Learning Processes, Behaviour and Performance - Recap (Video Notes)

What is Behaviour?

  • Behaviour defined as any outward or inward response to a stimulus
    • Includes actions, speech/language, feelings, and thoughts
    • All behaviour is elicited by a stimulus, whether observed or unseen, conscious or unconscious

Part 3: Changes in Behaviour from Experience

  • Repeated Stimulation
    • Repeated exposure to the same stimulus has two effects:
    • Habituation: decline in responding
    • Sensitisation: increase in responding
    • These changes are considered non-associative learning

Habituation

  • Startle magnitude experiments
    • Trials: 1 per day, every 3 seconds
    • Observations: long-term habituation vs. short-term habituation
    • Spontaneous recovery observed
  • Examples of habituation
    • Startle response: part of an organism’s defensive reaction to threat
    • Used to study fear and learning
    • Commonly studied in rats, mice, rabbits, humans using loud tones or bright lights
  • Additional examples
    • Salivation and hedonic ratings
    • Stimulus-specific and attention-dependent
    • Relevant for controlling eating behaviour
    • Involve trials and dishabituation
  • Infants and facial recognition
    • Infants habituate to familiar faces
    • Preferential, longer viewing time for novel faces, even when faces are shown at different orientations
  • Habituation parameters
    • With the same exposure to the stimulus, habituation occurs more rapidly with shorter interstimulus intervals
    • With the same number of presentations and interstimulus interval, habituation occurs more rapidly if the stimulus is presented for longer periods at each presentation
    • With the same interstimulus interval and duration, habituation occurs more rapidly if the stimulus is presented more frequently

Part 5: Emotions and Motivated Behaviour

  • Emotions can be elicited by stimuli in addition to external responses
  • Habituation and sensitisation extend to changes in emotions and forms of motivated behaviour (e.g., feeding, exploration, aggression)
  • Drug addictions are a particular area of interest
  • Solomon & Corbit (1974): biphasic intense emotional reactions
    • One emotion during elicited stimulus; the opposite emotion after stimulus termination
  • Emotional reactions change with experience
    • Primary reaction weakens; after-reaction strengthens
    • Habituation = drug tolerance
    • Extends to other emotion-arousing stimuli (e.g., love, attachment)
  • Opponent Process Theory (OP) (neurophysiological homeostasis)
    • Emotions are linked to neurophysiological mechanisms that maintain homeostasis
    • Behaviour is the net result of:
    • Primary process (direct effect of the emotion-arousing stimulus)
    • Opponent process (counteracts the direct effect)
  • Primary process (P-process) and Opponent process (O-process)
    • P-process: quality of the emotional state in presence of the stimulus
    • O-process (b-process): generates the opposite emotional reaction; lags behind the primary disturbance
    • Underlying opponent processes produce the manifest affective response observed during trials
  • After extensive stimulus exposure
    • Primary reaction weakens; after-reaction strengthens
    • Similar to habituation: primary reaction wanes, after-reaction grows
  • Addictive implications
    • Frequent users may not experience high highs
    • Addictions may serve to reduce aversiveness of after-reaction (aversive withdrawal states)
    • Neurophysiological evidence: reduced activation in reward circuits; increased activation in anti-reward circuits
  • Conditioning and drug tolerance
    • Can be applied to conditioned drug tolerance (a form of associative learning)
    • CS (cue) paired with US (drug effects)
    • After repeated CS–US pairings, the B (opponent) response activates upon CS alone
    • Tolerance forms as the body anticipates the US (needs more of the drug to achieve initial UR levels)
    • Novel stimuli (e.g., different environment) can disrupt homeostasis and increase overdose risk

BASIC CONCEPTS OF CLASSICAL CONDITIONING (CC)

Classical Conditioning Paradigm

  • Before conditioning
    • Unconditioned stimulus (US) naturally provokes an unconditioned response (UR)
    • Neutral stimulus (NS) is present but produces no conditioned response (CR)
  • During conditioning
    • CS is paired with US, leading to acquisition of a conditioned response (CR)
  • After conditioning
    • CS elicits CR similar to UR
  • Common examples
    • Food → Salivation (UR); Salivation becomes CR to CS (e.g., bell) over conditioning
    • Sexual arousal as CR to CS
    • Shock → Startle as UR; CS paired with shock leads to CR (startle) to CS

Extinction (Inhibitory Conditioning)

