EXTINCTION, RECOVERY AND AVOIDANCE - VOCABULARY FLASHCARDS

EXTINCTION

Reducing Responding – Everyday Examples

• Toddler crying at night ⟶ parent arrives (attention = reinforcer)
• Dog jumps up on owner ⟶ gets a treat
• Child mispronounces a word ⟶ receives praise
• Core issue: reinforcement is maintaining an undesirable operant response; goal of extinction is to eliminate that reinforcement contingency.

Basic Definitions

• Classical Conditioning (CC): Extinction = repeated presentations of the conditioned stimulus (CS) without the unconditioned stimulus (US).
• Instrumental / Operant Conditioning (IC): Extinction = performing the conditioned response (CR) in presence of the stimulus no longer yields the reinforcing outcome.
• Conceptually the opposite of acquisition; fits Rescorla-Wagner framework.

Extinction Dynamics in CC & IC

• CC example: Pavlov’s salivation – CS (bell) no longer followed by food; salivation diminishes.
• IC example: Light-on → lever-press → food dispenser blocked ✗; lever pressing eventually declines.
• Extinction burst: immediate, temporary increase in response rate when reinforcement is first removed.
• Response variability increases initially (Neuringer, Kornell & Olufs): subjects try novel topographies to regain reinforcement.
• Sustained non-reinforcement ⟶ gradual decrease in both rate and variability.
• Emotional by-products: frustration and possible aggression.

Rate of Extinction & Reinforcement Schedule

• Continuous Reinforcement (CRF) → rapid extinction once reinforcement halted.
• Partial Reinforcement → Partial-Reinforcement Extinction Effect (PREE): responding persists longer, extinction is slower.

Explanatory Theories for PREE

• Discrimination‐hypothesis:
  • Organism fails to detect the shift from partial to zero reinforcement; change from CRF → extinction is obvious, change from partial → extinction is subtle.
• Generalisation-decrement hypothesis:
  • Extinction context resembles past partially reinforced context → response generalises.
• Frustration theory:
  • Organism learns to respond through frustration; frustration itself becomes a cue that signals reinforcement.
• Key experimental comparison (Discrimination vs Generalisation):
  • Group 1 – intermittent reinforcement → extinction versus Group 2 – CRF → CRF → extinction; persistent responding in Group 1 supports discrimination difficulty.

What Happens During Extinction? Unlearning vs Forgetting vs Interference

• Unlearning: active erasure of CS–US association (Rescorla-Wagner sets \lambda = 0 during extinction, yielding negative \Delta V).
• Forgetting: passive time-based decay; occurs without explicit non-reinforcement; assessed with no-extinction “forgetting controls”.
• Interference: new inhibitory learning masks but does not erase prior excitatory learning; depends on similarity between phases.

Interference Types

• Proactive: Phase 1 learning impairs recall of Phase 2.
• Retroactive: Phase 2 learning impairs recall of Phase 1.
• Cue vs Outcome interference:
  • Cue: different cues compete for same outcome (A→B, C→B).
  • Outcome: same cue paired with different outcomes (A→B, A→C).
• Extinction = Retroactive Outcome interference: Phase 1 X+ (X→US), Phase 2 X− (X→no-US), Test X⟶CR? Retrieval disrupted by second‐learned association.
• More similarity (context, modality, timing) between the two phases → stronger interference.

RECOVERY FROM EXTINCTION

Key Message

• Extinction does not erase original learning; it installs an inhibitory S–R (or S–US) trace that is context-specific and fragile.

Forms of Recovery

1. Spontaneous Recovery
   • Response returns after mere passage of time.
   • Attributed to instability of inhibitory or second-learned associations.
2. Renewal
   • Change of physical context reinstates responding.
   • Variants:
     • ABA: Acquisition in A, extinction in B, test in A.
     • ABC: Acquisition A, extinction B, test C.
     • AAB: Acquisition & extinction in A, test in B.
3. Reinstatement
   • Non-signalled presentations of the US alone restore responding when CS is next encountered.
   • Context-specific: US must appear in, or generalise to, test context.
4. Facilitated Reacquisition
   • After extinction, re-pairings of CS–US are learned faster than initial acquisition (savings effect).

Bouton’s (1993) Retrieval Theory

• Frames extinction as retroactive interference; context (space & time) functions as a retrieval cue.
• If test context resembles extinction context ⟶ retrieve X–noUS (inhibition).
• If test context differs ⟶ proactive interference: earlier X–US resurfaces.
• Same mechanism explains spontaneous recovery (time = context change), renewal (space change), reinstatement (US presence alters contextual meaning).

Clinical Relevance

• Exposure therapy = extinction; relapse corresponds to spontaneous recovery, renewal, reinstatement.
• Research guides protocols to reduce relapse (Anderson & Insel 2006; Craske & Mystkowski 2006; Boschen, Neumann & Waters 2009).

ENHANCING EXTINCTION

Methods that Strengthen Extinction Learning

• Trial & session spacing: spaced > massed.
• Massive extinction: sheer number of CS-alone trials.
• Deepened extinction: extinguish CS in compound with a second excitor → larger prediction error \Delta V = \left(0 - [1.0 + 1.0]\right) = -2.0 ⟶ stronger inhibitory learning.

