete-Abstract framework, to the Holistic-Analytical, Impulsive-Reflective, Left Brain-Right Brain, Convergent-Divergent, Cognitive-Affective-Physiological –– one literature review identified 71 different models. As Kirschner and van Merriënboer grouse, if we consider each learning style as dichotomous (e.g. visual vs. verbal) that means there are 2 to the power of 71 possible combinations of learning styles – more than the number of people alive on earth.
They say that like it’s a bad thing. But if you happen to spend your time hanging around with human beings who are trying to learn things, 2 to the power of 71 different learning styles sounds about right. In other words, as astrophysicist Neil DeGrasse Tyson remarked recently, “In science, when human behavior enters the equation, things go nonlinear. That's why physics is easy and sociology is hard.”
Outside of a school setting anyone can easily observe these differences in action. I’ve spent a lot of time around rock climbers over the years, both kids who are learning and adults with a high level of expertise, in both formal and informal learning situations (in this sport, there’s often not much difference between the two.) Some kids watch the expert climbers like hawks, absorbing and internalizing every move, every gesture; they’ll dip their hands in the chalk bag and then shake off the excess with the exact same wrist flick the older climbers use. Other kids don’t watch. An expert climber can be doing a boulder problem right next to them and they won’t even glance that way. They’re not interested until they get on the rock themselves, feel it for themselves, their bodies exploring and experimenting and testing internally the dynamics of gravity and stone and muscle. One kid will want to talk to her coach the whole time she’s climbing, she’ll be asking for “beta” (information and advice), wanting a steady stream of encouragement and support. Another kid does not want to talk or to listen, and if somebody gives her too much unasked-for beta, she will quickly drop to the ground and walk away. She wants to figure it out for herself.
In other words, while admittedly the differences may be difficult to definitively categorize or to pin down, clearly the kids have different styles of learning. So how does something so intuitively obvious and readily observed cease to exist in the eyes of scientists?
As DeGrasse Tyson suggests, there are intellectual hazards inherent in the effort to turn the study of human behavior into a hard quantitative science. The social sciences are inevitably more subject to ideological bias than the physical sciences; as Noam Chomsky puts it, “It’s not that [physical] scientists are more honest people. It’s just that nature is a harsh taskmaster.” The laws of physics don’t yield to specious argumentation; if you get it wrong, the rocket ship falls out of the sky.
The correlate to this is that in the physical sciences there’s a way you can know things –– to have studies and data and statistical analyses that convince you that you really do know this thing –– and if you build a rocket ship according to what you know, the rocket ship will in fact fly to Mars.
In the social sciences, there’s a way you can know things –– to have studies and data and statistical analyses that convince you that you really do know this thing –– and a decade later not only are you proven completely wrong, but in retrospect it looks batshit crazy that you ever believed this for even five minutes. A good example of this: shortly before the economic collapse of 2008, Columbia University economist R. Glenn Hubbard co-authored a paper asserting that the explosion of derivatives in the financial system had actually "improved the allocation of capital and of risk throughout the US economy," leading to "enhanced stability in the US banking system” so that "recessions are less frequent and milder when they occur."
I think we all know how that turned out.
So what’s going on with learning styles theory? Well, I like to call the phenomenon “squeezing the kitten.” In other words, like a toddler who can’t hold a kitten without squeezing it to death, the education researcher takes hold of a perfectly sensible idea, defines his terms in a way that renders them rigid, narrow, and mechanical, constructs his experimental protocol in a rigid mechanical manner, quantifies his results, tabulates his data and — whoops! — discovers that the kitten was never alive in the first place.
Cognitive psychologist Daniel Willingham, for example, has squeezed the kitten very hard. His thinking, as outlined in a popular YouTube video called (you guessed it) “LEARNING STYLES DON’T EXIST” goes something like this:
1. If “learning styles” exist, then teaching in the student’s preferred modality will always be more effective than teaching in another modality.
2. So far, research has not shown that teaching in the student’s preferred modality is always more effective than teaching in another modality.
3. Therefore LEARNING STYLES DON’T EXIST!!
Sounds simple enough, right?
Let’s step back a minute.
