n-backing without software?

665 views
Skip to first unread message

revel

unread,
Mar 19, 2010, 6:43:32 PM3/19/10
to Dual N-Back, Brain Training & Intelligence
What do you people think about the prospective concept of a working
memory <i>meditation</i>? Is it possible to devise something similar
enough to DNB that we'd expect it to bring about the same sort of
cognitive gains that Jaeggi attributes to DNB? If not, what might be
essential about getting input from an external program, and what might
that say in turn about the mechanism of DNB training?

Meig...@aol.com

unread,
Mar 21, 2010, 12:52:48 PM3/21/10
to Dual N-Back, Brain Training & Intelligence
Well I'd say there's no way to "meditate" with working memory. It'd be
hard to explain I guess, the point of working memory is you use it
while _working_. To try and meditate using it somehow would mean
trying to actively hold an updating list of information, which is kind
of impossible on one's own.

Even if one thinks they can hold alot of information, you can't hold
more information than you can hold in working memory, and generally
speaking, the human brain can't focus itself inward enough to rewire
the brain the way intensive training can.

I'd say what's essential is the limited stimuli (small set of letters,
etc), update frequency, dual back, and the up/down when you get things
right/wrong enough. The way it would train your brain is pretty
obvious from the difficulty standpoint.

christopher lines

unread,
Mar 22, 2010, 1:22:39 PM3/22/10
to Dual N-Back, Brain Training & Intelligence
its an intersting idea...assuming a part of your mind could generate a
sequence n-back; for which you could train your mind to remember, I
dont really see how you can use another part of your mind to validate
your answers since that would infer somehow you have already committed
the sequence to memory.

I have noticed that before a training session on DNB I do sometimes
try to practise in my mind remember a random n-back sequence ...seems
to help if only to warm up those circuits.

Gwern Branwen

unread,
Mar 22, 2010, 1:26:09 PM3/22/10
to brain-t...@googlegroups.com

You can do N-back offline:
http://community.haskell.org/~gwern/static/N-back%20FAQ.html#offline-n-back

I don't know how well playing such a game accords with meditation.

--
gwern

Jonathan Toomim

unread,
Mar 22, 2010, 2:38:54 PM3/22/10
to brain-t...@googlegroups.com
While it's not dual n-back without props, this is still dual n-back
without software.

1. Grab a pack of cards. Ideally, the pack of cards would have 8
suits and 8 numbers per suit, but a standard 4/13 pack will do okay too.

2. Shuffle.

3. Pick N cards, one by one, look at them, then set them face down in
a pile.

4. Try to recall the suit and number of the card on the bottom of the
pile. Look at it. Score yourself.

5. Take another card from the deck, look at it, and put it on the top
of the pile.

6. Repeat steps 4-5 until there are no cards remaining in the deck or
you get bored.

7. Repeat steps 2-6 until you get bored.

Jonathan

revel

unread,
Mar 22, 2010, 4:24:20 PM3/22/10
to Dual N-Back, Brain Training & Intelligence
Jaeggi seems to recognize a broad set of cognitive regimens, including
DNB and meditation (and Nature Exposure : how Dao..), that may
transfer to general cognitive performance in a practically significant
way. Hence my context for thinking about hybridizing the two.

> Well I'd say there's no way to "meditate" with working memory. It'd be
> hard to explain I guess, the point of working memory is you use it
> while _working_.

Trust me, people don't always use WM at work ;)
I get your drift, though. Working memory is short term memory that can
be consciously manipulated in a nontrivial way. Absent the features of
WM, simple short term memory can still hold the pieces of a puzzle, so
to speak, but cannot shuffle them around in the different combinations
required to find the solution. This is my basic understanding, at
least.
Meditative practices will generally yoke attention in some way,
possibly even some aspects of short term memory, but I don't think
that they load much the ability to do working memory manipulations.
For example, holding a mantra in your head, even if it is by means of
repeating the mantra, is a comparatively static mind state. You will
be maintaining the mantra in memory, and possibly even updating the
contents of memory, something that is more clearly a part of
meditations that ask you be mindful of momentary sensations. But I'd
tend to argue that you will not be operating on the contents of
memory, you won't be juxtaposing one chunk against another.

