After 6 months of training, no changes in intelligence.

[jttoto]

Just wanted to give an update.

6 months ago I posted my IQ on this site after taking the Mensa Norway
test on this faq: http://community.haskell.org/~gwern/static/N-back%20FAQ#measuring-with-iq-tests

I scored a 135. After 6 months of dual n-back, triple n-back, and
quad n-back training, I took the same exact test. I scored exactly
the same, 135.

Granted, I took 7 less minutes to complete the test, but this was due
to familiarity of some of the questions.

That being said, I have been seeing significant increases in my digit
span and other WM gains, so while my aptitude on questions like the
Raven's may not have increased, my memory has.

[Gaël DEEST]

I once read that lobotomy has no effect on IQ, although it is known to
alter many cognitive functions. It seems to me like a striking
evidence that IQ, as relevant and predictive as it may be in many
situations, doesn't capture as much as it should of our mental
abilities. The prefrontal cortex (who's targetted by the lobotomy)
being at the heart of many complex operations, and n-back being
claimed to produce changes in the density of neuroceptors in the
cortical area (whether it is in the prefrontal cortex, I actually
don't know), is it all that suprising if you don't notice any change,
even after months of training ?

A few years ago, someone told me that the Mensa tests were actually
not very good, as they didn't test all the scales a professional test
supposedly does. I don't know if it's still true or even applies to
the test you're mentioning.

I know we're all here because some IQ-like test actually /did/ notice
something, so I'm not claiming that the benefits of N-back can't be
detected by an intelligence test - I actually don't know. But I'd like
to point out that it might be possible that other brain training
exercises / tools that do not seem to improve Gf / IQ / whatever may
actually work on another, still undetected scale that actually
improves one's ability to focus, work, solve problems, learn and
retain.

2009/9/28 jttoto <jtde...@uncc.edu>:

[zzzz]

Could you provide some details on your training routine?

[jttoto]

In the beginning of my training, I was extremely adamant about
increasing my n-back score. So for the first 3 days, I trained about
3-4 hours. After that, I became less enthusiastic about it so I
trained about 15-20 sessions 5 times a week. Some weeks I slipped up
and only did 3 times a week, but I can guarantee that I trained 5
times a week over 93 percent of the time.

My WM was never anything special, but now it is well above average. I
have seen significant gains in both my spatial and verbal WM.

[zzzz]

3 sec/trial?

[Gwern Branwen]

On Mon, Sep 28, 2009 at 11:46 AM, jttoto <jtde...@uncc.edu> wrote:
>
> In the beginning of my training, I was extremely adamant about
> increasing my n-back score.  So for the first 3 days, I trained about
> 3-4 hours.  After that, I became less enthusiastic about it so I
> trained about 15-20 sessions 5 times a week.  Some weeks I slipped up
> and only did 3 times a week, but I can guarantee that I trained 5
> times a week over 93 percent of the time.
>
> My WM was never anything special, but now it is well above average.  I
> have seen significant gains in both my spatial and verbal WM.

If we're all being skeptical here about this null result, perhaps you could just include your stats and config files as attachments and let the nay-sayers see for themselves.

That said, I'm curious: given no improvement in IQ, do you feel N-backing has been worthwhile?

----

On an entirely different topic, while I was adding jttoto's report to the FAQ, I decided to finally write up a terminology section (since we all like to use neologisms and acronyms). Here's what I have so far:

> N-back training is sometimes referred to simply as ‘N-backing’, and participants in such training are called ‘N-backers’. Almost everyone uses the Free, featureful & portable program Brain Workshop, abbreviated “BW” (but see the software section for alternatives).

> There are many variants of N-back training. A 3-letter acronym ending in ‘D’ specifies one of the possibilities. For example, ‘D2B’ and ‘D6B’ both refer to a dual N-back task, but in the former the depth of recall is 2 turns, while in the latter one must remember back 6 rounds; the ‘D’, for ‘Dual’, indicates that each round presents 2 stimuli (usually the position of the square, and a spoken letter).

> But one can add further stimuli: spoken letter, position of square, and color of square. That would be ‘Triple N-back’, and so one might speak of how one is doing on ‘T4B’.

> One can go further. Spoken letter, position, color, and geometric shape. This would be ‘Quad N-back’, so one might discuss one’s performance on ‘Q3B’. (It’s unclear how to compare the various modes, but it seems to be much harder to go from D2B to T3B than to go from D2B to D3B.)

