PSY260 Lecture 4: Classical Conditioning

Classical Conditioning (aka Pavlovian conditioning)

• Begins with an innate (unlearned) reflex

  • Unconditioned stimulus (US) - food

  • Unconditioned response (UR) - salivation

  • US → UR response does not need to be conditioned, it is innate

• A neutral stimulus (NS) [bell] is then repeatedly presented with the US → produces a new reflex

  • NS becomes CS (bell)

  • CS can now elicit CR (salivation)

  • Bell ringing → dog salivates before food is even presented

Classical Conditioning: Reasoning

• Helps an organism prepare for the future

  • Bell triggers behaviours to prepare for the food

• Classical conditioning represents an association between US and CS

• Chance to test Aristotle’s principles of association

  • Contiguity: same time/spatial location

  • Similarity: similar stimulus may elicit same CR (slighty different bell may still elicit salivation) → proportional to similarity of stimulus

  • Frequency: ↑ frequency of association → ↑ strength of association

Appetitive Conditioning

• When learning helps predict something good or appetizing

  • Elicits approach behaviour

• Ex. Quail Sex Conditioning

  • Exposure to female (US) → innate approach behaviour (UR) in male quail

  • NS = tone/light → becomes CS

  • After pairing US with CS, CS starts to produce CR

  • Male quail approaches light before female even shows up

Aversive conditioning

• New CS → CR reflex helps avoid noxious US

  • Elicits avoidance behaviour

  • Ex. fear conditioning → CR is unpleasant so you avoid it

• Ex. Fruit Fly Fear Conditioning

  • Exposure to shock (US) → innate escape/avoidance behaviour (UR)

  • NS = odour → becomes CS

  • After US is paired with CS, CS starts to produce CR

  • Works with only 1 trial

  • Flies move away from odour chamber before shock

Eyeblink Conditioning

• Puff of air to eye (US) → eyeblink (UR)

  • Innate (no conditioning needed)

• NS → CS = tone/light

  • Gradually comes to produce an eye closure (CR)

  • CR is in anticipation of air puff (unpleasant)

• Aversive conditioning: CS → CR prepares to avoid aversive US

  • CR = eye blink to avoid US = air puff when seeing CS = tone/light

• Works on rabbits and humans but takes many trials

Eyeblink Conditioning: Steps

1. Initially, NS (tone) → no response

2. NS paired with US (air puff) → UR (eyeblink)

3. NS becomes CS → produces CR (eyeblink) even without US

• Note: CR eyeblink is in anticipation of US (flinch) while UR is after airpuff (reflex)

Eyeblink Conditioning: Trials

Eyeblink conditioning: Timeline

• Day 1: initially, NS does not produce any response

• After extensive training, CR emerges as a slow squeezing shut of the eye before US (air puff)

  • CR is similar to UR but not the same → CR is anticipatory but both are blinks

• Other cases: CR can be very different from UR

Conditioned Compensatory Response

• CR that is the opposite of the UR, helping to balance/correct for the US-UR response

  • Overcorrection to maintain homeostasis

  • Body tried to get ahead of UR by counterbalancing the other way

• Ex. inject adrenaline (US) → HR increase (UR)

  • Repeate procedure in same testing chamber (CS)

  • Eventually, CS comes to produce a decrease in HR (CR) to maintain homeostasis → counterbalance against expected adrenaline injection

• This is an unconscious process

Extinction

• Breaking the association between the CS and US can extinguish the new CS → CR reflex

  • Present the CS alone repeatedly

  • Initially, CS → CR link is strong

  • Repetition: CS becomes less effective and association weakens

• Extinction probably doesn’t erase the CS-US connection, just inhibits it (inhibitory neurons)

  • Stress, new context, passage of time → can make the CS effective again

    • Easy to “wake up“ extinguished association

  • This suggests that the classically conditioned memory survives extinction

Rules of Classical Conditioning

1. Timing

2. Blocking

3. Latent inhibition

4. Associative bias

Timing

• If the US precedes the CS → no learning

  • Backward conditioning

  • Ex. eyeblink: if you puff air before tone → not effective, no preparation

• Best learning: CS then US very quickly

  • Ex. eyeblink: tone then puff immediately

• Worse learning: CS then US at a delay

  • Trace conditioning

Trace conditioning

• Delay in US after CS

  • Results in worse learning

  • More difficult to form association

Blocking

• Previously learned association can block the formation of a new association

  • Similar to proactive interference (old blocks new)

• Ex. Kamin’s Blocking Paradigm in rats

  • Control: trained to tone+light CS → shock US

    • After, both tone and light alone produce moderate CR (freezing)

  • Pre-trained: trained with just light CS → shock US

    • Then, same tone+light CS → shock US training

    • After, light produced large CR and tone produced none

• Tone wasn’t providing any new info that light hadn’t already provided

  • Tone learning blocked by previous light CS → shock US association

• Indicates that classical conditioning is about tracking information in the environment

