Re: Quintuple N-Back: Progress?

[Reece]

I haven't tried Quintuple N-Back yet, however I find Quad N-Back much
more difficult to chunk (gave up a long time ago on even trying to do
so) than DNB, so I can only imagine that will also hold true for
larger numbers of variables/modalities even if there is some way to
overcome this presumed "chunking limit".

One worry I have (and it's the same worry I have when playing quad n-
back) is that training quintuple or higher n-back might produce very
different results long term from what might have otherwise resulted
had we followed the research and stuck to DNB.

I've been playing Q2B and Q3B for the last 3 months and have very
little to show for it when I play DNB despite my scores on Q2B/Q3B
being substantially higher today than they were back then. It doesn't
seem like playing Q3B will make someone better at playing D5B or D6B
unless they are merely modifying chunking strategies they applied to
playing Q3B.

One person suggested awhile back that shortening the response time
window while keeping the time between trials at 3 seconds thereby
creating "dead time" between trials could be an effective way to cut
down on chunking, get the benefits (if any) of speed training, and not
risk sacrificing the benefits to be had by playing with the default 3
seconds per trial.



On 1 déc, 22:44, argumzio <argum...@gmail.com> wrote:
> Hello, everyone.
>
> I'm posting the first in what I hope is a long and fruitful discussion
> on the merits of, and pitfalls encountered while playing, 5NB. (I have
> already made a video demonstrating it, posted on youtube:http://www.youtube.com/watch?v=efSxgA7EpGk.) Any observations,
> thoughts, ideas, and so on, are welcome that aid in clarifying or
> interrogating the nature of working memory in the light of this new
> development.
>
> I have been reading some interesting information produced by Andreas
> Demetriou and others concerning the nature of WM and how it plays a
> vital role in cognitive development (specifically of a neo-Piagetian
> stripe). As indicated here (http://www.ncbi.nlm.nih.gov/pubmed/
> 15660854) it would appear that 5NB will be past the presumed limit of
> instantaneous processing of variables (i.e., four total). This poses
> an interesting theoretical question: seeing that working memory
> capacity develops by managing a maximum of complexity (experienced as
> a plateau in performance), and during which functional reorganization
> takes place such that a higher level of integration of information
> becomes possible, I would propose that all NB that exceed four mental
> operations (regardless of modality, but still bimodal at a minimum)
> will show no apparent differences in developmental progression. Of
> course, I certainly think it would be exciting to discover that this
> is not so, and that the total of mental operations could be increased
> without recourse to the all-too-familiar chunking effects; everything
> up to now, however, has indicated that the mind resists going any
> higher than this value, and does so by chunking information in
> various, dynamic ways. (One example of dynamic chunking that I have
> done (in a passive way) while playing QNB is changing which items I
> would temporarily group together based on moment-by-moment
> configurations, i.e., sometimes sound, sometimes color, etc.)
>
> My initial feeling is that with the binaural implementation of 5NB,
> and despite the added discriminatory complexity, the two audio streams
> are simply chunked in a spatial scheme rather than a temporal one, as
> would be the case of a monaural implementation. One way of overcoming
> this would probably be to use the same type(s) of information for both
> ears simultaneously.
>
> Anyway, I would also like this thread to enable us to relate our
> experiences and difficulties with this extraordinarily challenging NB
> mode.
>
> argumzio

[Michael Campbell]

argumzio wrote:


> TNB = 2^3 - 1 = 5
>

Small math nit:

2^3 - 1 = 7

[TheQ17]

I played 5nBack on nBack suite with movement as the 5th stimuli
instead of a second sound stimulus. I played it for a solid 2 months
and got relatively good at it and was able to get 3-5-Back in the 80%
range. My school work load picked up though and I eventually stopped.
I do not feel I received any benefit from the training.