> Currently, Quad is the highest Brain Workshop supports; Paul (BW’s developer) writes that a ‘Pent’ mode

>> “would require not-insignificant changes to the code (the addition of another modality) but this seems to be the next logical progression. I’ll look into this after the official release of 4.6.”

> Other abbreviations are in common use: ‘WM’ for ‘working memory’, ‘Gf’ for ‘fluid intelligence’, and ‘g’ for the general intelligence factor measured by IQ tests.

Did I miss any jargon?

--
gwern

[jttoto]

Done, I've been recently playing quad n-back so I haven't played dual
in a while. I can play dual today or tomorrow to report any
improvements.

On Sep 28, 11:58 am, Gwern Branwen <gwe...@gmail.com> wrote:

>  signature.asc
> < 1KViewDownload

[jttoto]

And to answer you question, yes I feel that it has been extremely
worth while.

My friends have always called me inattentive and absent-minded, but
since playing n-back no one has called me that for a while. I now
never forget where I park my car, when I used to do that nearly every
other day. I feel more attentive. Even if my ability to solve
problems hasn't improved, the gains in my memory are real and
measurable.

[Pontus Granström]

Well let us not throw out the baby with the babywater. I've taken many IQ-tests some that varied alot,
actually all of them. Does this mean that we should reject the G-theory based on one person! Maybe, maybe not. Read my example about being a good runner, this is the way I see dual-n-back, as a factor in the g-factor. I think we strengthen parts of our computational hardware and get accustomed to working at a higher intensity level.

I've experienced alot of difference in my daily life, greater mental stamina and greater confidence when experience demanding tasks (of course it could increase IQ) especially in math class, and I also just completed my  master thesis project that took alot of work, reading and thinking. I think dnb helped alot!
However after one day of computer science work one usually do not have the stamina to play dnb.

Let us be open minded and get many views and insights before jumping to any conclusion.

[Mike^2]

It's funny you mention that you've seen WM and Digit-Span increases
when these things have been shown -in studies- to correlate highly
with performance on IQ tests.

Perhaps the correlation is there with professional IQ tests, the tests
-or so i'd presume- such researchers would employ in their studies.

In any case, if IQ is used as a predictor of success later in life or
performance in school and academia, i think there has been a study
(which can be found here) that has found WM to be an even better
indicator of just this, undermining to a certain extent, the weight of
these IQ tests in light of other cognitive factors which are more
important.

On Sep 28, 10:04 am, jttoto <jtdem...@uncc.edu> wrote:
> Just wanted to give an update.
>
> 6 months ago I posted my IQ on this site after taking the Mensa Norway

> test on this faq:  http://community.haskell.org/~gwern/static/N-back%20FAQ#measuring-wit...

[childofbaud]

Interesting result, thanks for sharing.

What's the ceiling on that particular test?

On Sep 28, 10:04 am, jttoto <jtdem...@uncc.edu> wrote:

> Just wanted to give an update.
>
> 6 months ago I posted my IQ on this site after taking the Mensa Norway

> test on this faq:  http://community.haskell.org/~gwern/static/N-back%20FAQ#measuring-wit...

[Toto]

I think the ceiling is 145. Tests usually are unreliable for higher
levels, but I have similar experience with DNB. Training with TNB with
shapes, however, may have made a difference - I scored 99,9 % on
CFNSE. I learned later that it was inaccurate above 140, but it seems
that I'm able to handle more difficult problems, though I expected
that if there was any improvement, it would be only in speed.

[Pontus Granström]

You trained 6 months but which level did you reach?! Someone said that the realtion between N-level and DNB is 3=110, 4=120,5=130 etc +10 for each level.

[revel]

You mean n-level and IQ? Anyway, that relation is absurd, if you think
about it.

Jaeggi's paper, in the main, presents a time-dependant training gain
to a raw score on rapm-type problems.
In particular, the gains are still present even after controlling for
final n-level, and are not measured strictly in terms of IQ.

On Sep 29, 10:27 am, Pontus Granström <lepon...@gmail.com> wrote:
> You trained 6 months but which level did you reach?! Someone said that the
> realtion between N-level and DNB is 3=110, 4=120,5=130 etc +10 for each
> level.
>

[jttoto]

Hello, I got to the point where I can comfortable reach dual 6 back in
less than 8 tries.

If allowed 20 sessions, I can easily average 5.