This year scientists say they were finally able to detect gravitational waves, first predicted a hundred years ago by Einstein’s general theory of relativity. They based their claim on an experiment capable of measuring a distortion in spacetime less than one ten-thousandth the size of a single proton –– “equivalent to measuring the distance between the earth and the nearest star with an accuracy of the width of a human hair.”
Daniel Willingham claims that we now understand the human brain well enough to be able to state that LEARNING STYLES DON’T EXIST, based on experiments like this one, described in his YouTube video:
Here’s a visual learner, here’s an auditory learner. Suppose I give each of them two lists of words to learn. One list I read aloud — an auditory presentation — and for another list I show a series of slides — a visual presentation. Later, everybody gets a test. The prediction is straightforward: visual learners should learn the slides better than the words, and the auditory learners should learn the words better than the slides.
Wow. This sounds like a really awesome eighth grade science fair project. But as a way for grown-up adults to make sweeping statements about something as complex as human learning?
The problem here seems to be an error in thinking about what is sometimes called the map-territory relationship. In other words, Willingham is mistaking a finding about an extremely simplistic theoretical map of learning styles for a finding about the astronomically complex territory of learning itself. (He becomes comical at the point where he patiently explains how learning styles theory is refuted by the fact that auditory learners can’t learn the shape of Algeria through their ears, but I won’t even go into that.) The extraordinary difficulty in creating workable theoretical models of complex systems like the human brain has been called “Bonini’s Paradox,” expressed by Paul Valéry this way: “Everything simple is false. Everything which is complex is unusable.” In other words, if we try to create a model with 2 to the power of 71 different learning styles, it’s not really helpful; we’re as confused by our model as we are by reality. But when you simplify all the way down to only 2 different learning styles, inevitably your map fails to capture all of the information relevant to the territory you are trying to navigate. So one of the possibilities is that your model is too simplistic to have any predictive value.
Something you have to consider here is the ratio of the known to the unknown in the domain you are exploring. Columbia University neuroscientist Stuart J. Firestein began teaching a course on scientific ignorance “after realizing, to his horror, that many of his students might have believed that we understand nearly everything about the brain.” Daniel Willingham seems pretty confident that the learning styles question is settled because it has been studied for over forty years. (Remember that it took 100 years to detect gravitational waves.) Is it reasonable to think that in the past forty years we have definitively learned enough about the human brain to establish whether people do or don’t have different learning styles? It’s a bold proposition, to put it mildly. As Michael Smithson, a social scientist who teaches a course on ignorance explains, when we explore previously uncharted territories, another paradox comes into play:
The larger the island of knowledge grows, the longer the shoreline — where knowledge meets ignorance — extends. The more we know, the more we can ask. Questions don’t give way to answers so much as the two proliferate together. Answers breed questions.
So how do we think intelligently about systems that are so complex that our current knowledge is a tiny island in a vast sea of ignorance? Well, a first principle is that a little humility goes a long way. Richard Levins, in a classic essay regarding modeling in population biology, has argued that when dealing with systems with too many parameters to measure, we must inevitably simplify: but we must attempt to “simplify in a way that preserves the essential features of the problem.” How can we differentiate between what Levins calls “legitimate and illegitimate simplifications?” What Bonini’s paradox suggests is that if data generated by an extremely simplistic model seems to defy a substantial body of observed experience, we should go back to the drawing board to make sure that oversimplifications are not leading us to spurious conclusions.
In other words, if the kitten started out alive, we have to make sure we didn’t just squeeze it to death.
Willingham says that learning styles theory predicts that students will always learn better if taught in their preferred style. (This is known as “the meshing hypothesis.”) There are a couple of problems with this right out of the box. Leaving aside the obvious fact that you should pretty much always avoid the word “always” in connection with human behavior, we also have to remember that “learning” is not the same thing as “being taught.” Here’s where the whole argument gets a little squirrelly. Willingham and colleagues admit that people have fairly stable learning preferences. They also admit that people have variable abilities in visual v. auditory memory, etc. (When you combine preference with ability, that’s probably what most speakers of the English language understand by the term “learning style.”) But here’s where a sneaky elision occurs; suddenly the claim becomes that unless we can mechanically match instruction to these variations, LEARNING STYLES DON’T EXIST!! In a literature review of learning styles research, Pashler et al. state it this way: the theory of learning styles is only confirmed if we can successfully sort individuals into groups “for which genuine group-by-treatment interactions can be demonstrated.”
