Many companies are pushing you to buy their brain training software as a way to sharpen cognitive skills and ward off mental decline, but does this work? The question remains unanswered, but a study in the September 5 Nature provides some encouraging signs. Researchers led by Adam Gazzaley at the University of California, San Francisco, designed a video game that challenges people to steer a digital car while watching for road signs. Players have to pay attention to several stimuli at once. After 12 hours of this gameplay, cognitively healthy older adults dramatically improved their multitasking ability over control groups, and the boost persisted for up to six months, the authors report. The treatment group saw gains in untrained but related cognitive abilities, including attention and working memory. The heightened skills correlated with changes in brain wave patterns indicative of improved cognition. Though the study was small, the authors suggested that this particular game may benefit cognition.

Researchers praised aspects of the study design while cautioning that the results do not show that brain training helps people with daily life. “This is a step forward. I see this as promising exploratory data that is worth replicating in a larger group,” John Harrison at Metis Cognition, Warminster, Wiltshire, U.K., told Alzforum.

This study dealt with normal age-related cognitive change, leaving unclear whether the training would benefit people with cognitive impairment or dementia. Epidemiological evidence consistently links high lifetime cognitive activity with a lower risk of AD (see ARF related news story; ARF related news story) this finding has fueled interest in developing computer games that might delay dementia. So far, however, tests of specific interventions have always fallen short (see, e.g., ARF related news story). Nonetheless, many companies are pouring money into this area and seeking to validate their products in research trials (see ARF related news story).

Gazzaley and colleagues chose to focus on multitasking ability, which has been shown to slip during normal aging. In the NeuroRacer game, drivers on a hilly, winding course must press a button if a green circle pops up on their "windshield," but not a red circle or a green pentagon. The test requires paying close attention to both signs and navigation. The software adapts to keep the difficulty of both tasks at a level where the player gets them right about 80 percent of the time. Researchers assessed the mental cost of multitasking based on how much more slowly players identified signs while driving, compared to when stopped.

First author Joaquin Anguera and colleagues randomly assigned 46 cognitively healthy participants aged 60 to 85 to one of three conditions: the multitasking; an active control group who drove first then identified signs later; and a control group who did neither. When assessed on the multitasking scenario a month later, only the 16 participants who had performed both tasks together improved. While driving, they identified signs 64 percent more slowly at before training, but only 16 percent more slowly after training. They beat 20-year-olds playing the game for the first time—the whippersnappers logged a 37 percent drop in performance while multitasking. In addition, only the multitasking group showed gains on the test of variables of attention (TOVA), a standard neuropsychological measure of attention and on a delayed recognition task that tests working memory. This translation to untrained tasks is important because the benefits of most brain training programs extend no further than the test itself.

Electroencephalography (EEG) taken while the volunteers were ‘racing’ showed that only the multitasking group generated more powerful theta-band brain rhythms. The power of these 4-7 Hertz oscillations in the frontal cortex correlates with attention and working memory. The multitaskers also showed more coherence between frontal and posterior theta rhythms, suggesting better long-range brain connectivity. The extent of the improvement in theta predicted how well participants would maintain gains six months later. Intriguingly, frontal theta power correlates with deactivation of the default mode network (DMN). The DMN activates when people let their mind wander, turns off when people focus on external tasks, and has been linked to AD pathology (see ARF related news story; ARF related news story). The results imply that training on NeuroRacer may improve attention by helping people suppress DMN activity while doing tasks, the authors claim. Other studies have shown that connectivity in the DMN, as well as the ability to switch between functional networks, declines with advancing cognitive impairment (see ARF related news story; ARF related news story).

Gazzaley plans to test NeuroRacer in old people with cognitive impairment and depression, as well as in young people who have attention deficit disorder. In addition, he is developing four other video games to bolster different cognitive skills.

Gazzaley stressed that not all video games will help people. He believes games must be designed to tax a specific ability, and also be targeted to the right population. Meanwhile, Akili Interactive Labs in Boston, Massachusetts, a company Gazzaley co-founded, will develop NeuroRacer into a commercial product that could be used as a medical intervention for specific conditions, Gazzaley said.

Commentators noted that the use of an active control group and EEG, elevate this study over much of the previous work in the cognitive training field. “The changes in brain waves are consistent with the changes in cognitive performance. This is very exciting,” John (Wes) Ashford at Stanford University, California, told Alzforum. Others found the gains in untrained skills and the maintenance of benefits promising. Other studies, such as the Advanced Cognitive Trial for Independent and Vital Elderly (ACTIVE), have shown that improvements in skills targeted by cognitive training can last for years. Still, this does not necessarily translate to any benefit to daily functioning (see ARF related news story).

