Thoughts on Brain Training

LegoComputerLast Monday I had the opportunity to participate in a response panel as part of the Cottage Rehabilitation Hospital series on Empowerment Through Medical Rehabilitation. This year they brought Dr. Michael Merzenich in from UC San Francisco. Dr. Merzenich is an expert on neuroplasticity and is also the founder of several companies dedicated to cognitive enhancement through training. He is a research powerhouse, with hundreds of publications and a CV that is probably taller than I am. His most famous startup is likely PositScience, which you can find at the site

Dr. Merzenich gave a great 80-minute talk to a crowded room. His talk was pretty similar to others that I have watched on YouTube, so if you are curious about his research you might head over there. Here is a link to his Google TechTalk, which was quite good:

I had a whole series of points ready to bring up to try and foster some lively discussion during the panel. Due to time issues the panel was only about 20-25 minutes long. This didn’t leave much room for lively discussion after introductions were done. Good material for a blog post though…

What follows are some elaborations on the points that I had prepared for the response panel. They are all my opinion, occasionally backed by specific citations and research. They are deliberately intended to be controversial, which will hopefully make for interesting reading.

* Our brain function and structure certainly reflect their environment

Developmental psychologists and neuroscientists have known for some time that the brain was plastic and responds to its environment. After all, we research a period of life that is highly plastic, with a four-year-old brain having 10x as many connections as it will later in life. Changes in the brain of a child are sculpted by a mixture of genetic predisposition and environmental context. That is, some changes are hard-wired and will happen no matter what, while other changes are in response to the environment.

Even later in life it is recognized that the brain can change its structure and function and adapt to its surroundings. Among others, milestone neuroimaging studies investigating structural differences in the brains of London taxi drivers (Maguire et al., 2000) and changes following learning to juggle (Boyke et al., 2008) validated everything that we already knew: the brain is not a static organ. In that context brain training isn’t a revolution in our understanding of the brain, but a new method to mold that plasticity in the direction of our choosing.

* Everything, even your breakfast, changes your brain

The vast majority of functional neuroimaging is done with a defined cognitive task. The investigator will look at the relative difference in activity between different conditions of the task to determine indices of activity. A more recent investigative method, resting state imaging, is different. The subject is given no instructions, other than to sit quietly in the scanner and clear their head to the best of their ability. From the data collected during this scan you can begin to look at which brain areas have correlated activity, indicating that they tend to work together. By doing scans before and after a specific intervention,you can then learn about which functional networks are affected by treatment. So, what have we learned from these studies?

Everything changes the brain. Everything.

The Neuroskeptic had a great argument in a recent blog post, which is that even the act of opening your eyes can have a dramatic influence on the state of the brain. This can be seen even in commodity EEG equipment as a change in power in the alpha band. He was discussing a recent study that found a single dose of an antidepressant can have a significant effect on functional connectivity in the brain (Schaefer et al., 2014).

The challenge to neuroscientists is to separate the wheat from the chaff. If everything changes the brain, then which changes are actually relevant? Much of the brain training research does a good job of this, looking for measurable changes in behavior following training. As for the rest of neuroimaging, well, I have a feeling that we are going to see a lot more “crossing the street changes the brain” type of articles…

* Cognitive capacity is a mix of genetic and environmental factors

There seems to be an approximate 50/50 split in terms of what determines your intelligence. About half comes from your parents in the form of your DNA. The other half comes from the sum total of all experiences you have ever had in your life, including brain training. This is great, because it means that we have a lot of leeway to push our cognitive abilities around. Make positive changes and you will see improvements, make negative changes and you will see degradation. A second point should be made here though, which is that each of us carries genetic predispositions that will be difficult or impossible to change through training.

One extreme example of genetic factors at play is in psychopathy. There is evidence that it is difficult or impossible to rehabilitate true clinical psychopaths. I never got a chance to ask Dr. Merzenich about this counterexample. I have to believe that he would say that even clinical psychopaths could benefit from brain training. I have a different view, which is that a certain percentage of our cognition is hard-wired, and may or may not be able to be changed. Some of us have burdens that cannot be trained away.

