Found here:


From the author:

 Quick notes:

  • If you experience any technical problems running the game, please let me know what browser and OS you're using.
  • If you're unfamiliar with what Dual N-Back is, [1] this is a good place to start reading.



New Comment
9 comments, sorted by Click to highlight new comments since: Today at 4:45 PM

A decent (randomized, has placebo control) recent article finding no transfer here (pdf). Not saying this is the final word on dual-n-back but enough for me to spend time elsewhere until more evidence comes in....

That was an interesting read, thanks. But I laughed out loud when they explained how they increased the IQ variance of their sample. The original study worked with students from the University of Bern, a group too homogeneous with respect to intelligence. To increase diversity, the replication works with students from Georgia Tech, Georgia State University, and Michigan State University. That is playing to the stereotypes.

There's still at least 1 more study coming out, one involving Stephenson (whose 2010 thesis was one of the strongest datapoints, according to my meta-analysis). I've been told that it found transfer to fluid intelligence when using visual single n-back, but not audio n-back.

I wonder if a DDR version of Dual-N-Back could be devised?

It could! Exactly like the normal DDR, except you only stomp a pad if it's the same direction as N-back arrow was. The distribution of arrows would have to change somewhat.

Some context: Several frequently cited studies on working memory training using dual n-back, most famously Jaeggi et al. 2008 strongly indicated that WMT could reliably produce lasting benefits to fluid intelligence. These studies obviously provide great material for marketing cognitive training software. See for instance this page by Lumosity:

In 2008, Dr. Susanne Jaeggi, Dr. Martin Buschkuehl and colleagues at the University of Michigan showed that cognitive training with a task called Dual N-Back enhanced fluid intelligence – the ability to creatively solve new problems, and a critical component of IQ. This study involved healthy young adults, mostly university students. After as little as eight hours of training, young adults who trained saw significant gains in fluid intelligence and working memory. We have worked with the Michigan group to include a version of their training program on Lumosity. In addition, we have created a game-like version of their task called Memory Lane.

However, later studies seem to have found little evidence in support of these claims. For example, the 2012 meta-analysis Is Working Memory Training Effective? concludes:

The absence of transfer to tasks that are unlike the training tasks shows that there is no evidence these programs are suitable as methods of treatment for children with developmental cognitive disorders or as ways of effecting general improvements in adults’ or children’s cognitive skills or scholastic attainments.

To me, it seems that the case for dual-n back exercises actually yielding transferable improvements to intelligence and memory is much weaker after some years of scrutiny. The linked dual-n back game seems like an excellent alternative to existing n-back software. I'm not sure it's worth the time and effort.

My meta-analysis supports the claim that much of the transfer is an effect due to using no-contact control groups.

The absence of transfer to tasks that are unlike the training tasks shows that there is no evidence ...

They wouldn't have even run a study if there was no evidence. Finding no effect according to the study you made and the statistical tests you ran does not make eradicate that prior evidence from existence, and more importantly, it doesn't mean that you wouldn't find evidence if you looked outside the parameters of your study and the particular statistical tests you used.


I don't know why almost every DNB software I can find has some sort of massive, unsolvable problem. The sound doesn't work on this one, for me, if it's there at all.

[This comment is no longer endorsed by its author]Reply

New to LessWrong?