Evaluative Conditioning (Week 9)

Part 1 – Introduction to Evaluative Conditioning

• Evaluative Conditioning (EC)

  • Process of conditioning likes/dislikes toward an otherwise neutral conditioned stimulus (CS) by pairing it with a positive or negative unconditioned stimulus (US).
  • Responsible for many personal preferences (e.g., food, music).
  • Parallel to classical (Pavlovian) conditioning but:
  • Classical conditioning → pairs CS with biologically important US to elicit reflexive conditioned response (CR).
  • Evaluative conditioning → pairs CS with affectively valenced US to change attitude/valence toward CS rather than a reflex.
  • Formal contingencies:
  • \text{CS} + \text{Positive\ US} \rightarrow \text{CS\ Liked}
  • \text{CS} + \text{Negative\ US} \rightarrow \text{CS\ Disliked}

• Illustrative questions & everyday examples

  • “Which tea do you prefer?” shows attitude formation.
  • Red Bull can design: pairing can heighten positive evaluation of product.
  • Facebook post with humorous meme: pairing upbeat text with brand can transfer positive affect.

• Advertising & Politics

  • Celebrity endorsements, restaurant décor, political stage-craft all exploit association for attitude shaping.
  • “Mere exposure in advertising”: repeated non-valenced exposures alone can also raise liking, but EC leverages valenced co-occurrences for stronger effects.
  • Negative example: “Kill the messenger” – dislike for bearer of bad news.
  • Actors who portray villains may be personally disliked (“YOU WANTED HER DEAD MORE THAN VOLDEMORT”).

Picture-Picture Paradigm (Levey & Martin, 1975)

• Methodology
  • Phase 1: Rate 50 paintings as liked, disliked, or neutral.
  • Select two most liked + two most disliked images as USs; several neutrals become CSs.
  • Phase 2: Present CS–US pairs 20 × each (e.g., Liked US ↔ Neutral CS, Disliked US ↔ Neutral CS, plus Neutral–Neutral control).
  • Phase 3: Re-rate all pictures.
• Key finding: Previously neutral CSs
  • Paired with liked US → significant increased liking.
  • Paired with disliked US → significant decreased liking.

Part 2 – Generality of Evaluative Conditioning

Early verbal demonstrations

• Razran (1954) – “Conditioned evocation of attitudes.”
• Staats & Staats (1957)
  • Nonsense syllables (e.g., QUG, YOF, XEH, LAJ, WUH, GIW) used as CSs.
  • Each CS paired with positive or negative English words (USs), e.g., \text{YOF} \rightarrow \text{"beauty, success, money"} versus \text{XEH} \rightarrow \text{"worthless, sick, thief"}.
  • Post-pairing, syllables assume corresponding valence.

Visual domain

• Replication: Levey & Martin paradigm above; robust EC with pictures.

Gustatory domain

• Zellner et al. (1983) – Flavoured teas
  • CS+: Tea 1 + sugar; CS-: Tea 2 alone.
  • Post-test: Greater preference for CS+.
• Baeyens et al. (1990) – Fruit flavours
  • Positive condition: flavour + sugar vs. flavour alone.
  • Negative condition: flavour + Tween (bitter surfactant) vs. flavour alone.
  • Positive shift less reliable because (1) intrinsic pleasant tastes less extreme, (2) evolutionary negative-learning bias.

Cross-modal domain

• Todrank et al. (1995)
  • Phase 1: Rate odours & faces separately.
  • Phase 2: Present random face-odour pairs.
  • Phase 3: Faces re-rated – valence shifts toward that of paired odour.

Biologically significant USs

• Zanna et al. (1970)
  • CS+ word paired with electric shock; CS- word alone.
  • Later, CS+ words evaluated more negatively even though shock is primary US.
• Affective Priming Task (general method)
  • Prime stimulus presented milliseconds before target CS.
  • Response latency in categorising valence of target is faster when prime and target share valence—shows implicit EC even with strong biological USs.

Part 3 – Functional Characteristics of EC

Extinction (De Houwer et al., 2000)

• Condition group: 8\times \text{CS–US} pairings → immediate test.
• Extinction group: 8\times \text{CS–US} then 5\times \text{CS alone}.
• Finding: EC effects persist; attitude change is resistant to extinction compared with classical CRs.

