Rescorla-Wagner model challenges

what is extinction in the RW model

CS presented without US - negative prediction error (expected US, got nothing), CS-US association weakens

why is spontaneous recovery a problem for the RW model

if extinction destroys the CS-US association, the response shouldn’t recover

Rescorla (2004) - spontaneous recovery

used a clever design to deconfound time of testing from time since extinction.

renewal

after extinction in context B, returning to conditioning context A restores the conditioned response. shows extinction is context specific not true unlearning

Rescorla (2008) - renewal

extinction is context specific: conditioning in context A, extinction in context B and testing in context A shows renewed responding

why is renewal clinically important

found context induced relapse of conditioned responses to ethanol cues in rats

alternative view of extinction to the RW model (Pearce & Hall, 1980)

extinction doesn’t destroy the original CS-US association. instead, a new CS, no US association is learned that inhibits the original

how does the inhibitory learning view explain spontaneous recovery and renewal

the original CS-US link remains intact. the new inhibitory CS - no US link is context dependent, so changing context or waiting allows the original association to re-emerge

RW model’s assumption about stimulus compounds

the whole is equal to the sum of its parts. the associative strength of compound AB = V(A) + V(B)

feature positive discrimination

A - (no US), AB+ (US). animals learn to respond to AB but not A.

stimulus compound

when stimuli are presented as a compound (A+B), they compete for a limited associative strength which explains blocking and overshadowing

why is negative patterning a problem for RW

(A+, B+, AB-) is a problem because it requires the associative strength of AB to be lower than the sum of its individual parts A + B

Pearce (1987) - solution to the stimulus compound problem

the whole is not the sum of its parts, AB and A are unique configurations not additive combinations

Wagner (2003) - solution to stimulus compound problem

the whole is equal to the sum of its parts as well as a unique configural element. combines elemental and configural approaches

RW model’s view on stimulus processing

the effectiveness of the US diminishes as it becomes expected, which is why blocking occurs

what neuroimaging evidence supports RW’s US processing view

fMRI study: tone paired with noise 100% of the time (CS100) showed reduced amygdala response vs tone paired 50% (CS50+). UCR diminution correlates with UCS expectancy

what challenge does the CS processing pose for the RW model (Kaye & Pearce, 1984)

showed changes in orienting response to CSs during conditioning, suggesting CS processing also changes, not just US processing as RW assumes

what are the three main challenges for the RW model

1) extinction isnt true unlearning, 2) stimulus compounds aren’t purely additive 3) processing changes too not just Us processing