>To try and meditate using it somehow would mean
> trying to actively hold an updating list of information, which is kind
> of impossible on one's own.

This is actually possible, provided that you are doing the updates
yourself. Computers do it all the time: http://en.wikipedia.org/wiki/Linear_feedback_shift_register.

In DNB, you are asked to maintain an updating fixed-length list of
information, and at each update, make a comparison (i.e, perform an
operation) between certain elements in the list. A fibonacci-type
recursion over the integers modulo q would have these same basic
features. Note also that such a pseudorandom permutation of state
space would be sufficiently random to keep the task challenging; the
only easy way to know better than chance what the state will be like
far in the future is to actually perform the recursion.

>...update frequency, dual back, and the up/down when you get things
> right/wrong enough.

It's on these notes that contemplative tasks tend to fail. But working
at it with some discipline might make sure to keep the task steadily
challenging; and certainly discipline is kind of a necessary condition
anyway for getting something out of any training program,
contemplative, cognitive, or otherwise.

Still, computer software probably streamlines the cognitive task a
great deal by imposing a regular pace, and quite possibly that is a
good thing. Too much complication in the task could be discouraging,
ultimately acting more as distraction than as anything else.

I'm still pretty agnostic about the degree to which such an
internalized WM task might compare with DNB re transfer to general
cognitive performance. But it's good to know of such an easilly-
seeded, ultraportable WM exercise. WM, though it probably is a
cognitive bottleneck at many critical junctures, is something that we
probably don't strain too much at most moments of our lives, and in
that light is probably a good target for explicit training.

Gwern Branwen

unread,
Mar 22, 2010, 4:56:59 PM3/22/10
to brain-t...@googlegroups.com

The linear feedback idea is interesting; I had thought of math
exercises, but all the ones that came to mind involved increasing
quantities or repetition or memorizing (the digits of pi, calculating
exponents of 2, doubling a seed repeatedly, etc.)

Wouldn't want to work on binary sequences but base 10 sequences, with
a easy update. Wonder what formula would work?

--
gwern

revel

unread,
Mar 22, 2010, 6:39:00 PM3/22/10
to Dual N-Back, Brain Training & Intelligence
Finding a maximal sequence will probably be trickier when one of
either q or list length k is nonprime. There are still plenty of
sufficiently long sequences for any q and sufficiently large q^k,
regardless. Maybe I'll try to look into the algebra if I find the
time, or more likely, someone more knowledgeable than I will give the
answer, should there be any significant practical interest in the
exercise.


On Mar 22, 3:56 pm, Gwern Branwen <gwe...@gmail.com> wrote:

Xenomemo

unread,
Mar 22, 2010, 6:58:18 PM3/22/10
to Dual N-Back, Brain Training & Intelligence

Well, long before I heard about dual n-back, I experimented a bit with
processing my sensory input in n-back style.
In regular intervalls I took mental snapshots of my surroundings and
with every update i tried to visualise what I saw n-intervalls back.
In the same way i tried to memorize short bits of sound and recall
them n-intervalls later.

For me it was hard to keep doing this because my mind started to
wander after a short time and without any external stimulus that
pushed me to go on the game came to an end.
I have to admid that I didn't try it too hard back then.
After all it was just a game to pass some time while walking around or
taking the bus.

revel

unread,
Mar 22, 2010, 6:59:51 PM3/22/10
to Dual N-Back, Brain Training & Intelligence
You could also just take a pragmatic approach and forgo the math in
favor of a quick computer script to pick coefficients and initial
data, compute cycle length, and if it is found to be insufficient for
your needs, rinse&repeat.