  • Extinction = stop reinforcing a previously excitatory CS
  • Inhibitory associations = CS predicts the absence of the outcome
  • Extinction learning is slower than acquisition

Inhibitory Conditioning Procedures

  • Pavlovian/CS-based procedures
  • Variants include CI (inhibition of delay US), Explicitly unpaired, Backward conditioning, etc.
  • A+ / AX- style designs illustrate how one cue can inhibit another when trained in compound
  • Differential and CI (inhibition) approaches demonstrate inhibitory learning

Factors that Influence the Effectiveness of Conditioning

  • Contiguity (closeness in time) between CS and US
  • Contingency (predictive relationship) between CS and US
  • CS pre-exposure (Latent inhibition): pre-exposure to CS slows acquisition
  • Overshadowing: more salient CS dominates learning over less salient CS in a compound
  • Blocking: a previously learned CS blocks learning to a new CS when both are paired with the US
  • Salience: the conspicuousness or prominence of a CS affects learning rate
  • Latent inhibition and US pre-exposure show context-specific effects

Higher-Order Conditioning

  • Higher-order conditioning (SOC) allows learning without a direct CS–US experience
  • Second-order conditioning: CS1 → CS2 → US chain where CS2 can elicit CR via CS1
  • Sensory Preconditioning (SPC): Phase 1 stimulus pairings create S–S associations that later support CS–US conditioning
  • Shoulder of debate: extinction of first-order stimuli can affect second-order responding; SPC may rely on S–S rather than S–R associations

Rescorla and Wagner Model (1972)

  • Key idea: learning is driven by total error reduction; all cues present on a trial contribute to prediction

  • Model accounts for cue competition phenomena (e.g., blocking, overshadowing) better than contiguity alone

  • Learning rule (simplified):

    \Delta V = \alpha\beta\big(\lambda - \sum V_n\big)

  • Where:

    • \Delta V: change in associative strength for a CS on a given trial
    • \lambda: maximum associative strength (salience of US)
    • \sum V_n: total associative strength of all CSs present on that trial
    • \alpha: learning rate/associability of the CS (0 < a < 1)
    • \beta: learning rate/associability of the US (0 < b < 1)

  • The model can be written without explicit a and b in some presentations as a form of total error reduction

  • Example (blocking scenario):

    • If cue A has value 0.8 (A = 0.8) and X is introduced in compound with A
    • Trial-wise ΔV calculations show predictive error becomes zero after a few trials, blocking X from gaining associative value
    • Demonstrates that the total associative strength cannot exceed the US value

  • Example (overshadowing):

    • If cue A is more salient than X, A gains most of the US value early, leaving little for X
    • Demonstrates that a more salient cue can overshadow a less salient cue in learning

  • Problems with RW Model

    • Spontaneous recovery, latent inhibition, etc. indicate limitations of a simple error-c-correction framework

Reinforcement and Punishment

  • Role of the outcome: according to the Law of Effect, rewarding outcomes strengthen S-R associations; punishing outcomes weaken them
  • Reinforcement: increasing operant responding (e.g., food)
  • Punishment: decreasing operant responding (e.g., shock)
  • Both reinforcement and punishment can involve giving or removing an outcome
  • Outcomes defined:
    • Positive reinforcement: producing the operant response yields appetitive outcome
    • Negative reinforcement: producing the operant response results in the removal of an aversive outcome
    • Positive punishment: producing the operant response yields an aversive outcome
    • Negative punishment: producing the operant response yields removal of an appetitive outcome

Instrumental Conditioning Procedures

  • Positive reinforcement: instrumental response produces outcome; strengthens behaviour
  • Negative reinforcement: instrumental response removes a punishing outcome
  • Punishment: instrumental response produces an aversive outcome (reduces behaviour)
  • Negative punishment: instrumental response removes an appetitive outcome
  • Omitted or Differential Reinforcement of Other behaviour (DRO) as a form of negative punishment

Simple Schedules of Reinforcement

  • Fixed Ratio (FR): reinforcement after every nth response
  • Variable Ratio (VR): reinforcement after an average of n responses; variability around the mean
  • Fixed Interval (FI): reinforcement available after a fixed amount of time has elapsed; the first response after that interval is reinforced
  • Variable Interval (VI): reinforcement available after a variable/time-average interval; the time to reinforcement varies
  • Schedule effects
    • Steady, high response rates under ratio schedules with post-reinforcement pauses (ratio runs)
    • The length of the post-reinforcement pause correlates with the number of required responses
  • Continuous Reinforcement (CRF): every instance of the behaviour is rewarded; effectively FR1
    • Leads to steady, moderate response with brief pauses; relates to compulsive behaviours

