2013 Jan thru Aug
<p class="MsoNormal" >WALLIS, Joanathan,Frontal Lobes in Decision Making, Cal, Apr 18 2013</p>
<p>
he starts with a patient (Elliot)
made terrible decisions:</p>
<p>
prefrontal cortex and decision-making
EVR 0318 he shows huge frontal lobe lesions...</p>
<p>
which color do u prefer:
normals: have temporal consistency (over time)
but much more random in OFC and ACC damage but not in lateral PFC damage;</p>
<p>
he uses the awake behaving monkey (with electrode recording)</p>
<p>
macaque monkey: good homology</p>
<p>
see wallis 2011 in Nat Neurosci</p>
<p>
homology breaks down when u move to the rat...</p>
<p>
good homology esp in OFC and ACC</p>
<p>
OUTLINE of talk:</p>
<p>
func diffs btwn OFC and ACC</p>
<p>
expts done by Steve Kennerley now asst prof at UCL.</p>
<p>
OFC strongly connects with sensory areas but has few motor connections (unusual for frontal lobe struc)</p>
<p>
in contrast.. ACC strongly connects with motor areas but has few sensory connections
(it's just 2 synapses away from spinal cord)</p>
<p>
reward related connection</p>
<p>
when we started to study OFC...</p>
<p>
idea was OFC would determine value of things in eniviroment (money, cake)
it would compute abstract value signal...
idea is that the motor program then chooses the high value item and plans to get it...</p>
<p>
BUT... need lots of planning:
need to assess risk, effort, time, involved...</p>
<p>
so first we needed to get a handle on info involved</p>
<p>
task; monkey chooses a picture ass oc with difft probs of reward, amounts of juice and
of lever presses required to get a reward.</p>
<p>
monkey choose the more valuable picture on > 98% of trials...
easy task for our monkeys..</p>
<p>
kennerley 2008;</p>
<p>
we drop 32 electrodes per day in OFC, ACCl LPFC (a control area)</p>
<p>
results: one neuron:</p>
<p>
X is time; Y is firing rate</p>
<p>
for this neuron; when picture is presented:
this neuron fires with all 3: incr prob., incr payoff and low effort</p>
<p>
but other neurons encode difft combos of these 3 params.</p>
<p>
BUT... OFC was just as capable of encoding one param as another...
there was no diff among these 3 areas (ACC, OFC, and LPFC)</p>
<p>
what we are manipulating really drives the ACC (but not so much the others)</p>
<p>
(the ACC has 30% of neurons that would encode all three variables)</p>
<p>
so, how does OFC encode values
eg from what u have for lunch vs what car u buy...
the firing goes from zero to 20 hz in these neurons...</p>
<p>
proposed mechanism is
an adaptive gain control of value signal...</p>
<p>
he illustrates with difft food:</p>
<p>
eg if at taco bell: (low value enviro)
then only get one single spike from taco...
so system turns up the gain and u get brisk firing (20 hz)
with tasty salmon... (and vice versa)</p>
<p>
Value contrast in OFC:</p>
<p>
now we see a signal present in the OFC, not present in the other two areas;</p>
<p>
*</p>
<p>
now what happens when animal finds out the outcome 9above was the choice phase)</p>
<p>
firing is similar to reward prediciton error (fires when outcome is better than expected) =
actual outcome - predicted outcome...</p>
<p>
(but dopamine neurons show an inhibition of response when something worse than expected happens)</p>
<p>
two separate populations of neurons here: + prediction errors and negative prediciton errors...</p>
<p>
</p>
<p>
compare the 3 areas: all 3 areas had about 17% of neurons encoidng choice phase only</p>
<p>
summary: OFC keeps a detailed repr of expected value, including how the available outcome
relate to the value enviro</p>
<p>
but the ACC encodes the relationship btwn expectd and actual outcomes,
a caolcution that may be impt for leaning and modifying future behavior...</p>
<p>
Part 2 of talk</p>
<p>
func organiz within OFC
work done by Erin Rich...</p>
<p>
in OFC v detailed cytoarch maps but
how does function map onto this anatomic complexity...</p>
<p>
perhaps two axes in OFC:</p>
<p>
mediolateral axis: rewards are medial and punishments perhaps more lateral
and concrete more posterior and abstract more angterior;</p>
<p>
but morrison and salzmann 2009 j neuro
did not find that hypothesized organizaiton...</p>
<p>
our task: use secondary reinforcer (could accumul a juice account)</p>
<p>
they studied post and negat reinforcement separately...</p>
<p>
two monkeys C and M
performed better on posit trials</p>
<p>
(they can cash in their accounts for juice)</p>
<p>
always cashed in after a fixed # of trials ;</p>
<p>
neg trials decr siz eof bar (how much juice u will get) and vice versa.</p>
<p>
they could find neurons that respond mainly to + pictures and some
that repsond mainly to - pictures...</p>
<p>
two main sulci in OFC: MOS and LOS (medial and lateral)
and 2 types of neurons were completely randmly dispersed</p>
<p>
BUT... there are neurons that encode siz eof bar and the # of trials...</p>
<p>
lateral of LOS almost all the neurons encode the size of the bar</p>
<p>
but medially... its about external tracking (eg # of trials)</p>
<p>
now we studied the abstract-concrete axis;
some trials give a juice reward (rather than secondary reinforcement)</p>
<p>
2 to 1 (2 in the bush is worth 1 in the hand) approx.</p>
<p>
found reinforcer independent neurons (did not matter whether reward was juice or juice account)
these tended to be located more anteriorly...</p>
<p>
conclusion: no support for valence axis (medial to lateral)</p>
<p>
but did find support for concrete to abstract and for external tracking;</p>
<p>
OFC anterior: more abstract rewards...</p>
<p>
conclusion: our results support a hierarch organiz of OFC with progressivley more abstract value info encoded by more anterior arreas.</p>
<p>
This organizaiton scheme parallels that in lateral frontal cortex, where progressivley more abstract relationships
between stimuli and actions are represented by more anterior frontal regions.</p>
<p>
Jay Q: relationship btwn consummatory behavior and the reward representaiton eg
might represen. be quite different for difft kinds of rewards;</p>
<p>
his comment later: OFC lesion families are mainly concerned abt social deficits;
he wonders: do our jucie rewards map onto eg social rewards (Michael Platt: eg the monkey porn study)</p>
<p>
Bill N questions monkey/ human homology of OFC anatomy/ functional mapping;</p>
<p>
JW comment: in humans the responsive eregion is usu in VMPFC;</p>
<p>
</p>