  • When CS is redundant to what is already known → no new learning

  • No new info → no new learning

Latent Inhibition

• First, pre-expose animal to NS/CS repeatedly

  • Ex. eyeblink = only tone, no air puff

  • No learning happens - animal learns that NS is not important

• If you then pair NS/CS with US, learning is inhibited

  • Lag in learning because animal learned that NS doesn’t mean anything and stopped paying attention

• Later overcomes this lag and is able to learn association, just delayed

Associative Bias

• Some associations are innately easier to make:

  • Tone vs. taste paired with poison: only taste provokes CR

  • Tone vs. taste paired with shock: only tone provokes CR

• In nature, tastes go with getting sick, sounds with getting hurt

• Seems we have some innate preferences for forming associations that can override statistical correlations

Models of Classical Conditioning

• The rules for classical conditioning are quite complex

• Researchers have turned to models of classical conditioning - formal sets of rules for when learning will occur

  • Mismatch between model and data provides clues to what we still need to understand

• 2 popular traditions in models of classical conditioning:

  • US modulation approaches - Rescorla-Wagner

  • CS modulation (attentional) approaches - Mackintosh

US Modulation Theory: Rescorla-Wagner

• Learning from errors

  • The amount of learning depends on prediction error: how much reality differs from expectation

    • ↑ prediction error → ↑ learning

• Is the US totally unexpected?

  • Large prediction error → lots of learning

• Is the US accurately predicted by the CS?

  • No prediction error → minimal learning

• BASICALLY: learning only occurs through failure of your expectations

Rescorla-Wagner Model: Math

• Each stimulus has a weight (W) for predicting a US

  • Weights go from 0 (NS) to 1 (certain)

• Sum of the stimulus weights is the animal’s prediction

  • Accumulation

• When the predictino is wrong, the weights are adjusted

  • Error: difference between expected and actual outcome

    • Prediction error: actual US - expected US

  • Adjustment: each active stimulus adjusted by a proportion of Error

    • Change in stimulus weight: Learning rate x Prediction error

    • When error is 0 = no learning occurs (asymptote) → means that W = 1.0 and what you expected was already true

Rescorla-Wagner Model: Blocking

• Can be used to explain blocking

  • Blocking study:

    • tone+light CS → shock US

    • light CS → shock US then tone+light CS → shock US (only light elicits CR)

  • RW: second stimulus adds no new info

  • No/low prediction error once association with the initial CS is learned, so no real need to learn about the new CS

CS Modulation Theory: Mackintosh

• Proposed that CC focused on attention and the way the CS is processed

  • Stimuli have a salience → determines attention

• Amount of learning depends on salience and relevance of CS

  • Cues that are better predictors warrant greater attention and have higher associability as a result

    • Ex. bell is previously paired with food: bell has high salience and should be paid attention to (you’ll get food!)

• BASICALLY: how much attention each stimulus should be paid (salience)

Mackintosh Model: Latent Inhibition

• Can be used to explain latent inhibition

  • Repeated exposure with no consequences → decreases salience

    • Form of habituation

    • Pre-exposure to CS decreases attention for that stimulus

  • Makes the CS harder to learn about in the training phase (reduces associability)

Mackintosh Model: Associative Bias

• Some stimuli have more salience when it comes to certain outcomes/effects

  • Recall that taste/smell is highly associable with sickness/poison

  • Tone is more highly associable with physical pain

Brain subtrates

• Since Pavlov, there has been interest in understanding how classical conditioning works at the neural level

• 1980s: Thompson and colleagues discovered that eyeblink conditioning in rabbits depends on cerebellum

Cerebellar substrates

• CS-US association may be stored in:

  • Cerebral interpositus nucleus

    • Site of association between CS and US

  • Purkinje cells of cerebellar cortex

    • Carry signals related to discrepancies between expected and actual outcomes

    • Influence timing and magnitude of eyeblink CR

• Recordings from cerebellum before/after learning help demonstrate where the memory is stored

Interpositus Nucleus

• We see huge increase in neuronal firing during period between CS and US

  • Involved in prediction of US and production of CR

Purkinje Inhibition of Interpositus Nucleus

• Inhibitory fibres

• Well-trained animals: Purkinje cells switch OFF in response to CS

• Shutting off Purkinje inhibition of interpositus nucleus → enables CS to generate CRs

  • Purkinje fibres OFF: CS → CR ON

Hippocampus

• Removal of the hippocampus (with medial temporal lobe)

  • Treatment for epilepsy

  • Doesn’t alter basic classical conditioning paradigms

  • Eliminates latent inhibition and trace conditioning

• Plays an important role in processing/binding contextual information

  • Ex. context fear in fear conditioning → the room you’re in, who’s involved etc

    • Gives valuable info to predict something scary

Where is Conditioning Taking Place?

• We don’t have a single clear answer, and there seem to be many areas working together

• Maybe:

  • US modulation (Rescorla-Wagner) occurs in cerebellum

  • CS modulation (Mackintosh) occurs in hippocampus and medial temporal lobes