[personaldev] Expanding working memory

[mattbell]

I've now played with the Dual N Back task for expanding working memory for ~20min a day for a couple of days.  I can now do 4-back fairly well and stumble through 5-back at a marginal level of competence.  My main complaint is that the program keeps upping the level the moment I do reasonably well even once, so I feel like it's continually pushing me into a zone of incompetence.  Based on the graphs of user performance from the initial study, the biggest boost is in the first couple of days, so if the study's correct I'll be gradually pushing my way up to 6 and 7 back if I keep this up for a couple of weeks.

I do feel by short-term memory pushing to accommodate the additional data in the harder tasks.  It doesn't feel like I'm chunking the databut perhaps I am doing so subconsciously.

Comments

[easwaran.livejournal.com]

As I was telling spoonless, on my second try I suddenly had a realization that made it much easier, which is that I can try to remember things in blocks of N, rather than remembering the most recent N at any time. It does mean that towards the end of a cycle, I'm keeping nearly 2N things in my memory, but somehow this is easier than "shifting" everything in my memory by one slot each time. I've found that 6 is a little too hard of a block to do this way though (though there was one time that I managed to stay on 6 twice in a row instead of immediately moving back down to 5). But the realization made a big improvement - instead of struggling on 3 and being hopelessly overwhelmed on 4, I now get perfect scores on 3 and generally do well enough on 4 to move up to 5.

[easwaran.livejournal.com]

Also, the fact that I could improve so much with what seems to be a very task-targeted trick makes me feel less confident that I'm doing anything useful and real by practicing. Perhaps I should look at the original study and figure out more about what this "gF" thing is that they claim improves, and how they measured it.

[nasu-dengaku.livejournal.com]

I started blocking things like that too, though the block system is too long to work beyond 5. When I got to six I started doing two blocks of three, which let me throw away the first block once I was halfway through the next set of 6.

I do feel like my brain keeps mentally repeating the blocks to keep them fresh, and I'm gaining skill in knowing when to repeat and how much I need to divert resources to remembering, not just evaluating, new blocks as they come in. It may be that I'm mentally creating algorithms for optimizing use of existing working memory instead of actually increasing working memory.

[spoonless.livejournal.com]

When I got to six I started doing two blocks of three, which let me throw away the first block once I was halfway through the next set of 6.

Wow, that is exactly what I started doing today when I finally got to dual 6-back. Hadn't read your post till now though.

I agree that it's annoying how it keeps pushing you into the incompetent zone.

It seems that now that I've done it a while, my audio and visual have reversed in strength back to what I would have expected. I was able to stay on audio 8-back for a good bit, getting mostly in the 60-70% range and one time 77% (thereby almost making it to 9-back). But for position, I could only get up to 6-back. Considering 6-back is also now the highest I've gotten in dual mode, it seems that visual, not audio, is my limiting factor.

[nasu-dengaku.livejournal.com]

Interesting. I assumed my audio memory was better than my spatial memory, but I was able to surprise myself by blowing through to position 9-back fairly fast even when I'm super tired. It was helpful to count off the positions as they came in (up to the N of the N-back I was doing) so that I knew quickly how far back in my memory to look.

You could potentially do just as well on the spatial N-back by converting the information to audio information (as in you give each space a different sound). Although this could be seen as defeating the purpose of improving spatial working memory, it's possible that your brain might be able to better reason about spatial data by incorporating audio information.

I need to go back to the main paper and make sure all our strategizing isn't just useless domain-specific adaptation.

[spoonless.livejournal.com]

It was helpful to count off the positions as they came in (up to the N of the N-back I was doing) so that I knew quickly how far back in my memory to look.

yup, I did exactly the same thing. That's the second time you've described a strategy and it's the same strategy I had started using independently. Must either mean we think alike, or these are really important tricks to use... probably more the latter :)

You could potentially do just as well on the spatial N-back by converting the information to audio information (as in you give each space a different sound).

I'm pretty sure this is not the case. If I am repeating one set of letters in my head, and I hear another set then my brain will automatically check for matches as it's going along. If I have to convert each spacial location into letters while I'm holding onto other letters that were converted before, that's a much harder task... I think. Then again, maybe I should try it and see, since I've been surprised before by what is easy or hard to do on these.

[spoonless.livejournal.com]

I assumed my audio memory was better than my spatial memory, but I was able to surprise myself by blowing through to position 9-back fairly fast even when I'm super tired.

By the way, what was your longest audio N-back? I am very impressed with your position 9-back, I had so much trouble with 6-back I can't even imagine getting anywhere close to 9. I did the position 6-back last after all the other stuff... so I had already been trying to focus for about an hour and was getting kind of tired. But I'd be surprised if that accounted for more than 1 level. I've always known that I'm pretty bad with spacial relationships though... I get lost very easily because i can't picture maps very well in my head, and geography was the only class in grade school that I got a D in.

[nasu-dengaku.livejournal.com]

On audio after about 15 minutes of trying I can get up to 8-back. I had no trouble with N<=6, but 7 and 8 are tough.