[Bartweiss]

This is interesting, and kind of odd.

The article claims the failure of previous brain-training studies came from choosing the wrong techniques, but the evidence on dual n-back specifically has been conflicted for quite a while; this isn't the first study on it.

What's interesting here is that a lot of the best results on dual n-back in the past have concluded that it's mostly promising to fight cognitive decline or boost recovery. Previous work on boosting memory in young, mentally-healthy people has come up with basically nothing.

I'm hoping that this is the better and more-powerful study, and that n-back does boost general working memory. But the cynic in me is wondering if the chosen working-memory metric was simply more vulnerable to training effects from n-back than from complex span.

https://www.gwern.net/DNB-meta-analysis

[arkades]

My understanding of the literature is largely that n-back training doesn’t generalize effectively, mostly just improving ones ability to perform n-back exercises.

[Bartweiss]

Mine also. This article implied they did see generalization, at least to other working memory tests - but I can't find the full text to check what they used.

[arkades]

The only paper I’ve seen so far with any generalization used timing feedback to incentivize closer attention. Generalized Outcomes were believed to be due to improved executive function/attentiveness, not memory, which is what the time pressure aimed for.

[DigitalJack]

I thought the unique aspect here was comparing techniques against each other. Not that the techniques hadn't been studied.

It didn't seem like a great distinction to me, but maybe it eliminates control issues when comparing two studies of different techniques vs comparing two different techniques in one study.

[Bartweiss]

Yeah, I think the head-to-head between techniques is what's novel about the paper. But what confused me is that they got a strong positive off dual n-back in the first place; other studies of that exact technique have regularly come up ambiguous-to-negative.

[JamesBarney]

The studies on Dual-N-Back increasing IQ have been all over the place. Some positive some negative. With the larger more careful trials coming up negative.

But IQ isn't working memory, attention, or executive function. And the findings on Dual-N-Back and working memory are less ambiguous and more positive.

Also there have been some more interesting studies done on Dual-N-Back in terms of changing Type 1 dopamine receptor density, as well as volume changes in certain parts of the pre-frontal cortex. So it really does seem like it's doing something even if that doesn't translate into an IQ change.

[Bartweiss]

> the findings on Dual-N-Back and working memory are less ambiguous and more positive

Huh, I guess this is what I was missing. I had the impression that even on working memory in particular, dual-n-back had mostly failed on a practical level - it just created training effects that broke the validity of certain tests.

If that's not the case, I'm suddenly much more interested.