I started switching to triple n-back when I found out I started
reaching 7 as my max. I might even average 6 now, if given 20
sessions, although I haven't tested this yet. (something I plan on
doing tonight)

On Sep 29, 11:27 am, Pontus Granström <lepon...@gmail.com> wrote:
> You trained 6 months but which level did you reach?! Someone said that the
> realtion between N-level and DNB is 3=110, 4=120,5=130 etc +10 for each
> level.
>

[jttoto]

And I do remember someone saying that, but I believe he was describing
averages rather than max, and his theory hasn't been formally tested.

On Sep 29, 11:27 am, Pontus Granström <lepon...@gmail.com> wrote:

> You trained 6 months but which level did you reach?! Someone said that the
> realtion between N-level and DNB is 3=110, 4=120,5=130 etc +10 for each
> level.
>

[Reece]

I have my doubts that IQ would increase anywhere near such a large
rate at higher n-backs -- some of the top n-backers are doing dual 10+
n-back which would put their IQ over 200. Those performing best at n-
back certainly do seem to be far above average in intelligence,
however I strongly believe most people interested in intelligence
amplification would score well above average.

IQ doesn't measure everything -- success at school/college is stronger
correlated to motivation levels than to IQ, as are things such as the
ability to focus and concentrate (eg. difference in perfomance between
ADHD children on stimulants versus not on stimulants).

For me, the biggest problem I have in life is a terrible short term
memory. If dual n-back can (and it does appear to) help with that,
then I'm more than happy to continue dedicating several hours per week
to it.

[Pontus Granström]

Okej interesting! Well I scored about what you do and I increased my score on iqtest.dk to 133 and I scored 135 on the Mensa.no. Have in mind though that Mensa.no is not a very professional test rather one that has been made by a student. I think that we need to find other evidence for or against dnb and Gf gains. Neuroimaging etc. Never the less there are many studies point to that training on what can seem a simple task can transfer into general performance. Just like upgrading your CPU architecture would make it possible to speed up calculations etc and make it seem that the computer is alot more advanced.

[Raman]

I scored 130 on Mensa.no and 112 on iqtest.dk . so how do we reconcile
this huge difference in my IQ scores? And both the tests were after 19
sessions of DnB where I reached 7 and in the last session even got to
9 (though I was promptly busted back to 8!!). I don't think there is
any correlation between n level and IQ scores - Jaeggi herself said
that n-level was not significant - it was the training that was
significant.

As far as upgrading the brain architecture goes, I guess it would take
years of practice to make a permanent change in brain. I think there
were some studies on changes in Buddhist monks brains due to
meditation and the most changes happened to monks who had been
practicing for years and years.

How many ppl are doing QnB regularly? I am trying to put 150 trials
per session but its extremely hard work (I get around 65% on Q2B and
40% on Q3B and 20% on Q4B with the 5 minutes session - I have put the
time as 2.00 sec/trial). And with the 5 minutes session I feel intense
psychological pressure to give up mid-way - and my brain is appears to
be extremely taxed. Does this mean am making progress?

[Gwern Branwen]

On Tue, Sep 29, 2009 at 2:47 PM, Raman <raman...@gmail.com> wrote:
...

> As far as upgrading the brain architecture goes, I guess it would take
> years of practice to make a permanent change in brain. I think there
> were some studies on changes in Buddhist monks brains due to
> meditation and the most changes happened to monks who had been
> practicing for years and years.

...

That seems a little pessimistic. Neither of the 2 DNB studies showing
physical brain changes (the fMRI and neurotransmitter studies) made
students play DNB for years. And IIRC, I believe I've seen meditation
studies which trained over only a few weeks or months and saw real
changes. Those experienced monks certainly may have more changes, but
that's not the same thing as it taking years/decades to make any
change. (There's likely diminishing gains going on there; one year of
meditation will make a bigger difference to someone just starting than
to someone whose practice is a score of years old.)

--
gwern

[Reece]

Hi Raman,

I play Q2B/Q3B daily with 100-200 trials and my scores are similar to
yours. I've been getting a high pitched "buzzing" sound in my head a
lot lately -- seems to be from Quad n-back, as I can never recall such
happening so frequently in the past. The only other time I ever get
this with any frequency is after listening to theta wave music for
extended periods of time.

[jttoto]

Strange, I scored higher on the mensa.dk (I took it today), and
personally thought it was easier (considering the questions were very
similar to the Mensa.no)
It is still significant that you scored above-average however.