What are “group-by-treatment” interactions? Well, in this scenario instruction is conceived as a “treatment” applied to a learner who is completely passive; the teacher first diagnoses and sorts the learners into groups, applies a randomized “treatment” to each group, and then administers a test to determine which “treatment” worked better –– like a drug trial. The learner has no agency or choice in this; her role is simply to take her medicine as directed. But when you remove the agency of the learner, it becomes unclear whether you are measuring something about the learning or something about the teaching. Anyone who actually hangs around with real human children knows that instruction designed to achieve a certain goal may not actually achieve that goal; in other words, there’s no guarantee that making those “auditory learners” perform a rap song about the Taft-Hartley Act is actually going to be more effective than a regular old history lesson about the Taft-Hartley Act. (Don’t get me wrong; I’m sure there are some awesome rap songs about the Taft-Hartley Act out there.)
But here’s the next problem: excluded from the list of methodologically acceptable studies is anything that involves the kind of creative project-based activities that teachers (or students) might come up with to address the needs of diverse learners (including rap songs about the Taft-Hartley Act.) Why? Because there’s no way to be sure that they aren’t just better or more interesting than regular classroom lessons. From the standpoint of strict scientific method, this is, of course, correct; your experimental protocol should control every variable except the one you are testing. How can you achieve this? By further simplification, of course: by creating a lesson so lacking in complexity that it can’t possibly be interesting to anyone. Like memorizing a random list of words.
Here’s where you run into what cognitive psychologist Frank Smith pointed out long ago; that much education research takes the form of collecting data on people’s ability to learn nonsense. The problem with this is that data about how people memorize a meaningless list of words may or may not have much to do with how they learn in complex, meaningful contexts. Many studies have shown that people (like rats and pigeons) can be induced to perform mechanical learning tasks at a low level in response to rewards or punishments, but that in the long run (for people, at least) rewards and punishments seem to lessen intrinsic motivation to learn. So maybe most people can memorize a short random list of words equally well whether the words are presented visually or verbally. That, in itself, is a marginally interesting minor finding. But what does it really tell us?
It’s important to remember here that we don’t know why people have such strong preferences about learning modality, just that they do have them. One obvious possibility is that if people are able to approach a complex learning challenge in their preferred way, they will be more likely to enjoy the activity and want to continue to higher-order learning, and if not, they will be more likely to find it tedious and unpleasant and want to quit as soon as possible. One neuro-imaging study has suggested that people who self-report as preferring information presented either verbally or visually may mentally convert information to their preferred modality, essentially picturing the words or naming the pictures in their minds. How does this affect their subjective experience of learning? We don’t know. What we do know is that learning can either feel energizing and engrossing, where you are immersed in a “flow” experience of concentration, or it can feel stressful and exhausting, where you have to use enormous will power to force yourself to focus on something you dislike. The former kind of learning is “better,” not because it will necessarily show up as superior on a short test of memorization, but because people who experience the former are more likely to pursue the given activity to a high level over time, because they love it. Of course if the test situation involves learning that no one could love, like memorizing a random meaningless list of words, that effect won’t show up at all. But by the time you claim it has no educational significance, as Willingham and others do, the kitten gets very, very quiet.
Maybe too quiet.
The interesting thing about all of this is that Willingham and friends admit that learning styles may in fact exist; better-designed studies may provide convincing evidence in the future, and of course they know this. Pashler et al. concede that many possible versions of learning styles “have simply not been tested at all.“
So apparently the mantra that LEARNING STYLES DON’T EXIST!! is designed, like the sound bites that emanate from ideological think tanks, to be picked up by the media and repeated, blogged and re-blogged, tweeted and retweeted, to influence the ignorant public. And, of course, to influence policy that affects millions of children.
So why are they selling it this way when they actually know it’s not true?
It’s an interesting question.
The distortion of science for ideological purposes is one of the great social ills of our time; we are all familiar by now with the strategies of those who have attempted to deny the scientific consensus about climate change. But there’s a reverse misuse of science that is currently in play — the claim there is scientific consensus where there is no such thing, that an open, ongoing area of scientific inquiry has reached a settled conclusion, and that anyone who disagrees about this is as irrational and “anti-science” as a climate denier. There are three parts to this claim, which you currently see in the popular media about hotly debated issues ranging from GMO’s and nuclear power to phonics and learning styles:
Claim #1 is that the research is all on one side; that a consensus exists among reputable scientists and that there is little or no empirical support for opposing views.