Motivational Mechanisms

  • Three-term contingency: Context (S) – Instrumental response (R) – Outcome (O)
    • S-R association: driven by the Law of Effect; reinforcer stamps in the S-R link
    • S-R-O: the basic motivation to perform the instrumental response via contextual cues
  • S-O or S-S associations: learning that outcomes are expected or anticipated by cues
  • Two-process theory (Hull, Spence): instrumental responses increase because S evokes the response directly via S-R; or via S-O expectancy that motivates the response
  • Two-process theory suggests both S-R and S-O/motivational expectancy contribute to instrumental performance

Premack Principle and Response Deprivation

  • Premack Principle: higher-probability (more likely) activities can reinforce lower-probability (less likely) ones
    • If H (high probability) follows L (low probability) after an opportunity, L is reinforced
    • If H follows L, H does not reinforce L; essentially, the more probable activity can serve as a reinforcer for the less probable one
  • Applications
    • Eating, sex, etc., are not inherently special beyond their likelihood
    • Explains individual differences (e.g., videogaming as a reinforcer)
  • Response-Deprivation Hypothesis
    • Restriction of access to a normally available activity creates reinforcement for other activities
    • Instrumental procedures inherently restrict access to some reinforcer, affecting its reinforcing value

Extinction (Revisited)

  • Extinction in CC: CS no longer predicts the US; CR diminishes
  • Extinction in Instrumental Conditioning: CR in the presence of CS no longer leads to US
  • Interference effects include outcome interference, cue interference, punishment, etc.
  • Extinction is not erasure but new inhibitory learning specific to context

Recovery from Extinction

  • Extinction is not permanent; recovery of response can occur after extinction
  • Mechanisms include:
    • Spontaneous recovery: return of responding after time without training
    • Renewal: recovery when testing occurs in a different context (ABA, ABC, AAB)
    • Reinstatement: return after re-exposure to US
  • Bouton’s Theory of Retrieval (1993)
    • Extinction reflects retroactive outcome interference; second-learned X–O2 disrupts retrieval of X–O1
    • Proactive interference: test context not similar to extinction context can lead to proactive interference
    • Retrieval theory links context, time, and space to recovery

Generalisation vs Discrimination

  • Discrimination: determining whether instrumental behaviour is controlled by a specific stimulus; different aspects of the same stimulus control different responses
  • Generalisation: responding to similar stimuli; the extent of generalisation is shown by gradient steepness
    • Generalisation gradients illustrate stimulus control: steeper gradient indicates strong control; shallow gradient indicates weaker control

Evaluative Conditioning

Introduction to Evaluative Conditioning

  • Evaluative Conditioning (EC): conditioning of likes and dislikes via association with positive or negative stimuli
  • Similar to classical conditioning but the outcome is a change in attitude/valence rather than a physiological reflex
  • Examples involve preference changes (e.g., tea preference) rather than reflexive responses

Generality of Evaluative Conditioning

  • Acquisition of likes/dislikes via associative processes
    • Classical conditioning: neutral CS paired with US elicits preparatory/defensive responses (reflexes)
    • Evaluative conditioning: the valence of a CS changes due to US associations; CS becomes liked or disliked
  • CS = neutral stimulus; US = affective stimulus; DV = changes in valence or liking

Models of Evaluative Conditioning

  • Conceptual-Categorisation Account (Davey, 1994)
    • EC is not purely associative learning but concept learning
    • CS contains both likeable and unlikable elements; pairing with US highlights congruent features
    • CS recategorised based on features, not solely on affective valence
    • Problems: cross-modal conditioning and US revaluation effects
  • Holistic Account (Martin & Levey, 1970s)
    • EC is a basic form of learning shared across species
    • CS-US presentation yields a holistic representation including the evaluative nature of the US
    • CS evokes US representation, explaining resistance to extinction and contingency awareness
    • Problems: inconsistent with sensory preconditioning
  • Referential Account (Baeyens et al., 1992)
    • Classical conditioning: associative changes in preparatory responses to CS
    • Evaluative conditioning: associative changes to the valence of the CS
    • Distinguishes signal/expectancy learning (classical) from affective evaluative learning (referential)