Ben Epstein

unread,
Mar 22, 2010, 8:45:46 PM3/22/10
to brain-t...@googlegroups.com
I think a playing-card game mode should be added, using all 52 cards as variables. Maybe add to this the variables of position and sound.

Jonathan Toomim

unread,
Mar 23, 2010, 2:56:43 AM3/23/10
to brain-t...@googlegroups.com
Oh no. Did I just make more work for myself?

Gwern Branwen

unread,
May 2, 2010, 10:08:20 PM5/2/10
to brain-training
So how simple could we make the procedure? I was tinkering around, and
I thought of this one:

remember a sequence of 4 numbers. Start with (1,1,1,1); add the last
two modulo 10 to get the new #1, shift #1,2,3 right. So if we were
using Haskell:

> let f (a,b,c,d) = (let x = ((c+d)`mod`10,a,b,c) in [x] ++ f x) in take 100 $ f (1,1,1,1)

[(2,1,1,1),(2,2,1,1),(2,2,2,1),(3,2,2,2),(4,3,2,2),(4,4,3,2),(5,4,4,3),(7,5,4,4),(8,7,5,4),(9,8,7,5),(2,9,8,7),(5,2,9,8),(7,5,2,9),(1,7,5,2),(7,1,7,5),(2,7,1,7),(8,2,7,1),(8,8,2,7),(9,8,8,2),(0,9,8,8),(6,0,9,8),(7,6,0,9),(9,7,6,0),(6,9,7,6),(3,6,9,7),(6,3,6,9),(5,6,3,6),(9,5,6,3),(9,9,5,6),(1,9,9,5),(4,1,9,9),(8,4,1,9),(0,8,4,1),(5,0,8,4),(2,5,0,8),(8,2,5,0),(5,8,2,5),(7,5,8,2),(0,7,5,8),(3,0,7,5),(2,3,0,7),(7,2,3,0),(3,7,2,3),(5,3,7,2),(9,5,3,7),(0,9,5,3),(8,0,9,5),(4,8,0,9),(9,4,8,0),(8,9,4,8),(2,8,9,4),(3,2,8,9),(7,3,2,8),(0,7,3,2),(5,0,7,3),(0,5,0,7),(7,0,5,0),(5,7,0,5),(5,5,7,0),(7,5,5,7),(2,7,5,5),(0,2,7,5),(2,0,2,7),(9,2,0,2),(2,9,2,0),(2,2,9,2),(1,2,2,9),(1,1,2,2),(4,1,1,2),(3,4,1,1),(2,3,4,1),(5,2,3,4),(7,5,2,3),(5,7,5,2),(7,5,7,5),(2,7,5,7),(2,2,7,5),(2,2,2,7),(9,2,2,2),(4,9,2,2),(4,4,9,2),(1,4,4,9),(3,1,4,4),(8,3,1,4),(5,8,3,1),(4,5,8,3),(1,4,5,8),(3,1,4,5),(9,3,1,4),(5,9,3,1),(4,5,9,3),(2,4,5,9),(4,2,4,5),(9,4,2,4),(6,9,4,2),(6,6,9,4),(3,6,6,9),(5,3,6,6),(2,5,3,6),(9,2,5,3)]

This probably has bad properties vis-a-vis randomness, but I'm not
sure any of them matter. If we have a hard time remembering 5 or 6
back, we're certainly not going to be worried about remembering a few
hundred or thousand cycles ago!

Trying it out for 10 minutes, it seems to work well enough. Even just
4 is initially quite a challenge for me.

--
gwern

--
You received this message because you are subscribed to the Google Groups "Dual N-Back, Brain Training & Intelligence" group.
To post to this group, send email to brain-t...@googlegroups.com.
To unsubscribe from this group, send email to brain-trainin...@googlegroups.com.
For more options, visit this group at http://groups.google.com/group/brain-training?hl=en.

Reply all
Reply to author
Forward
0 new messages