What makes a good CS and US: novelty of stimuli
- classical conditioning is slower when the stimuli used are familiar to the subject (novel stimulus should have higher rates of responding)
→ latent inhibition (CS pre-exposure) effect: caused by repeated exposures to the CS before the CS is used in conditioning trials (harder to form associations)
→ US pre-exposure effect: caused by repeated exposures to the US before the US is used in conditioning trials
What makes a good CS and US: salience of stimuli
- The stimuli to-be-conditioned must be noticeable
→ intensity of US and CS
→ biological relevance to animal model (more naturalistic CS = more salience)
→ biological relevance to animal state (food for hungry animal, water for thirsty animal)
What makes a good CS and US: CS-US relevance (belongingness)
- Selective association: do the CS and US “go together” naturally
→ taste “goes with” sickness
→ audio visual “goes with” shock
Learning without an US
- Classical conditioning would be very limited if direct exposure to the US was required for learning to occur
→ there are situations where learning (associations) occurs without a US - higher-order conditioning: CS may serve as a “US” once conditioned
- sensory preconditioning (CS that are paired → aversive US → aversive CR to any of the CS)
US as a determining factor for the CR
- The response depends on the US that elicits it
→ ex. Pigeons given the same CS (light) but different US (food or water), gave different responses for each US
Stimulus-substitution model (Pavlov)
- CS activates neural circuits previously activated by the US
- CS becomes a surrogate (substitute) US
Ex. Bell → bone (new neural connection formed after conditioning), bone → saliva (pre-existing neural connection)
CS as a determining factor of the CR
Rat example
- rat used as a CS to signal food US
→ predictions based on stimulus - substitution model: rat should bite at or made chewing motions
→ actual results: CS elicited social affiliate CRs (pawing, grooming, crawl-over)
CS - US interval as a determining factor of the CR
Quail example
- conditioned male quail with either short or long CS-US interval
→ CS = stuffed quail, US = access to female quail at the end of interval
→ short interval spent more time near CS (went right to consummatory behaviour)
→ long internal was engaged in locomotor behaviour (general / focal search, appetative)
S - R learning
New stimulus-response connection between CS and CR
- US becomes unimportant; the CS directly triggers the CR, bypassing the need for the US. The US is no longer needed for the response
- early in conditioning: CS → US → UR
- after extensive conditioning: CS → CR
S - S learning
CS activates a mental representation of the US (in line with Pavlov’s stimulus-substitution)
- US remains important
- early conditioning: CS → US → UR
- after extensive conditioning: CS → US representation → CR
US devaluation paradigm
Phase 1:
Experiment group 1 and 2 receive same thing (CS → US → CR)
Phase 2:
Experiment group 1: US gets devaluated
Experiment group 2: value of US stays the same
Test:
If S-R, both groups should respond at high levels to CS. If S-S, experiment group 1 should respond at lower levels to CS compared to experiment group 2.
Results:
Support S-S model of learning (devaluation caused a change in behaviour); US is important
Blocking effect
Interference of the conditioning of a novel stimulus because of the presence of a previously conditioned stimulus
→ before blocking was discovered it was thought that temporal contiguity (pairing two CS together that predicted a US) was sufficient for learning associations
R-W model
- Standard mathematical model for classical conditioning
- Expectation of US related to associative properties of stimuli preceding it (CS)
→ less surprising as trials go on
Symbols:
λ = maximum possible associate strength (what occurs, when learning plateaus)
V = current associate strength (what is expected)
k = related to the salience of the US
(λ - V) = “surprisingness” of US after presentation CS
→ when λ = V, there is no more room for learning (no surprise)
R-W and blocking
- When λ = V, one stimulus already perfectly predicts the US
→ when a novel stimulus is presented alongside the old stimulus (which predicts the US), there will be no surprise or associative value
The comparator hypothesis
- Comparator cues: other cues present when the target CS is being conditioned
- in blocking paradigm:
→ context: previously trained CS(A)
→ target: CS(B) - comparator hypothesis states that the notion of responding to CS(B) is blocked, not learning CS(B) itself
The comparator hypothesis and blocking
- CH revaluation procedure: after phase 2 ([A + B] → US), eliminate responding to CS(A) through extinction ([A] → no US)
- test for CR to CS(B), a response demonstrates that performance was blocked rather than learning
