Battista, Christian
<p>Jan 25, 2014 02:04 PM cbattist@stanford.edu</p>
<p>
below is his FriSem of Jan 24 2014. in Vinods Lab (post-doc); age 32; formerly at U West. Ontario.</p>
<p>
<b>Battista, Christian, Neural Mechanisms of Arithmeti</b>c, Stanford, Jan 24, 2014
This is work that Christian did while he was at <b>Univ of Western Ontario</b>.
He is now a <b>post-doc in Vinod Menon's Lab</b> at Stanford (on Arastradero).</p>
<p>
why study arithmetic?.. It is phylogenetically recent and unique to humans.</p>
<p>
What we know...</p>
<p>
parietal cortex is important..</p>
<p>
see eg Gerstman 1945...</p>
<p>
Dehaene 2003 <b>angular gyrus involved in fact retrieval</b>...</p>
<p>
but <b>intraparietal sulcus IPS is involved in procedural calculations</b>.</p>
<p>
(but this is NOT a strict dichotomy,)</p>
<p>
factors modulating brain activity during mental arithmetic:</p>
<p>
Delazer 03...</p>
<p>
compare trained vs untrained arithmetic problems...</p>
<p>
untrained... <b>doing untrained problems exhibits more frontal and parietal activation</b>...</p>
<p>
but in problems that have been trained, you get more activation just of <b>angular gyrus</b>...</p>
<p>
and same for strategy...</p>
<p>
see Grabner 2009...</p>
<p>
one question: did u just retrieve from memory or were there actually procedures involved?</p>
<p>
so, with <b>retrieval get angular gyrus activity</b> but
with <b>procedural retrieval get IPS and frontal activation</b>..</p>
<p>
how abt individual diffs...
Rivera 05;</p>
<p>
eg press button if correct is 6 + 2 = 8 correct; how abt 3 +2 = 6 ... not correct</p>
<p>
get positive correlation with age (ie, more fact retrieval as you get older)</p>
<p>
Outstanding Qs...
does arithmetic training effect reflect a shift in strategy...</p>
<p>
do recently learned arithmetic facts differ from remotely learned arithmetic facts?</p>
<p>
and do strategy and training interact?</p>
<p>
But... indiv variability in strategy use could confound results (as in Campbell)</p>
<p>
Christian developed <b>CAT = customized arithmetic training</b></p>
<p>
start out with individually tailored training sets</p>
<p>
strategy use is balanced at start of training</p>
<p>
CAT fMRI...</p>
<p>
20 adults ages 23 to 30</p>
<p>
calibration stage: training set (do 12 reps per day over 6 days (S does this at home.)
4+3 =? is the answer 7 or 8..
indicate with a left button press for 7...
now describe your strategy... (did you memorize the answer or or did you use mental arithmetic)
this was self report..</p>
<p>
want to compare to 10 year old kids... so this was just simple integer arith...</p>
<p>
stage 1: present probs to S, and incr size of operands to 45 to 50 then decrease them.</p>
<p>
stage 2: use info from stage 1 to generate potential calculated and memorized probs
(eg peope have trouble with 6s and 7s...
he shows <b>RTs = reaction times</b> vs solution size (the sum of operands)</p>
<p>
<b>the memorized ones are FAST</b> and take only 2.5 to 2.8 seconds (eg for sum below 20 to 30)
vs the <b>procedural ones (sum of between 40 to 80)</b> and RTs of 2.7 to 3.4 seconds...</p>
<p>
he shows one particular subject:</p>
<p>
big number operand on Y axis and small number on X axis
and <b>he shows RT by bubble size</b>...</p>
<p>
small numbers are fast and big numbers are slow
(but he looked at numbers with tens in them and cut them out)</p>
<p>
he shows a slow guy versus a fast guy...
(fast guy would go really fast with all the numbers up to 30...</p>
<p>
huge diff in profile...</p>
<p>
(but jay cites this example: 23 + 64 (add in parallel 20 + 60 and 3 + 4)</p>
<p>
noah: how abt test/ retest reliabilities for RTs..</p>
<p>
he says...yeah... might just also be getting better at calcs (faster)</p>
<p>
subjects work on this 12 repetitions per day for 6 days...</p>
<p>
at home do web based training...</p>
<p>
herb: were they supplied with the correct answers... No.. only an X if wrong..</p>
<p>
first calibration; ten training sets and then the fMRI test...</p>
<p>
strategy types...</p>
<p>
calculated - untrained problems solved by calc (28 trials)</p>
<p>
and had (remote) memory... untrained probs solved by memory</p>
<p>
and recent memory.. calc probs memorized thru training
trained calcualted...
etc</p>
<p>
fMRI untrained, calculated probs are the slowest...</p>
<p>
Outstanding Qs..</p>
<p>
will the CAT produce viable results?</p>
<p>
yes... <b>memorized problem do activate angular gyrus</b>...</p>
<p>
and <b>calculated lights up frontal and parietal areas</b> ...</p>
<p>
next Q... does the training effect reflect a shift in strategy?</p>
<p>
comment: perhaps this is just testing working memory WM
(follow on to my Q abt cog burden and arithmetic carries)</p>
<p>
comment: why is primary visual lighting up? (cuz we present info visually)</p>
<p>
*
did we replicate the grabner study... yes we did...</p>
<p>
recently memorized lights up angular gyrus vs calculated (involves IPS and frontal areas)</p>
<p>
next Q.. do recently learned arith facts facts differ from remote arith facts...</p>
<p>
ie recent memory (just learned) vs remote memory...</p>
<p>
there is more frontal activation with recent memory</p>
<p>
recently memorized still looks more like calc probl with IPS activation and frontal activ...</p>
<p>
discussion (jay and herb) when they self reprot "memorized" was it actually memorized or calculated.
a lot of prior lit on this... his A: "describe your steps"
Jay elaborates: this might be an internal RACE between memory and calculation with
both going on in parallel. If memory wins, the subject says "I just knew it"</p>
<p>
he shows HC = hippocampal ROI (some problems at a midpoint btwn mem and calc)</p>
<p>
do strategy and training interact?
(eg is calculation affected by training)</p>
<p>
ROI angular gyrus... involved in symbol reference. </p>
<p>
its not all about a shift in strategy...</p>
<p>
<b>surprise finding: putamen activation</b>:
assoc with procedural memory... trained calculation (still called calc but doing much faster)</p>
<p>
yes... strat and training do interact in angular gyrus and in putamen...</p>
<p>
(angular gyrus can be modulated by training... hence... faster memory
even tho it still involves angular gyrus)</p>
<p>
future studies...
use CAT to control diffs in strategy use..</p>
<p>
do DTI data and learning rates
Run CAT on younger subjects...
explore encoding success...</p>
<p>
Advisor for this work: <b>Dan Ansari..(Western Ontario)</b> this was Christian's PhD work;
(Western Ontario may be #1 center for cog neuropsych in Canada:
Adrian Owen and Mel Goodale there. (McGill: more oriented toward medical apps).)</p>
<p>
noah... u asked them how they did it? yes... how did u do the prob.
calculation is a v broad category... eg overall workload, etc.
we do not have the power to look at that (cannot do that (break it down by specific strategy)</p>
<p>
Q how abt eye movements while doing this? no... not done..</p>
<p>
</p>