* Individual differences can create very different training outcomes

We typically look at the results of an fMRI study, see the blobs, and immediately assume that the entire group of research subjects must have been using those areas of the brain. This is somewhat true in terms of the statistics, but there is a richer story to be understood. First, some subjects do not show increases in activity, and some even show reductions in activity! Further, the patterns of activity across the rest of the brain will often be wildly different from one individual to another. My postdoc advisor, Michael Miller, has done a lot of research on this, going so far as to call each person’s pattern of activity a “neural fingerprint”.

What does this mean for brain training? Well, if everyone is using a highly varied mix of functional networks as they navigate a cognitive task then we would likely expect to see a highly varied response to brain training as a consequence. For example, if you leverage executive working memory to a greater degree than I do while completing a task and we then complete a working memory brain training program then you would likely demonstrate a greater benefit than I would.

One issue I had with Dr. Merzenich’s talk was his assertion that every ability could be improved at any age. This is a very strong argument. He cited an animal study which showed that cognitive training in rats led to a range of behavioral and physical improvements, right down to improved immune function. Knowing that each individual human could have a different response to each form of brain training, I would put forward that most cognitive abilities could be improved using brain training. I would further argue that there will be a varied generalization of brain training to other tasks that remain untrained.

* The relevant networks for training might be counter-intuitive

It will not always be immediately apparent which functional networks in the brain should be targeted with brain training to derive a practical cognitive benefit. This will be particularly true as you try to train cognitive abilities that lie higher and higher in the hierarchy of function. For example, how would you train an adolescent to better handle risky situations? The common refrain here would be to train inhibitory control, so that the teen could better inhibit a prepotent response. That may have some benefit, but evidence is accumulating that one representation of social risk is modeled by an ‘as-if’ loop, which attempts to predict body and emotional state following a decision (Preuschoff et al., 2008). It may be that something like mindfulness meditation may be more beneficial here relative to brain training (but I am only guessing).

* Why aren’t more people using brain training?

Dr. Merzenich suggested that there were immediate benefits following brain training that lasted for years, transferred to other cognitive abilities, and had a significant effect size. He argued that brain training was the logical path to enhanced cognitive capacity and cognitive rehabilitation. Sounds persuasive, yes? If all of that is true, then why isn’t every hospital and school clamoring for access to the training programs? Insurance companies aren’t fools, contrary to the opinion of most Americans. If they can obtain the same benefit from $1k of brain training instead of $10k of traditional treatment, do you not think that they would be pushing doctors to prescribe it? What’s the hangup?

* Social interaction as a highly complex cognitive task

I thought that one subject that didn’t get a lot of play in Dr. Merzenich’s talk was the cognitive requirements of social interaction. Social interaction is not frequently viewed as a highly complex cognitive task. That is largely because we are so damned good at it. You can model the current internal state of multiple individuals while simultaneously predicting how that internal state will change based on some tidbit of information that you are about to communicate. That is simply incredible. There is a reason that it takes us over two decades to become ‘adult’ in terms of our behavior, and a large part of that is having the processing capacity and social experience necessary to interact with others as an adult. It takes way more than 10,000 hours…

I bring this up mostly because I think that social interaction is a critical part of our daily lives, and I see it as critical during the rehabilitation process as well. Which patient has a better overall outcome: the one who does brain training alone at home twice a week or the one who sees a therapist for training twice a week? My money is on the latter. I say this not only for the obvious emotional benefit of social interaction, but because navigating the social landscape is one of the most cognitively difficult tasks that we, as humans, accomplish.

* The need to define a ‘challenging’ cognitive task

I think that much of the benefit that comes from the brain training has relatively little to do with the tasks that are presented and more to do with the progressive difficulty of the software that drives them. One major point in Dr. Merzenich’s talk that deserved additional attention was that there was very little benefit to be had with brain training if the participant is not adequately challenged. That means that the task cannot be too difficult, causing frustration, and cannot be not too easy, meaning no benefit. Instead, you want the participant to get the majority of trials correct (~80%) and still have room to grow.

When I was in grad school my roommate dropped me in an MRI scanner for 2.5 hours to complete a mildly complex task. To answer your questions, yes, I did that task for over two hours and, no, it was not fun at all. What the data showed at the end was fairly amazing though. At first, when the task was difficult, I had to utilize a range of frontal and parietal resources to complete the task. It took effort. Two hours later, when I didn’t even have to think about the task anymore, the majority of activity was subcortical. It had become automatic.

