How David Cameron's yellow Y-fronts gave me an astonishing memory

Hannah Devlin Science correspondent
What’s the best way to learn? The Memrise Prize is an international competition aimed at identifying the optimal strategy for learning. Photograph: Getty Images

A surreal scene is unfolding in my home: David Cameron is sitting at my kitchen table in a pair of lemon yellow Y-fronts. He’s having his hair trimmed by his Ethiopian barber, who wears an Adidas tracksuit. Out in the garden, Trinny Woodall is lying in a toboggan swigging Spanish port from the bottle and Alan Carr is blithely wading into the pond, where we keep the pet piranhas.

No, I am not a modern-day Gatsby. These are the motley inhabitants of my newly constructed memory palace, designed to help me memorise 63 world capitals I didn’t know, for a test set by my editor. Cameron is Cameroon, the yellow undies are Yaoundé and so on (I’ll leave you to guess the rest). By translating the country-capital pairs into vivid images in specific locations within a familiar environment, the facts will be indelibly imprinted on my brain – at least, that’s the aim.

This art of remembering is ancient, with the invention of the memory palace, or “method of loci”, credited to the ancient Greek poet, Simonides of Ceos. But the science of learning is only in its infancy.

A version of the memory palace method has been announced as the winner of the Memrise Prize, an international competition aimed at identifying the optimal strategy for learning, providing compelling new evidence for the power of the technique.

The winning team, a group of scientists from Radboud University in the Netherlands, combined the ancient visualisation method favoured by Roman orators and contemporary memory athletes alike with repeated testing and an algorithm designed to tailor the rate of learning to the individual.

Nils Müller, a member of the Radboud team, says the strategy works by encouraging the learner to map new, potentially boring, facts onto something meaningful. “A completely isolated fact is so hard to learn,” he said. “The method of loci can produce amazing results, especially when learning things in isolation that otherwise would be very hard.”

The competition was launched by Ed Cooke, CEO of the online learning company, Memrise and two senior neuroscientists at University College London, in an attempt to fill what they saw as a knowledge vacuum about the act of learning itself.

“What’s the best way to learn is one of the most fundamental questions that science could ask,” said Cooke, who rates finding an answer to this “as important as anything discovered in physics”.

“There’s no scientific consensus,” he added. “In other areas of human life, there have been big advances in our understanding, but there haven’t really been in human learning. Science has discovered a lot of ingredients, but no one’s really tested a recipe for learning.”

The $10,000 Memrise Prize was designed to fill this gap.

Eleven teams, mostly from world-leading neuroscience laboratories, took part in the challenge of designing the ultimate system for learning 80 Lithuanian words in an hour. Around 10,000 participants were funnelled through the competing methods in the finals, giving the results a level of statistical power that cognitive scientists can normally only dream of.

When tested one week after the initial learning session, participants trained using the Radboud strategy remembered more than twice as many of the Lithuanian words as those taught using the control method of flashcards (nearly 30, compared to less than 15).

Winning strategy

The Radboud team, who this week published a scientific paper linked to their work, began by showing participants a brief video about how to create a memory palace. It features Boris Konrad, a neuroscientist who is ranked 24th in the World Memory Championships and whom Cooke describes as “an old nemesis” from his own time on the international memory circuit.

Simonides, legend has it, came up with the loci method after a near-death experience in which the grand banquet hall in Thessaly collapsed just after he left the building. Some of his fellow diners were crushed beyond recognition, but amidst the rubble, Simonides realised he had near-perfect recollection of who was who based on the seating plan. For most, it would have been a gruesome moment to forget, but for Simonides, it was a serendipitous discovery of how memory can hijack the brain’s navigation system to our advantage.

“From my perspective as a user of the technique, it’s about using something that the brain does really well,” said Konrad. “In our past, location has always been really important. Where is protection, where is food, where is danger, where are our friends?”

Why memory palaces work

Chris Summerfield, a neuroscientist at the University of Oxford whose rival entry into the Memrise contest came joint second, agrees. “The brain has evolved particular ways of dealing with space,” he said.