On Dec 2, 11:00 am, argumzio <argum...@gmail.com> wrote:
> On Dec 2, 9:19 am, Michael Campbell <michael.campb...@unixgeek.com>

> Thanks. I can always count on those more attuned to arithmetic to get
> it right. ;-)
>
> I always did find it tediously boring (not especially difficult), so I
> never developed the tendency to double check my work. Another excuse:
> I was in a hurry.
>
> Ye math gods, spare this puny mortal your almighty fury!
>
> argumzio

[Gwern Branwen]

On Wed, Dec 2, 2009 at 3:54 PM, TheQ17 <jros...@gmail.com> wrote:
> I played 5nBack on nBack suite with movement as the 5th stimuli
> instead of a second sound stimulus. I played it for a solid 2 months
> and got relatively good at it and was able to get 3-5-Back in the 80%
> range. My school work load picked up though and I eventually stopped.
> I do not feel I received any benefit from the training.

Do you mean that N-back in general gave you no benefit, or PNB* in
particular didn't? (I don't have you in my testimonials list either
way.)

* I hereby rename 'Quintuple N-back' to 'Pentuple N-back', so as to
avoid confusion with Quadruple N-back

--
gwern

[Paul]

On Wed, Dec 2, 2009 at 5:53 PM, Gwern Branwen <gwe...@gmail.com> wrote:
> * I hereby rename 'Quintuple N-back' to 'Pentuple N-back', so as to
> avoid confusion with Quadruple N-back
>
> --
> gwern
>

Good suggestion - this will be in the official release.

Paul

[Reece]

One thing I was thinking about last night was how training with 2
sound stimuli might help in situations where there are more than one
conversation going on at the same time. They seem to be rather similar
tasks, especially if one were to speed up the settings to say, 1s/
trial (eg. 2 sound stimuli per second = 120 sounds per minute) so that
it is closer to the rate at which people converse.

On 3 déc, 15:30, argumzio <argum...@gmail.com> wrote:
> On Dec 2, 6:53 pm, Gwern Branwen <gwe...@gmail.com> wrote:

>
> > On Wed, Dec 2, 2009 at 3:54 PM, TheQ17 <jross3...@gmail.com> wrote:
> > > I do not feel I received any benefit from the training.
>
> > Do you mean that N-back in general gave you no benefit, or PNB* in
> > particular didn't? (I don't have you in my testimonials list either
> > way.)

> > --
> > gwern
>
> My question precisely, Gwern. The cognitive load, on the other hand,
> of a 3-visual, 2-auditory mode to my mind would be far greater than a
> 4-visual, 1-auditory one. Another issue that needs addressing is
> whether TheQ17 tried to reach beyond his cognitive limits (something I
> assume plays an important role in NB training.) Should I write
> something to add to your testimonials list, by the way? I believe I
> said something in an earlier thread, but I don't remember where
> exactly.
>
> I also thought about modifying the game's nomenclature, but opted for
> the rather clumsy and inconsistent numerical solution of 5NB. PNB
> works very well.
>
> argumzio

[Reece]

Saccading + DNB? in one :-)

On 3 déc, 21:23, argumzio <argum...@gmail.com> wrote:
> blank (if memory serves) actually had the same idea, right after I
> came up with the dual-channel mode.
>
> I support it completely as a very interesting addition. As for its
> benefits, we won't know until we have people plugging away at it for a
> while, speaking generally.
>
> Paul, are you reserving this dualDNB for a 4.9 or even 5.0?
>
> argumzio
>
> On Dec 3, 9:08 pm, Enoch Alien <alienenoch...@googlemail.com> wrote:
>
>
>
> > BW rocks. I like the dual-channel addition to 4.7.
>
> > I've got an idea I'ld like to know what you guys think about it.
> > Since I've got two laptops I'm considering extending the dual-channel
> > idea
> > to the visual realm where I run two instances of BW on two laptops
> > with a
> > division that isolates one eye per laptop screen. Would that have any
> > positive effects?
>
> > On Dec 4, 2:36 am, argumzio <argum...@gmail.com> wrote:
>
> > > Yeah, I thought the same when I came up with the idea of dual-channel
> > > training. Hopefully the skill can work. Luckily I'm ambidextrous, so
> > > it isn't unlikely that my hemispheres are less lateralized than
> > > others', meaning that my two hemispheres could probably better handle
> > > two auditory streams at the same time (where linguistic functions
> > > distribute between the two). We'll see...
>
> > > Incidentally, I find that whenever I listen to something where I can
> > > control which ear receives the input (e.g., audiobook with
> > > headphones), my left ear performs significantly better than the other.
> > > Hmm... Maybe my ears aren't so ambidextrous.
>
> > > argumzio