[akprasad]

Sigh. This completely misses the point of _controlling for n-level_,
as jtotto pointed out. If your only goal is to increase n-level, you
can do that quite easily by teaching yourself mnemonic tricks.

On Sep 29, 8:27 am, Pontus Granström <lepon...@gmail.com> wrote:
> You trained 6 months but which level did you reach?! Someone said that the
> realtion between N-level and DNB is 3=110, 4=120,5=130 etc +10 for each
> level.
>

[Raman]

Really - try using mnemonics with interval set to 1.5 or 2s with
QnB!!! If you can use mnemonics at this speed then you are already a
genius in my books :)

[jttoto]

True, but I took the Mensa.no not too long after and scored a 135. I
can take three different IQ tests and have dramatically different
scores, which is why I wanted to take the exact test to see if there
were any improvements. (The European IQ Test on Gwern's faq says I'm
a super-genius, and I'm just not that smart) While my n-back score
improved, my fluid intelligence on the exact test didn't. Does that
say something?

The professionally measured IQ test was filled with knowledge based
questions, which was not an accurate measure of my IQ. This is due to
the fact that I didn't pay attention in grades 1-12.(I cleaned up my
act in college however). Doing enough just to pass really puts a dent
on the amount of general knowledge one acquires over time. On that
same test, my best scores were in the fluid sections, not the
knowledge (or crystallized) sections.





On Sep 30, 5:29 am, blank <peisistrat...@gmail.com> wrote:
> jttoto, you took a professionally administered iq test around 17
> months ago and scored a 126

[Pontus Granström]

I am the same I scored 3 stanine points above my fluid intelligence score. That test consisted of three different measures of Gf, visual reasoning, bongard and number series (originated from 1947 by Guildford and Lacey) the swedish psychology professor made his own version. I scored different on all of those three, which 6 stanine points difference in maximum. This however was for a scale of above average, meaning that a low score put you above average anyways, but still! One of the reason why I strongly believe in DNB is because I took this test! It was built in a different way where you had to remeber the answer in your head and they were shown separate from the questions. Something that takes alot of working memory. The test lasts for roughly about the time of 20 DNB sessions sometimes a few minutes less sometimes a few minutes more. It was also highly speeded which meant you had to solve a question in 20-40 seconds then solve the next one you could not administrate your own tempo, go back or anything! Just like in DNB.

The number series tests was very hard and required long calculations something that also takes alot of working memory. Overall DNB could really boost your score on these kind of tests!

It would be interesting though if we could measure the exakt load on WM when taking a Gf tests. Even though they claim to measure G it is easy to see that they do infact rely on a strong WM. Imagine to benchmark IQ-tests based on WM just like a computer monitoring program does with a running process.

[Pontus Granström]

I scored 3 points above my Gc score on the Gf test I mean.

[Pontus Granström]

I appologise for the bad language! XD

[jttoto]

And yes, it seems my memory was a little hazy on the subject, and I
sincerely forgot that I started n-backing a little before I took the
Mensa.no. I also incorrectly entered my average on that post.
Numerical typos are harder to catch.

But, even if we assume that n-backing on the first 3 or 4 days did
improve my IQ score, this still shows that for nearly 6 months of
training after had no effect on my score after. I don't think I need
to expound on what this means, but this is not a good thing for those
that have trained months, like me, or years. Has anyone seen
continuous improvements on their IQ and the amount of time they've
been training? Does anyone have hard evidence that improvements in IQ
are directly proportional to the amount of time spent?

[Pontus Granström]

Isn't the study the evidence?! Beyond this only what has been reported here.

On Wed, Sep 30, 2009 at 2:19 PM, blank <peisis...@gmail.com> wrote:

> But, even if we assume that n-backing on the first 3 or 4 days did
> improve my IQ score, this still shows that for nearly 6 months of
> training after had no effect on my score after.

ok then, i need to repeat myself. please clarify what you say here:

"In the beginning of my training, I was extremely adamant about
increasing my n-back score.  So for the first 3 days, I trained about
3-4 hours.  After that, I became less enthusiastic about it so I
trained about 15-20 sessions 5 times a week."

nothing personal, just like the truth.

[jttoto]

Yes Pontus, I meant to say after 20 days. I apologize for not making
that clear.