Almost one in four participants dropped out of the study, and commentators noted it must be repeated in a much larger group. Above all, it is unclear if the results have any relevance for Alzheimer’s disease, Ashford said. He pointed out that the intervention mostly acted on frontal lobe functions, while AD primarily attacks the posterior temporal lobes. Gazzaley agrees that more work is needed. “It’s a nice proof-of-concept, but it’s only the first step,” he told Alzforum.—Madolyn Bowman Rogers.


  1. This is a well-designed video game training study reporting improvement in cognitive control in older adults. It was conducted by a well-established and respected group of investigators. The findings, as the authors acknowledge, are preliminary, given the limited sample size of 46 healthy older adults with 16 receiving multitasking (MMT) training and 15 receiving single task (SST) training.

    Some strengths of the study include: 1) the baseline study conducted prior to training study to examine performance levels of two key tasks across a wide age range; 2) development of an adaptive staircase algorithm that guided the adaptive training paradigm; 3) examination of transfer to working memory and attention tasks; 4) 6-month follow-up; 5) inclusion of an active training condition that involved training on 2 single tasks (signs, driving); and importantly the examination of neural activity related to training improvement.

    However, to review and evaluate the training findings, it is important to understand the actual focus of the training and what types of transfer measures were examined. From the Nature article and supplementary material, the primary focus of MMT was on the speed component of the multitasking performance of the subjects, not to improve accuracy on multitasking per se. Prior to training, thresholds for the sign and drive tasks were determined for each individual such that each participant would perform at 80 percent accuracy at baseline. That accuracy is a high baseline criterion, and is often the criterion set for end of training success for studies focusing on improvement in accuracy on a cognitive task. The supplementary material indicates that several subjects were not enrolled due to inability to meet the 80 percent accuracy criterion at baseline. In future work with MCI or dementia subjects, a focus on improvement in accuracy of multitasking, rather than solely on speed of multitasking would need to be considered.

    Given the focus on improved speed of performing multitasking, it is not surprising that the working memory and attention tasks that did show transfer effects also had reaction time as a dependent variable. The tasks also involved recognition rather than recall in working memory. Many of the standard working memory (updating) tasks focus on accuracy and recall, not on reaction time solely.

    Training on a simulated driving task added some real world value to the findings and appears to have increased participant motivation and interest in the training. However, future research would need to validate the training task with more direct measures of driving performance.

    As in much cognitive training research, the subject sample is positively selected. Notably, subjects who were on hypertension medications were excluded (Methods Summary), which would result in exclusion of a considerable portion of cognitively normal elderly.

    The design of the baseline study that compared performance at different ages is cross-sectional is cross sectional and thus the findings show age differences in multitasking cost, not intra-individual longitudinal change. The slope (Fig 2) is steeper at age 40 - 60 years than reported by longitudinal studies that have examined working memory with accuracy as the dependent outcome in cognitively normal elderly.

    The examination of neural correlates of training improvement is to be highly commended. Focus on the default mode network and the medial prefrontal node has the potential to contribute to current research on the default mode network as an early measure of preclinical dementia.

    Lastly, the maintenance of training effects at 6 months is encouraging. The ACTIVE trial has recently reported maintenance of training effects at 10-year follow-up (Rebok, G. et al., in press).

    View all comments by Sherry Willis

Make a Comment

To make a comment you must login or register.


News Citations

  1. Not Just Fun and Games—Cognitive Engagement Deters Amyloid
  2. Add Mental Exercise to Potential AD Protection
  3. Brain Training Falls Short in Big Online Experiment
  4. Tech Revolution: Behavioral and Cognitive Interventions
  5. Tracing Alzheimer Disease Back to Source
  6. Cortical Hubs Found Capped With Amyloid
  7. Network News: Images of AD Brains Reveal Widespread Snafus
  8. Communication Breakdown: Multiple Networks Decline in AD Brains
  9. The ACTIVE Trial—Long-Term Effects of Cognitive Training

External Citations

  1. Akili Interactive Labs

Further Reading

Primary Papers

  1. . Video game training enhances cognitive control in older adults. Nature. 2013 Sep 5;501(7465):97-101. PubMed.