[TheQ17]

I personally don't feel a difference, although I thought I did at the
time. It may have been that I did go over my limit, but I did work my
way up to that level. I started with Dual Nback and made it to D5B
before switching to T3B. Within a couple of weeks the QNB was released
and I switched to that. I played that for most of summer and practiced
on Q4B the majority of the time. When Shamanu released his version
with PNB I played that usually at P3B or P4B. Out of curiosity I just
played for the first time in close to 3 months and scored a 60% on P2B
for 10 mins. I always practiced in 10 minute intervals I found it was
very easy to move up that way, especially if you test yourself at 1 or
2 minute intervals like the normal DNB trials.

At any rate, I don't feel studying is any easier although it wasn't
really difficult to begin with for me. Perhaps I'll give it another go
over break and report back. My goal originally was to get to P5B
before adding a second sound stimulus making a Sextuple Nback but I
don't know if Shamanu made an updated version to make that any easier.
I'm also kind of on the fence about the effect on the depth of
training. It may have been more beneficial to do higher N levels
instead of more stimuli.

On Dec 3, 4:30 pm, argumzio <argum...@gmail.com> wrote:
> On Dec 2, 6:53 pm, Gwern Branwen <gwe...@gmail.com> wrote:
>

> > On Wed, Dec 2, 2009 at 3:54 PM, TheQ17 <jross3...@gmail.com> wrote:
> > > I do not feel I received any benefit from the training.
>
> > Do you mean that N-back in general gave you no benefit, or PNB* in
> > particular didn't? (I don't have you in my testimonials list either
> > way.)

[milestones]

I think this Quint back is the perfect "ultimate" n back challenge but
probably the maximum one would want to go as far as increasing of
number of modalities go.

What is the highest a human could possibly hit on this? My guess would
be 6, maybe 7.

[TheQ17]

I think you'd be quite surprised how high people could go. This game
hasn't been out for very long and a very tiny percentage of the
population even knows about it, let alone tried it compared to
something like basketball or chess whom people have dedicated
ridiculous amounts of time practicing. I wouldn't be surprised at all
if people were able to keep track up to 10 stimuli for 5-6 steps back.
It's all about the time you put in to it. It only took me 4 months to
get to moderately successful levels of P4B and that is with only
playing 2 10 minute sessions a day for the most part. I wouldn't doubt
there are people far superior to my intellect and much more motivated
than I am that could completely surpass my levels in even quicker
time.

My theory is that your brain simply learns to consolidate the
information better. The more exposure your brain gets to the stimuli
the better the information can be recalled. It's a lot like opening up
a biology book and reading a few chapters for the first time. You
won't grasp or remember all of the information but if you continuously
overlap the information you can get it down to the point of almost
photographic recall of the figures and concepts shown throughout the
chapters. The only real benefit I for see DNB giving is something like
that of mindfulness meditation where you have to focus on a single
task or thought for a period of time. If you read the actual research
by jaeggi the whole claim of the program increasing IQ is rather
dubious considering the methods employed to come to that conclusion.

[Paul]

This is the sequence which is repeated at each trial:
- A random choice is made for each modality, 1/8 chance (even the
numbers are 1/8 because a subset of 8 numbers is chosen at the start
of the round).
- A random number is generated between 0 and 1
- If that number is below CHANCE_OF_GUARANTEED_MATCH, a random
modality is chosen and the stimulus is set to what it was N trials ago

This does not apply to Jaeggi mode which generates a sequence prior to
the start of the round which contains a constant number of matches of
each type (4 audio, 4 position, 2 simultaneous).