Normally, as memories are laid down, they are lightly encoded across a disparate network of neurons. But linking a memory to a specific location creates a “sparser” distribution, with memories essentially boxed up by tighter clusters of neurons. According to Summerfield, this helps prevent new memories from overlapping and interfering with each other – essentially providing a filing system for information that is otherwise devoid of context.

“If you’re navigating you really need to be able to distinguish one location from another,” said Summerfield. “It’s a way of putting the memories in unique boxes.”

The creation of images may also have helped users engage with the task – notably, of the five finalists, the Radboud strategy was the one that users reported as being most enjoyable.

In my own task of recalling the capital cities, I discover that coming up with a suitably memorable image is the real challenge – not the act of remembering itself. This transforms the dry task of rote learning into a creative venture.

Optimised spacing

The moment we have learned a new fact, we begin to forget it again. In order to cram the most information in a short time period, simple arithmetic suggests the ideal approach would be to test each fact just as the learner is on the brink of forgetting, making them retrieve and reinforce the memory. There is also evidence that we learn best when tested right up to the limit of our ability.

Rosalind Potts, a neuroscientist at UCL and Memrise prize founder, said this was the approach taken by several of the most successful strategies. “A test that requires you to search your memory for the answer strengthens the memory more than an easier test such as multiple choice where the correct answer is in front of you,” she said.

One method, which used multiple choice tests, did no better than the control learning task, the competition found.

Intriguingly, a strategy based around getting participants to make an incorrect guess at the meaning of the Lithuanian words (before they were given the initial translation) also performed well in the contest. Perhaps due to the counterintuitive nature of their strategy, users rated it as no more effective than the control method, even though, in reality, it was about twice as good.

“There are many examples … of a dissociation between what people believe is effective for learning and what really is effective,” said Potts, adding that her ultimate goal is to help correct such cognitive illusions.

Individualised learning

The Radboud team also used a basic artificial intelligence (AI) algorithm, designed to tailor the presentation of words depending on their difficulty and on a person’s own learning speed.

Gesa van den Broek, the graduate student who led the Radboud entry, said: “We made the spacing of the retrieval trials adaptive with a computational model that predicts how quickly an individual learner forgets and which words are difficult and must be presented more often.”

Summerfield believes the use of similar (slightly more sophisticated) AI was crucial to his own team’s approach, which was ranked closely behind the winning method. “If I throw lots of information at you, you get all flustered; if I drip-feed you stuff to slowly, you’re wasting time,” he said. “I’m sure that’s the classroom experience of many children.”

It is not a great logical leap from this conclusion to a future classroom in which pupils are all hooked up to individualised robotic teachers, drip-feeding them information at precisely the optimal rate. But Cooke is quick to push back on this vision.

“I think that personalisation is much less important than the underlying technique,” he said. “You can choose almost any bicycle in the world and it will help you go five times faster than walking. It’s the design of the learning that makes the difference. Even if you do the same for everyone it could still be a revolution in learning.”

“Three of the entries were twice as effective as the control, which is really quite amazing,” he added. “To double the quality of your learning is quite something.”

Test results

After a couple of hours of cultivation, my memory palace is teeming with bizarre juxtapositions of celebrities, politicians and old school-mates with names like “Chad” that are conveniently matched to my target countries. When I’m tested on the next two mornings, I get 63 out of 63 both times and – I say this as someone who does not excel at quizzes – it’s a breeze. However, scientists say that the palace technique, while a good way to get the facts into your head, is only the first step to truly consolidating the information into long-term storage.

“If you now memorise your groceries on the same route, you will probably overwrite the capitals unless you’ve practiced the capitals enough to become independent of the route,” Van den Broek advises.

Is this really a problem, though? I’m fairly ready to expunge the memory of Steve Bannon (Lebanon) barking (baying=Beirut) outside my front door.

By using Yahoo you agree that Yahoo and partners may use Cookies for personalisation and other purposes