Factors Influencing Conditioning • Intensity • Attention • Contiguity (aka “when”) • Relevance • Surprise • Contingency (aka “whether”) Next
Factors Influencing Conditioning Intensity Attention Contiguity (aka “when”) Relevance Surprise Contingency (aka “whether”) Next.
Dec 19, 2015
Welcome message from author
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
Factors Influencing Conditioning
• Intensity
• Attention
• Contiguity (aka “when”)
• Relevance
• Surprise
• Contingency (aka “whether”)
Next
CS Intensity Affects Rate but not Asymptote of Conditioning
cs US
CS US
Weak CS
Strong CS
Suppression and CS Intensity
Another CS Intensity Effect
Overshadowing – the more salient CS wins if two CS are trained in compound
Group Treatment Test xOvershadow Ax US crControl Ax US CR
US Intensity Affects Rate and Asymptote of Conditioning
CS us
CS US
Weak US
Strong US
Suppression and US Intensity
Back
CS Preexposure Experiment
(Latent Inhibition or LI)Group Phase 1 Phase 2 Test CS Experimental CS CSUS cr Control ---- CSUS CR
• Because the CS is a benign stimulus it will lose the capacity to command attention if preexposed
• Relation to schizophrenia
Back
CSUS
Delay
CSUS
Trace
US
Explicitly Unpaired
Wea
ker
cond
itio
ned
resp
ondi
ngTemporal Contiguity
CS
CS
US
Simultaneous
Back
Is forward contiguity sufficient [enough]?
CS-US Relevance
From Garcia & Koelling, 1966 Back
Blocking and Surprise
Group Stage 1 Stage 2 Test ResultBlocking AUS ABUS B? crControl ABUS B? CR
Back
A Contingency Experiment
Positively Correlated
CS
US
Chance of US per CS = 2/4 = .5
Chance of US outside CS = 0/10 = 0
A Contingency Experiment
Uncorrelated
CS
US
Chance of US per CS = 2/4 = .5
Chance of US outside CS = 5/10 = .5
2/4 = .5
A Contingency Experiment
Negatively Correlated
CS
US
Chance of US per CS =
Chance of US outside CS = 5/10 = .5
0/4 = .0
It’s a little like…
Animals are scientists, trying to make causal predictions.
…trying to determine whether the US is contingent on the CS
Quantifying
• p(US|CS) = proportion of CS trials with a US
• p(US|no CS) = proportion of “background” only trials with a US
p = p(US|CS) - p(US|no CS)
Some Examples
p(US|CS)
• 20/20 = 1.0• 15/20 = .75• 10/20 = .50• 10/20 = .50• 0/20 = 0
p(US|no CS)
• 0/60 = 0• 6/60 = .10• 30/60 = .5• 45/60 = .75• 60/60 = 1.0
• 1.0• .65• 0• -.25• -1.0
p
12345
P(US/no CS)
P(U
S/ C
S)
0 1.0
1.0
Negative
Positive
1
2
3 4
5
+1.0+.65
-.25
-1.0
Consequences for Controls
• Selection of appropriate control depends on your theory– explicitly unpaired (CS pairings)– uncorrelated/truly random control
(contingency)– CS alone (sensitization)– US alone (sensitization caused by arousal)
Rats as Statisticians?
US
CS
no US
no CS
P(US/CS)
P(US/no CS)
Better Idea
• Background becomes associated with the US
• Background competes with CS for association with the US
Related Documents