Blank, I understand it isn't personal, but could you please describe
what you want me to elaborate on? I think my explanation is pretty
lucid, but I will clarify. On the first 3 days I trained far more
than 20 sessions. It could be more than 3 days, could be less. I
remember it being for what felt like hours. I posted my stats which
are far more objective (and reliable) than my memory.

You are saying that my n-back score did improve my intellect due to
the intense training I had before. If you are implying that training
15-20 sessions 5 times a week goes against the Jaeggi experiment, then
you are correct. You are implying that training this way does not
improve WM and therefore does not provide any further gains in Gf,
then I can't agree with you. My WM and n-back score still continued
to improve despite training , as shown in the stats I posted.

On Sep 30, 8:23 am, Pontus Granström <lepon...@gmail.com> wrote:
> Isn't the study the evidence?! Beyond this only what has been reported here.
>

[jttoto]

I also want to make a formal apology. At the time I posted this I
remembered taking the mensa IQ test before n-backing, but now, looking
at my old post, I remember otherwise. Understand that the two events
were in close proximity to each other and it was 6 months ago.
Because of this, the conclusions that can be drawn from my original
post are contaminated. I think this is a good example of how
unreliable anecdotal information can be.

[Gwern Branwen]

On Wed, Sep 30, 2009 at 7:55 AM, Pontus Granström <lepo...@gmail.com> wrote:
> I am the same I scored 3 stanine points above my fluid intelligence score.
> That test consisted of three different measures of Gf, visual reasoning,
> bongard and number series (originated from 1947 by Guildford and Lacey) the
> swedish psychology professor made his own version. I scored different on all
> of those three, which 6 stanine points difference in maximum. This however
> was for a scale of above average, meaning that a low score put you above
> average anyways, but still! One of the reason why I strongly believe in DNB
> is because I took this test! It was built in a different way where you had
> to remeber the answer in your head and they were shown separate from the
> questions. Something that takes alot of working memory. The test lasts for
> roughly about the time of 20 DNB sessions sometimes a few minutes less
> sometimes a few minutes more. It was also highly speeded which meant you had
> to solve a question in 20-40 seconds then solve the next one you could not
> administrate your own tempo, go back or anything! Just like in DNB.
>
> The number series tests was very hard and required long calculations
> something that also takes alot of working memory. Overall DNB could really
> boost your score on these kind of tests!
>
> It would be interesting though if we could measure the exakt load on WM when
> taking a Gf tests. Even though they claim to measure G it is easy to see
> that they do infact rely on a strong WM. Imagine to benchmark IQ-tests based
> on WM just like a computer monitoring program does with a running process.

Interesting that you focus on the speeding aspect, and that jttoto
reports a 7 minute decrease in his test-taking time.

Didn't Moody's critique of Jaeggi focus on how the administered
Raven's was artificially sped up? (This suggests that 'all' DNB does
is speed up your thinking rather than put it on a whole new level. As
a sci-fi fan, I can't help but think of the distinction between 'weak'
superintelligence and 'strong' superintelligence!)

(On a side note, I've uploaded a fresh copy of the FAQ, with the last
couple weeks/months of changes. It's been a while since I refreshed
the static HTML.)

--
gwern

[Pontus Granström]

On the other hand a common opinion today is that speed reflects intelligence and that's why there are time limits at the tests.

[Ron Williams]

I suppose someone has said this already (I haven't been able to find where they did, if so), but it seems to me that the modus operandi of IQ tests that measure crystallized intelligence relies on the observation that smarter people know more stuff (in a standardised, normed way, of course :)).

So what this tests for is what you've used your current intelligence _to learn_ over the years. I would have thought that a test that incorporates a significant component of this kind would be totally inappropriate for measuring _changes_ in intelligence, since this correlation was introduced into testing in an era in which IQ was thought to be fixed, immutable.

Such a test cannot measure what you _will_ learn as a result of an increase intelligence, over time going forward. They should be avoided in this context (plasticity of 'g'), IMHO. At the very least, the 'crystallized' portion should be identified and discounted in re-testing post-training.

[Gwern Branwen]

All very true; what we want is a culture-neutral IQ test. Raven-style
tests seems to be the closest we can get to pure Gf tests; I include
links to just Raven-style IQ tests in the FAQ (or at least IIRC I
included disclaimers about the ones that struck me as having
crystallized-testing components).

I believe jttoto's and most of the other reports use only these
Raven-style tests.

--
gwern