Psychology Today article on working memory training

[XFMQ902SF]

This is a somewhat old but good article by William Klemm, a
psychologist. Klemm writes about studies done in Japan showing
transfer of working memory training to fluid intelligence in children.
The working memory tasks used are described as keeping a list of
numbers in mind and remember order. Ex. 5389, and the children would
have to say which number came fourth? or "what place is number 5?"
Gwern wrote(i think) about n-back simply being a visual spatial task
which familiarizes the trainee with visual tasks and the RAMP is a
visual IQ test, therefore removing its g-load. Well, this study from
Japan is a working memory training method which is not visual and
shows transfer to IQ afterwards. Klemm did not cite the study showing
transfer at the end of the article.

Specifically from the article
>training method could increase the working memory capacity of children. While they were at it, they tested for any effect on IQ. Children ages 6-8 were trained 10 minutes a day each day for two months. The training task to expand working memory capacity consisted of presenting a digit or a word item for a second, with one-second intervals between items. For example, a sequence might be 5, 8, 4, 7, with one-second intervals between each digit. Test for recall could take the form of "Where in the sequence was the 4?" or "What was the 3rd item?" Thus students had to practice holding the item sequence in working memory. With practice, the trainers increased the number of items from 3 to 8.

After training, researchers tested the children on another working
memory task. Scores on this test indicated in all children that
working memory correlated with IQ test scores. When first graders were
tested for intelligence, the data showed that intelligence scores
increased during the year by 6% in controls, but increased by 9% in
the group that had been given the memory training. The memory training
effect was even more evident in the second graders, with a 12% gain in
intelligence score in the memory trained group, compared with a 6%
gain in controls. As might be expected, the lower IQ children showed
the greatest gain from memory training.<
http://www.psychologytoday.com/blog/memory-medic/201203/training-working-memory-why-and-how

[Pontus Granström]

You shouldn't really pay too much attention to Gwerns statement, his dogma has been proven wrong too many time. The link between n-back and IQ might not be working memory per se, but rather the ability to update the working memory. This is a executive function that shows a high g-load (inter correlation) even higher than many sub tests of 0.6 or something like that. Once again, the rather logistical thinking isn't really  functional when it comes to neuro science. It might be other reasons as well, perhaps reducing anxiety, self regulate, improve the mental stamina and also strengthening overlapping circuits.

--
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.

[Gwern Branwen]

So they trained digit span over 2 months in young kids? We've already
discussed how kid and even teenager IQs are variable, and from the
sound of it, the difference was small.

More importantly, nothing was mentioned of how randomization or how
controls were done. We know from Melby-Lervåg & Hulme 2012 that
omitting active controls alone changes the effect size of WM exercise
on non-verbal ability (IQ) from _d_=0.00 to _d_=0.38 - which sounds
roughly right for 6% vs 9%.

> As might be expected, the lower IQ children showed the greatest gain from memory training.

Yes, if only from regression to the mean...

--
gwern
http://www.gwern.net

[☉]

On May 27, 10:14 am, Gwern Branwen <gwe...@gmail.com> wrote:
> We've already
> discussed how kid and even teenager IQs are variable, and from the
> sound of it, the difference was small.

When we did, it was further pointed out that the study design _did
not_ account for maturation or effects of history, meaning such
studies fail to be of any viable import (valid).