Paul

On Sun, Dec 6, 2009 at 7:45 PM, argumzio <argu...@gmail.com> wrote:
> Recently in another thread the issue of CHANCE_OF_GUARANTEED_MATCH was
> raised. Recalling that this wasn't the first time, I thought about
> this matter in private shortly thereafter and wasn't too sure
> regarding the particulars, but now that I have tested my hypothesis
> first-hand, I think there is something to be said for it, and I hope
> Paul and Jonathan are reading.
>
> For the lower NB modes, like DNB and TNB, higher occurrence of matches
> leads to a perceptibly easier session. This significantly reduces the
> executive load since one need not update n-items in memory as
> frequently. (This updating procedure is probably the most central
> aspect to DNB-style training.) However, with the likes of QNB and PNB
> (more so the latter), the nature of the game has changed. When I had
> passed P2B on the 3rd of December, that was when I had eliminated the
> guaranteed match factor (viz., set it to zero); my performance felt
> significantly easier than it did before, when it was at the default
> value of 0.25. Amazed and puzzled, I inquired into the cause of this.
> After having played a few trials where guaranteed match is set to
> 0.90, I have dropped my performance below threshold levels, and the
> difficulty of the task has increased dramatically, so much so that I
> am again at P2B.
>
> This leads me to think a few things:
>
> (1) 1/8 for n options leads to an uneven distribution of answer
> probability, such that (1/8)^5 would be the actual value for an answer
> like PCIAA in PNB, whereas it's ideal frequency would be something
> like 1/32 (based on 2^n = nNB, where n=5)
> (2) Weighted probabilities would better reflect a more evenly
> distributed occurrence of possible answers.
> (3) The cognitive load of recalling 5 items as opposed to 2 items
> instantaneously must be accounted for in the software design; since
> recalling 5 items has such a minimal chance of happening (where
> guaranteed match = 0), decreasing chance of guaranteed match would NOT
> make the task more difficult but the reverse: easier.
>
> Here are the possible outcomes of PNB (PCIAa form):
>
> {Nil response}, {P, C, I, A, a}, {PC, PI, PA, Pa, CI, CA, Ca, IA, Ia,
> Aa}, {PCI, PCA, PCa, CIA, CIa, CAa, PAa, PIA, PIa, IAa}, {PCIA, PCIa,
> PCAa, PIAa, CIAa}, {PCIAa}
>
> Generalizing therefrom, we get:
>
> Nil response: 1/32
> Single response: 5/32
> Double response: 10/32 = 5/16
> Triple response: 10/32 = 5/16
> Quadr. response: 5/32
> Pentu. response: 1/32
>
> The program may have to reflect the actual possible answers'
> distribution, because as it stands this is what it looks like with
> guaranteed match set to zero:
>
> P = 1/8 (8 spaces)
> C = 1/8 (8 colors)
> I = 1/8 (8 pentominoes)
> A = 1/8 (8 letters)
> a = 1/8 (8 letters; 14 numbers -> 1/14)
>
> (Thus, probabilities are analogous to flipping 5 octahedral dice at
> once. Deviating from the item types, as in the case of numbers, would
> serve to amplify the problem, if their value is greater than 8 total.)
>
> CHANCE_OF_GUARANTEED_MATCH = 0.375 would lead to 1/8 -> 1/2, which
> would reflect the actual possible answers' distribution (including nil
> response). Even if this were the setting, it may still be too easy,
> because the overall distribution would not quite be weighted to the
> more difficult triple, quadruple, and quintuple response items. So
> something like .4 would be the minimum to achieve a balance between
> cognitive load and non-repetition.
>
> Be all this what it may, this is simply a general outline of the
> issue, and may not reflect an actual solution to what I see as a
> problem. Note that I am following this description in the config.ini
> in addition to my observations of the nature of the task:
>
> # The chance that a match will be generated by force, in addition to
> the
> # inherent 1/8 chance. Setting this to 1 will guarantee at least one
> match
> # each trial (and will cause some repetitive sequences to be
> generated).
> # The value must be a decimal from 0 to 1
> # Increasing this value will make the n-back task significantly
> easier.
>
> Any clarifications (e.g., the nature of "force" in the above, etc.),
> thoughts (e.g., refinement of the design, etc.), etc. would be very
> helpful on this matter. Since I have clearly thought this over, I do
> not think a "variable match" (as I had said earlier) would be adequate
> to ameliorating the situation; the matter is a little more subtle than
> pure "randomness" since we have to figure in cognitive demands (not
> just pure probabilities).
>
> argumzio
>
> --
>
> 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.
>
>
>