Claim #2 is that cognitive biases, emotions, denial, irrationality, etc., are what prevent untrained people from accepting this conclusive body of scientific data.
Claim #3 is that the only legitimate path forward is to set aside our childish intuitions and false beliefs and act in rational obedience to research. In other words, data good; intuition bad.
In the case of learning styles, Claim #1 is false on its face: there is no scientific consensus that learning styles are a myth. Noted cognitive scientists like Robert Sternberg and Li-fang Zhang, who have spent decades investigating more complex models of learning styles, remain convinced that they are real and that we should continue refining our models and methodologies in the effort to understand them better. In other words, as Grundmann and Stehr articulate in their excellent article Social Control and Knowledge in Democratic Societies, “the basic rift in such debates is not between lay people and experts but between two alliances that advocate different courses of action based on divergent basic values and knowledge claims.”
Claims #2 and #3 are not only false, but they embody and promote for public consumption a dangerously debased understanding of science and its role in public discourse. In an era when scientific research is often seriously compromised by corporate and political influence, encouraging the informed lay public to submissively set aside their own thoughts, perceptions, and moral and ethical instincts and allow policy to be dictated by the current state of (usually corporate-funded) research is a recipe for further undermining democracy, essentially handing over all power to those who have the resources to fund research while silencing other essential perspectives.
Beyond this, it is a crude misrepresentation of the nature of scientific inquiry. Any serious scientist knows that data and intuition do not stand in opposition to one another, but in a more complex interrelationship; both data and intuition can lie, and both data and intuition can point toward truth. Bias can infect the process of data collection and interpretation at every stage of the process, and intuition can be the product of a thousand empirical observations; that’s why active, creative research scientists so often rely on hunches and instincts to generate new testable hypotheses. Ultimately it is the intellectual ability to hold inconclusive and sometimes conflicting information, observations, and intuitions in an open mind which enables genuine scientific inquiry to proceed. As Max Planck, originator of the theory of quantum mechanics, put it:
Again and again the imaginary plan on which one attempts to build up order breaks down and then we must try another. This imaginative vision and faith in the ultimate success are indispensable. The pure rationalist has no place here.
That’s not being “anti-science.” That’s science.
There’s an Orwellian chill to the spectacle of cognitive scientists reasoning from their own inability to meaningfully address the differences in our children to a finding that our children don’t have meaningful differences after all. (In fairness, Willingham admits that our children do have differences –– in ability. Some of them are smarter, and some are dumber. So there’s that.)
Why does this matter? Because every year our schools fail millions of healthy intelligent children who don’t learn well through normal classroom methods. These kids are not stupid, and the way they are made to feel stupid, day after day, year after year, decade after decade, is child abuse. Let me repeat that: it is child abuse, occurring on a mass institutional scale. When you look at the lifelong psychological, social and economic damage done to intelligent people who do not learn well in schools, you quickly realize that finding better ways to cope with human learning differences is not an academic question, but a problem of the utmost moral urgency.
So cognitive scientists have found that presenting a memorization task or a test of verbal comprehension either visually or auditorily does not solve this problem?
Our children have complex, multi-dimensional learning differences that require complex, multi-dimensional solutions — solutions that go well beyond “sorting” them into groups so that we can apply different “treatments” to them. Ultimately there is something deeply unsettling in the fact that anyone would look at human children and human learning in such a schematic, diminished fashion. But there is a long history of schematic, diminished thinking in the cognitive and behavioral sciences, from intelligence testing to Skinner boxes. The scientists today who proclaim that LEARNING STYLES DON’T EXIST are reminiscent, in both their arid logic and their cold, haughty tone, of the scientists of decades past who dismissed Jane Goodall for “anthropomorphizing” when she insisted, based on her own observations and intuitions, that chimpanzees feel affection and loneliness and joy and grief very much as we do.
Those were the silly intuitions of a soft-headed untrained woman, these men of science decreed; any rational scientific person knew that animals don’t have emotions. Why, they don’t even feel physical pain as we do!
Of course, in retrospect, the fact that scientists ever believed that for even five minutes looks batshit crazy.