Information Processing View on Classical Conditioning: 5 Experiments & Implications, Slides of Public Health

Download Information Processing View on Classical Conditioning: 5 Experiments & Implications and more Slides Public Health in PDF only on Docsity!

An Information Processing

Perspective on Conditioning

Five Classic Experiments

• Rescorla’s background conditioning

experiment (the truly random control)

• Kamin’s blocking experiment• Reynold’s overshadowing experiment• Wagner’s relative cue validity experiment• Conditioned inhibition experiments

Blocking

Result: No conditioning to the CS that first appears in

compound (the blocked CS) Implications• Temporal pairing not sufficient• Conditioning occurs only if US “surprises” subject

Overshadowing

•^

Two pigeons trained withtwo compound cues: one(+) yielded food whenpecked; the other (-)yielded nothing

-^

Tested with individualelements

-^

One bird pecked only attriangles, the other, only atred

-^

Implication: when twopredictors (red & triangle)are redundant, one isdisregarded

Conditioned Inhibition

Implications:• Temporal pairing neither necessary nor sufficient for

conditioning

• Negative contingency just as effective as positive

contingency

So what’s the problem?

"We provide the animal with individual

events, not correlations or information, andan adequate theory must detail how theseevents individually affect the animal. That isto say that we need a theory based onindividual events.”--Rescorla (1972, p. 10)

Shannon meets Pavlov

• From an information processing

perspective, conditioning is driven byinformation

• The information an event provides to a

subject is measured by the reduction in thesubject’s uncertainty--Shannon, 1948

Principles

-^

Subjects respond only to informative CSs

-^

CSs inform to the extent they change uncertaintyabout the time to the next US

-^

Bandwidth maximization by minimizing numberof information carrying CSs

-^

Information in a protocol is carried by thetemporal intervals between events and thenumbers of events

-^

Weber’s law: uncertainty (noise) in therepresentation of intervals and numbers isproportional to their magnitude

Can Shannon Information

Predict Acquisition?

Preliminaries•^

The appearance of aconditioned responsein the course ofconditioning is abrupt

-^

As if it were theoutcome of anevidence-baseddecision process

Conditioning Protocols

Accumulated Bits at

Acquisition

N

a

log

CS

B

N

a^

is the number ofreinforced CSpresentations prior tothe onset ofconditioned behavior

Contribution from Fixed CS-

US Interval

H

log

2

e

(^2)

^

Entropy of a Gaussian

1 2

log

2

e 2 

 



log

2

w

Additional info available

N

a^

log

2



CS  B  

 

2

N

a

where w = Weber fractionCumulative Bits at Acq

Cumulative Dist of G&B Data

• Cumulative distribution

shows the fraction of thesubjects that had begun torespond to the CS as afunction of the amount ofinformation so farcommunicated; the longerconditioning lasts, themore bits communicated

• The median amount of

information at acquisitonis roughly 100 bits(median = # of bits atwhich fraction = 0.5)

Partial Reinforcement

-^

Partial reinforcementhas no effect onhence, it should haveno effect on reinf toacq, hence no effect onbits cumulative bitsconveyed atacquisition



CS



B