7
$\begingroup$

The Dual N-Back task is the only task I'm aware of that has empirical support showing that it improves working memory. It appears to improve working memory through multitasking, although this multitasking must fit certain criteria, namely that you can't learn the task so well that you can eventually "automate" it without going into working memory.

Thus, I'm wondering whether other forms of multi-tasking such as playing Starcraft II (see my earlier thoughts) and driving with a cellphone can also improve working memory?

A possible follow-up question: if not, what differentiates the type of multitasking found in the Dual N-Back task from other types of multi-tasking?

$\endgroup$
  • 2
    $\begingroup$ I'd like a ref on how multitasking improves WM via N-back. Also, the conclusion to your question seems kind of obvious: there are many differences between N-back and Starcraft (what's the difference between a sardine and a Ford Pinto?) Further, multitasking might be pursued via a number of literatures: perceptual load; switching costs; mixing costs, all of which bear on this question, but all of which presume extremely restricted task sets. Comparing their WM effects to those produced by a video game seems apples to oranges. $\endgroup$ – shanusmagnus Jan 20 '12 at 4:05
  • $\begingroup$ Added another link, and added the part about using a cellphone while driving a car. Regarding the Starcraft II relationship - someone from LessWrong says something very similar (lesswrong.com/lw/6sj/experience_with_dual_nback/4jme) and I'd like to see that investigated more rigorously. $\endgroup$ – InquilineKea Jan 22 '12 at 0:49
  • 2
    $\begingroup$ @InquilineKea I've given the question a bit of an edit; I've tried to remove assumptions and couch claims more tentatively; I've also tried to create a structure whereby there is an initial discussion of the motivation for the question culminating in a clear question. Feel free to edit. $\endgroup$ – Jeromy Anglim Jan 23 '12 at 22:59
9
$\begingroup$

In general, I'd hypothesise that "memory-training" programs will not lead to domain-general increases in fluid intelligence nor working memory.

As general background, you might want to check out the literature on expert memory.

  • Practice is very effective at improving performance on the practised task. Transfer is real and does exist, but it is often small in effect.
    • Let's take the study by Ericsson, Chase, and Faloon (1980). After 230 hours of practice a participant, SF, was trained to increase their digit span (i.e., the sequence of random numbers they could recall) from 7 to 79 numbers. One might think that SF had increased his working memory. However, talk aloud protocols suggest that SF was using sophisticated domain specific mnemonic strategies (e.g., linking running times to random numbers). The skill did not generalise to other memory stimuli.
    • The expertise literature is filled with examples of experts seemingly defying the limits of human ability. Yet, such achievements tend to be domain specific and achieved through substantial practice. For an excellent review of the expertise literature, see Ericsson, Krampe, and Tesch-Romer (1993).
  • It is very difficult to modify very large domain general abilities.

Thus, my general advice is for people wanting to improve their working memory or fluid intelligence is to instead focus on what domain specific ability they want to improve and focus on practicing that.

References

  • Ericsson, K. A., Chase, W. G., and Faloon, S. (1980). Acquisition of a memory skill. Science, 208(4448):1181-1182. FREE PDF
  • Ericsson, K. A., Krampe, T., and Tesch-Romer, C. (1993). The role of deliberate practice in the acquisition of expert performance. Psychological Review, 100:363-406.
$\endgroup$
2
$\begingroup$

In my mind the dual-n-back (aka Jaeggi training) is still controversial. A meta-analysis by Jaeggi and co-authors (2015) found in favor of their original paper, but others have gone on record to disagree with their conclusions, Melby-Lervåg & Hulme (2016) in particular who had conducted a meta-analysis concluding the opposite in 2013; Jaeggi and co-authors disagree with that criticism, of course. Dougherty et al. (2016) were also unconvinced of the meta-analysis of the Jaeggi group.

There's a 2016 editorial in Nature on the state of affairs on this line of research, with the subheading:

Conflicting results are expected in a young field, but what do you do when even the meta-analyses do not agree?

An empirical study by Lawlor-Savage and Goghari (2016) published after those meta-analyses and which (according to its introduction) took methodological concerns outlined in the meta-analysis into consideration for its own experimental design, also failed to find improvements from dual-n-back. I'm not saying this is the last word on the issue, just that replication has proven difficult.

So I think it's somewhat premature to ask what else works like this, when we're not sure this works.

$\endgroup$
0
$\begingroup$

I think the "automate" problem is caused by several reasons in my experience.

1: brainworkshop's design on fixed time intervals between trials is bad. Some of the time, we increase our n-back level without some seconds more to recall the letter or position, which ruins the rest of the trials. And so, we chose to remember letters and positions and finally "automate" this bad practice.

2: It is bad to have too many letters in brainworkshop. If you can, imagine how we tried to remember a long word while we were still children. So we suck the long chain of letters into our minds instead of recalling what we saw.

3: Positions are not the best way to practice n-back, just like in the last reason. The human brain can easily recognize this kind of 3x3 spatial pattern. While we are training, we can't avoid learning the 3x3 patterns eventually. How hard to stick a 3 position set and a 3 letter set in mind, right? But while we are sticking those, we have less parts of our minds available to recall memories.

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.