Statistical contingency (Baeyens et al., 1993)

• Perfect: Only \text{CS} \rightarrow \text{US}.
• Partial: CS sometimes alone (equal CS-only trials).
• Composite: CS-only & US-only trials added.
• Result: EC occurs even with degraded contingency; mere co-occurrence suffices, unlike predictive learning.

Contingency vs. demand awareness (Field, 2000; Baeyens et al., 1990)

• Demand awareness: Participant guesses experimenter’s aim.
• Contingency awareness: Participant can explicitly report CS–US pairings.
• Findings:
  • EC can arise without demand awareness.
  • Size of EC effect not strongly linked to contingency awareness (e.g., 77 % vs. 18 % CS–US memory groups showed comparable shifts).

Counterconditioning (Baeyens et al., 1989)

• Baseline ratings → Acquisition (10× CS–US) →
  • Counter group: CS paired with opposite-valence US.
  • Control: No further pairings.
  • Extinction: CS alone.
• Counterconditioning can reverse valence, but initial learning shows persistence.

Post-acquisition US revaluation (Baeyens et al., 1992)

• After CS–US pairing, US itself is re-rated (inflated or deflated) without presenting CS.
• CS valence shifts correspondingly → supports idea that CS taps stored representation of US value rather than simple stimulus–response chaining.

Part 4 – Theoretical Models of Evaluative Conditioning

1. Conceptual-Categorisation Account (Davey, 1994)

• EC = concept learning.
• CS contains many features; pairing highlights US-congruent features, causing recategorisation.
• Problems:
  • Cross-modal EC (odours→faces) hard to explain via feature overlap.
  • Fails to predict US revaluation effects (shouldn’t matter once CS recategorised).

2. Holistic Account (Martin & Levey, 1970s)

• EC = basic primitive mechanism; classical conditioning requires higher cognition.

• CS–US presented together → single holistic memory trace including affect.

• CS later re-activates full trace → valence retrieval.

• Explains: extinction-resistance, contingency-awareness independence, US revaluation.

• Weakness: cannot deal with sensory preconditioning.

  • Sensory preconditioning demonstration (Hammerl & Grabitz, 1996):
  • Phase 1: \text{CS}1\text{ – }\text{CS}2 association.
  • Phase 2: \text{CS}_2 – \text{US} pairing.
  • Test: \text{CS}_1 acquires valence despite never co-occurring with US.

3. Referential Account (Baeyens et al., 1992)

• Distinguishes signal learning (classical) vs. referential learning (EC).
• Classical: Learner encodes expectancy that US will occur → drives preparatory CR and needs awareness.
• Evaluative: Learner encodes that CS “refers to” valenced US; does not need predictive contingency or awareness.
• Provides conceptual split:
  • Classical conditioning: \text{CS} \Rightarrow \text{Expect US}.
  • EC: \text{CS} \Rightarrow \text{Refers to good/bad}.

Comparative summary of models

• All capture some empirical regularities (resistance to extinction, lack of contingency awareness requirement, US revaluation, etc.) but none integrate boundary conditions fully.
• Still largely descriptive; neural mechanisms, evolutionary rationales remain active research areas.

Practical, Ethical & Real-World Relevance

• Marketing & consumer manipulation: strategic pairing can mold preferences subconsciously.
• Political campaigns: flags, music, or popular figures adjacent to candidate imagery.
• Clinical possibilities: counterconditioning may help reduce phobias or prejudices.
• Ethical caution: covert affective manipulation can violate informed-consent norms.

Numerical / Methodological Quick Reference

• Repetition typical range: 10\text{ – }20 pairings sufficient for robust EC.
• Liking scales: -100 \text{ (max dislike)} \rightarrow +100 \text{ (max like)}.
• Affective-priming RT facilitation often \sim20\text{–}50\,\text{ms} when prime & target share valence.
• Contingency-awareness tasks: forced-choice memory accuracy compared to 50\% chance.

Concise Take-Home Points

• Preferences can be conditioned through mere co-occurrence with valenced stimuli; predictive contingency and voluntary awareness are not required.
• EC effects are generally durable, cross-modal, and observed with both secondary (pictures, words) and primary (taste, shock) USs.
• Extinction and contingency manipulations weaken but rarely abolish learned valence.
• Competing theoretical accounts differ on whether EC is associative (referential or holistic) or conceptual; sensory-preconditioning evidence challenges a purely holistic view.
• Understanding EC is vital for fields from advertising ethics to clinical psychology.