We may not get a lot of cognitive benefit out of social interaction because we are already experts in social interaction. Likewise, when you first start doing crosswords everyday you will likely see a cognitive bump, but as you get better at crossword puzzles the marginal increase you see will diminish. This is the secret sauce of brain training. If you had a set of crossword puzzles that increased in difficulty as your skills improved I think that you would continue to observe improvements over time.

* Brain training seems to have mixed results

First, I haven’t read the entire body of literature on brain training, so I am limited to the handful of papers that I have reviewed. One consistent thread that I have observed is that the training can offer improvements that confer long-term benefits on a scale of many years. Still, the effect size of the studies I reviewed isn’t always huge.

A recent study by Wolinsky et al. (2013), showed an effect size of around 0.25 for most of their training, as indexed by the Cohen’s d statistic. This was across several experimental conditions one year after the initial training.

Another, larger study from the NIH found that cognitive benefits can stay around for over a decade. The Advanced Cognitive Training for Independent and Vital Elderly (ACTIVE) study followed 2,832 participants for ten years following an initial six-week brain training program (Rebok et al., 2014). After ten years approximately 70% of subjects in the training group were above their initial performance baseline. This is relative to 50-60% in the control group. The effect sizes varied depending on the condition, but ranged from approximately 0.20 to 0.50.

Head on over to the Cohen’s d effect size interpreter at and look at the overlap of the treatment and control group distributions for a d of 0.2. Yeah, there is a benefit, but there is not a dramatic shift. Effect sizes of 0.5, which were observed in the ACTIVE study, are better. Still, you are not jumping several standard deviations up the curve.

The authors of both studies discus that the practical effect of brain training is a 2-7 year delay in certain aspects of cognitive decline. Again, not super huge. You are not turning a 70-year-old brain into that of a 30-year-old. Still, when I am 70 I will take any improvement I can get.

* Exercise and nutrition as global cognitive benefits

It is no secret that I am a huge fan of exercise and mindfulness meditation as ways to improve your brain. Many of the points in my impromptu post on running still ring true, and there is an increasing body of evidence that exercise, nutrition, meditation, sleep, and medication can improve the function of the brain. Dr. Merzenich accepted that these had positive influences, and mentioned them early in his talk. Still, he was somewhat dismissive of their ability to significantly improve cognition or aid in rehabilitation, saying that they formed part of a “complete package” of rehabilitation.

I took strong issue with his stance on this issue. He stated that exercise and nutrition provide “essentially no benefit” relative to brain training on the tasks that they had investigated. I have a hard time completely believing that, especially given the range of benefits that have been published in scientific literature.

* Warning against false prophets of brain training

Brain training has a huge buzz right now, and a number of startups have been founded provide brain training tools. PositScience is not alone in a market that includes Lumosity, Cogmed, CogniFit, LearningRx, and many others. Which ones really work?

How many programs are just peddling mildly difficult games with an advertisement that it improves the brain? I sleep better at night knowing that Dr. Merzenich is the Chief Scientific Officer of PositScience, but how many companies do not have a solid scientist at the helm, or the research to back up their claims?

* Realtime neuroimaging techniques could play a role in the future

Brain training certainly has some effect. Further, there is very little downside to using brain training tools. The only possible negative I could see is the opportunity cost. We all have a finite amount of time during the day, and if you are in your back office engaged in brain training instead of exercising or interacting socially then you may be missing out on other ways to improve your cognitive function.

I am most excited about the market for realtime feedback on cognitive performance based on neuroimaging data. Christopher deCharms has been doing a bit of this using realtime fMRI to help control pain in patients. He founded a startup, Omneuron, to investigate this technique and capitalize on it. What I would like to see is a consumer-grade EEG system (like the Emotiv headgear) that integrates realtime brain state with brain training. I think that this is really the Holy Grail, because you are no longer basing the progressive difficulty of a task on the behavior of the trainee, but instead you could potentially increase difficulty based on the actual cognitive load.

Perhaps it is time for me to start up a company of my own…

Leave a Reply