Methods that Improve Retrieval of Extinction Memory

• Multiple extinction contexts (Gunther, Denniston & Miller 1998): extinction in B,C,D produced less renewal vs single-context B.
• Increasing similarity between acquisition & extinction contexts: gradated context shifts reduce AAB renewal (Laborda et al., 2011).
• Reminder (extinction) cues during testing (Brooks & Bouton 1994) – external or internal signals present during extinction are re-presented at test to facilitate retrieval of inhibition.

PUNISHMENT

Definition & Forms

• Positive punishment: add aversive event after response (e.g. shock, scolding).
• Negative punishment: remove positive event (e.g. time-out, point loss).
• Laboratory practice: first establish high response rate with positive reinforcement, then pair response with punisher.

Key Variables Determining Effectiveness

1. Intensity
   • Higher intensity → greater suppression up to a limit.
   • Graduated (escalating) punishers < single severe punisher (habituation to mild shocks).
2. Schedule (Proportion of responses punished)
   • Punishment need not be continuous; however, lower probability ⟶ weaker suppression.
   • On FR reinforcement schedules, punishment lengthens post-reinforcement pause but little effect on ratio run.
3. Contingency (response–punisher correlation)
   • Greater contingency → stronger suppression (Goodall: instrumental‐contingent vs yoked CS).
4. Contiguity (delay)
   • Longer R–punisher delays dramatically reduce effectiveness (“Just wait till your father gets home…”).
5. Competing Positive Reinforcement
   • Behaviour may persist if still reinforced (drug self-administration models: cocaine or sucrose maintained lever-pressing despite shock).
6. Punisher as Discriminative Stimulus for Reinforcer
   • If reinforcement occurs only in presence of punishment, response increases (Holz & Azrin pigeons: pecking higher during punishment because food delivered only then).
7. Availability of Alternative Reinforcement
   • Providing other ways to obtain positive reinforcement (Thompson et al.) enhances suppression of the punished behaviour.
8. Presence of explicit discriminative stimulus (S+)
   • Instrumental response may return when S+ absent; punishment effects are stimulus-specific.

Theoretical Accounts

• Conditioned Emotional Response (CER) theory (Estes):
  • Stimuli preceding response become CS; punishment = US; fear (CER) interferes with behaviour.
• Negative Law of Effect (Thorndike extension):
  • Punishment weakens S–R association; empirical work shows punishment can be ≈3× more potent than reinforcement.

Side-Effects & Ethical Considerations

• Aggression, escape/avoidance of punishing agent.
• Behaviour may only suppress in presence of punisher.
• Moral, welfare, legal constraints; reinforcement preferred where possible.

AVOIDANCE

Distinction from Punishment

• Punishment: positive contingency between R and shock → suppresses behaviour; “passive avoidance.”
• Avoidance: negative contingency – performing R prevents shock; increases behaviour; “active avoidance.”

Discriminated (Signaled) Shuttle Avoidance

• Trial initiated by warning CS.
  • If animal responds before US → avoidance trial (no shock delivered).
  • If fails → escape trial (shock occurs until R emitted).
• Variants: two-way shuttle, one-way shuttle.
• Free-operant avoidance: no explicit CS; shocks at regular intervals unless response emitted.

Extinguishing Avoidance

• Response blocking / flooding: prevent avoidance response, forcing exposure to CS without US; extinction of avoidance.

Motivation Across Training

• Early: escape from actual shocks creates fear ➜ motivates responding.
• Later: fear decreases, but responding persists (fear & avoidance dissociate).
• Two-process theory: (1) Classical conditioning of fear to CS/context; (2) Instrumental negative reinforcement – response terminates fear CS.
• Safety-Signal Hypothesis: Feedback cues following avoidance act as conditioned inhibitors of fear and become secondary reinforcers sustaining behaviour (proprioceptive, spatial, tactile cues).

OVERALL SUMMARY

Extinction

• Negative CS–US contingency ⟶ response decline.
• Not unlearning; subject to spontaneous recovery, renewal, reinstatement.
• Inhibitory associations are context-specific; extinction strengthened via spacing, massed trials, deepened, multi-context procedures, reminder cues.

Punishment

• Positive response–US contingency; suppression depends on intensity, timing, contingency, concurrent reinforcement, alternative reinforcers.
• Explained by CER theory & negative law of effect; ethical & side-effect concerns.

Avoidance

• Active response to prevent US following CS.
• Initially driven by fear (escape); maintained by removal of fear or presence of safety signals.
• Extinguished through response blocking/flooding.

Mathematical / Formal Points Included

• Rescorla-Wagner error term: \Delta V = \alpha\beta ( \lambda - \sum V )
• Extinction sets \lambda = 0 → negative \Delta V (inhibitory learning).
• Deepened extinction predictive error example: (0 - [1.0 + 1.0]) = -2.0
• Outcome interference phase diagram: Phase 1 X–US, Phase 2 X–noUS.

Ethical & Practical Implications Spelled Out

• Clinical exposure therapy mirrors extinction; relapse resembles recovery phenomena.
• Punishment protocols require balancing efficacy and welfare; alternative reinforcement enhances outcomes.
• Flooding techniques derived from response blocking in avoidance research.