It's kind of amusing to me that people think children can be used as
legitimate examples of N-back training and its alleged transfer on
I.Q. tests - especially since it is well known that their I.Q.'s tend
to be variable, mostly in the upward direction, simply as they get
older.

http://groups.google.com/group/brain-training/msg/1d1a874ef7d84d22

Bad study design is a big no-no when you're trying to say an
intervention has a serious effect like "far-transfer" (which probably
isn't being used in any rigorous manner).

argumzio

[XFMQ902SF]

If we stretch our ability to keep bits of information in our working
memory, we may very well increase fluid intelligence. According to
Oregon psychologist Edward Awh quantity of items stored in short term/
working memory is correlated with IQ, not so much clarity. SO, if the
children in the study did in fact expand their working memory from 3
items to 8 items simultaneous then it might make sense that IQ could
also improve. This is assuming that they did in fact improve working
memory after training and the study is not flawed as others suspect.

http://www.ncbi.nlm.nih.gov/pmc/articles/PMC3050565/

[XFMQ902SF]

My new theory in regards to n-back effectiveness, is that trainees
must continually improve in n-back level to improve intelligence.
Continue to climb in n-level to experience fluid intelligence
enhancement. This is in line with Awh's work and Jaeggi 2011; where
only the children who improve in n-levels-those who stretched their
working memories, gain IQ points.

[Gwern Branwen]

On Tue, May 29, 2012 at 4:46 AM, XFMQ902SF <kei...@aol.com> wrote:
> This is in line with Awh's work and Jaeggi 2011; where
> only the children who improve in n-levels-those who stretched their
> working memories, gain IQ points.

And this idea based on post hoc analysis is not in line with Redick et
al 2012, as I already pointed out.

--
gwern
http://www.gwern.net

[XFMQ902SF]

Jaeggi 2011 vs Redick 2012. Same analysis, differing results... You
know what would be interesting? If Redick or someone were to conduct a
study and show the control group improving on tests of fluid
intelligence vs the n-back group. We have yet to come across a study
where the control group has greatly outperformed the n-back group, the
way the experimental n-back sometimes outperforms the control. If we
cannot show a consistent IQ gain after 8, 12, 17 days of training for
control groups, the way Jaeggi 2008 did for the n-back group, then it
seems unlikely that Jaeggi 2008 is due to Hawthorne effect what have
you.

On May 29, 9:04 am, Gwern Branwen <gwe...@gmail.com> wrote:

[☉]

On May 29, 8:40 am, XFMQ902SF <keif...@aol.com> wrote:
> We have yet to come across a study
> where the control group has greatly outperformed the n-back group, the
> way the experimental n-back sometimes outperforms the control.

But isn't it precisely here that we should suppose that that is
because the tasks are in an important sense similar, such that the n-
back group basically gets more practice at doing the kind of cognitive
juggling required of them during a speed-oriented (time-based) matrix
test instead of supposing that there is some poorly defined "far-
transfer" taking place? After all, not all tests showed increased
scores for whatever n-back group in whatever study. Your use of the
phrase "greatly outperformed" is rather curious, because one could
just as well say the same of some studies regarding the n-back group.

To me, this seems to be more a matter of making hasty, ungrounded
assumptions based on limited evidence. And this is only partly
scientific, because they didn't do the *real work* to show whether or
not their guess was right, but instead just stick to it and follow
along with "questionable research practices" to put it mildly
euphemistically. Moreover, one also has to consider if the fallacy of
reification was committed, not only because did they not use a test in
a way that would be representative of a normative sample (and so give
one I.Q. scores), since they just point to the numbers as if that
shows anything that isn't a fluke is going on. It is alarming, to say
the least.

argumzio

[Gwern Branwen]

On Tue, May 29, 2012 at 9:40 AM, XFMQ902SF <kei...@aol.com> wrote:
> You
> know what would be interesting? If Redick or someone were to conduct a
> study and show the control group improving on tests of fluid
> intelligence vs the n-back group. We have yet to come across a study
> where the control group has greatly outperformed the n-back group, the
> way the experimental n-back sometimes outperforms the control.

Yes, it would be interesting, but we only have 9 studies or so. Even
if there were zero publication bias, tweaks, modified methodology,
etc. and the control and experimentals were exactly equal, at p=0.05
how often would you expect to see the control outscoring the
experimental with significance?

Incidentally, in Chooi, the experimental mean was 12.235 and the
control was 12.24.

--
gwern
http://www.gwern.net