Email Icon Facebook icon Twitter Icon GooglePlus Icon Contact

User Top Menu

Mental Training Evokes Superior Memory, Transforms Brain Networks

19 Mar 2017

by Gwyneth Dickey Zakaib

This article appears by special arrangement with Alzforum. See original article with additional links/commentary.

At first blush, extraordinary memory seems to require genius-level smarts. However, people who quickly memorize hundreds of random words or recall the exact order of a deck of cards in less than a minute may not have been born with superior brains, after all. In the March 8 Neuron, scientists led by Martin Dresler, Radboud University Medical Centre in Nijmegen, the Netherlands, and Michael Greicius, Stanford University, Palo Alto, California, report that a unique pattern of brain network connectivity found in top memory athletes can be evoked in ordinary people after just six weeks of intense memory training. The results suggest that the average person can learn to perform extraordinary memory feats, and that doing so transforms connections in the brain. Researchers are unsure if such training might help treat or prevent dementia.

“What’s exciting about this paper is that not only are they showing something special about brain network connectivity in memory athletes, but that they could achieve a similar pattern in trained naïve subjects,” said Emily Rogalski, Feinberg School of Medicine at Northwestern University, Chicago. “It’s a very clever and well thought-out study for exploring the ways in which [neural] plasticity can manifest.”


Despite decades of study, much of it in people who had specific brain lesions that caused amnesia, the mechanics of memory are still unclear. To get a better handle on it, Dresler and colleagues decided to study people with superior memories. Researchers have rarely examined brain changes underlying top memory performers, though two small functional MRI (fMRI) studies examined brain connectivity patterns in memory athletes who used a prominent mnemonic technique called the “method of loci.” In a nutshell, the athletes map a real-world route in their mind. Then, into points along that route, they mentally tuck information that they can recover as they retrace their steps later. The mnemonic device has helped world champions achieve such mental feats as memorizing the order of a deck of cards in fewer than 20 seconds, or 480 random words in five minutes. For comparison, people taking the widely used ADAS-Cog test are asked to remember just 10 words in about 20 seconds. 

The MRI scans suggested very few structural brain differences between controls and either top performers in the World Memory Championships or people recently trained on the method of loci. However, users of the technique activated more neurons in visuospatial areas and the hippocampus during a memory task (Maguire et al., 2003; Nyberg et al., 2003). Neural networks incorporating these regions are key for spatial learning.

To examine these connectivity patterns further, Dresler and colleagues, including co-first authors Boris Konrad and Nils Müller from Radboud, and William Shirer at Stanford, gathered 23 of the top 50 memory athletes in the world and studied their brains by structural and functional MRI. They compared them to 23 controls matched for age, sex, and IQ. During the fMRI scan, athletes and controls completed a memory task, in which they saw a series of 72 words for two seconds each, and later had to recall as many as possible. The athletes remembered almost every one, while controls could only manage about 40. The researchers then calculated the connectivity between various brain areas and 71 regions in brain networks associated with memory and visual processes, including the medial temporal lobe. They constructed a matrix of hundreds of connectivity differences that distinguished athletes from controls.

To determine whether these differences were innate or came from training, the researchers then recruited 51 new volunteers, average age 24, who were naïve to any mnemonic strategy. During a baseline fMRI, they completed the same 72-word memory task. Thirty-four volunteers then underwent six weeks of training—17 on the method of loci, and 17 practicing a working memory task without using any specific strategy. The remaining 17 people received no training at all. Over six weeks, the method of loci-trained group more than doubled the number of words they recalled. They did almost as well on the same task after four months without any additional practice. In contrast, the active and passive control groups saw little to no improvement.

At the end of the training period, all again underwent fMRI imaging during the word memory test. The researchers then calculated the connectivities among the same 71 regions, and found a tight correlation between the method of loci-trained group and the memory athletes. The most robust changes seen in both groups involved stronger connections between two specific network hubs, the dorsolateral prefrontal cortex and the medial prefrontal cortex, and with regions of the DMN, visuospatial network, and the MTL, such as the left parahippocampal gyrus, bilateral retrosplenial cortex, posterior cingulate cortex. Other connections were suppressed, such as those made with the angular gyrus. In general, regions toward the front of the brain strengthened in connectivity, while connectivity in the back weakened. The more the trainees’ connectivity resembled the pattern observed in athletes, the more words they recalled. By contrast, connectivity changes in neither control group correlated with the pattern exhibited by memory athletes.

Taken together, the results suggest that the superior memory afforded by the mnemonic training in controls associates with a unique functional connectivity signature shared by top memory athletes. It also supports the idea that memory is not just contained in one or a few brain structures, as it has been thought in the past, said Greicius. “It’s not just the hippocampus that’s important, but a more distributed memory network,” he said.

Do any of these findings have implications for Alzheimer’s disease or other forms of dementia? In the last few years cognitive and memory games have become increasingly popular among older adults, with little indication that they actually help stave off cognitive decline. Whether the work has implications for cognitive impairment in old age remains uncertain, Greicius said. The group now plans to delve deeper into the data they collected, including analyzing the network processes involved in memory retrieval.

“It has never been shown this elegantly that brain organization can be altered,” said Prashanthi Vemuri, Mayo Clinic in Rochester, Minnesota. That these researchers could reproduce a signal seen in memory athletes by training controls was remarkable, she said. While the data hint that brains may be more easily “rewired” than has been thought possible, a prospect that might bode well for people with neurologic disease or even dementia, it has yet to be seen if older individuals or those who have amyloid or vascular pathology would be able to rewire their brains as easily, Vemuri noted. Some literature suggests practice effects are weaker in people who have such pathologies, she said. Still, Paul Frankland, Hospital for Sick Children, Toronto, was struck by how long-lasting the effects were. “That reminded me how plastic the brain is in a non-pathological state,” he said. However, he agreed with Vemuri that network changes could be harder to realize as the brain ages and deteriorates.

Li-Huei Tsai, MIT, Cambridge, wondered whether this type of training alters normal oscillations in the default mode network, which has been associated with amyloid production, and whether those changes might decrease amyloid in this region. She recently reported that synchronizing neurons to γ-frequency oscillations reduced amyloid production in a transgenic mouse model of AD. 

David Jones, also of the Mayo Clinic in Rochester, said that the pattern of network changes in those with superior memories bears striking resemblance to one he sees in aging brains and in people headed toward dementia, namely, increasing connectivity in the front of the brain and decreasing connectivity at the back (Jones et al., 2011; Jones et al., 2016). It may seem counterintuitive, he said, but it supports the idea that a functional pattern that helps memory may compensate in AD as medial temporal lobe-based systems deteriorate. “It was surprising just how similar the patterns were,” he told Alzforum. “I don’t think that’s a coincidence.” Greicius agreed that there is growing support for the idea that increases in frontal connections may compensate for reduced MTL and posterior parietal connections in AD. Jones is unsure how activating this pattern early on would affect risk of the disease. “The question is, what healthy network pattern might yield the best long-term outcome?”