Classical Conditioning Part 2

Review of Contingency Theory

• Mechanistic, minimalist model of Pavlovian conditioning.

  • Assumes CS–US relations are stored as either:

  • Excitatory (produces CR).

  • Inhibitory (suppresses CR).

  • “Null” (no information ➔ no learning) when experience with and without CS is perfectly balanced.

• Predicts behavior from the statistical contingency between CS and US only; makes no claims about cognitive operations (e.g., attention, memory for place, comparisons).

• Supported by early findings showing:

  • Standard excitatory conditioning (CS → US → CR).

  • Inhibitory conditioning (CS predicts absence of US).

  • Zero contingency → no net learning.

Problems Exposed in Later Work

• Animals still learn something when overall CS–US contingency is $0$ (equal trials with and without CS), contradicting the diagonal “no‐learning” prediction.

• Need for a model that:

  • Allows simultaneous acquisition of multiple associative links.

  • Acknowledges attention, context processing, and comparison operations.

Emergence of Comparator Hypothesis (Miller & Matzel $1989$)

• Core idea: Every cue in an episode (explicit CS, background context, other stimuli) can gain excitatory associative strength with the US.

• Expression of a CR depends on relative strength of the CS–US link vs. competing links (especially Context–US).

  • If $V<em>{CS} > V</em>{Context}$ ➔ CR appears.

  • If $V<em>{Context} \ge V</em>{CS}$ ➔ little or no CR, even though CS–US learning occurred.

• Introduces overtly cognitive elements:

  • Animal detects and stores information about physical setting (light, odor, sound, spatial cues).

  • Animal can direct attention to the most predictive cue.

  • Performance is a comparison (hence “comparator”) among stored associations at test.

Concept of “Context” (Illustrative Example)

• Library:

  • Quiet, bright, smells of old books.

  • Behavioral expectation: speak softly, sit still.

• Bar/Disco:

  • Dim, loud music, smell of alcohol.

  • Expect dancing, loud talking.

• Humans instantly switch expectations across contexts; comparator hypothesis claims non-human animals do likewise.

Key Assumptions vs. Contingency Theory

Experimental Evidence Supporting Comparator Hypothesis

Caspera et al. $1987$

Design Overview

• Subjects: Rats, water-deprived; CR measured via lick-suppression (time to complete $25$ licks).

• US: Foot-shock.

• CS: Tone.

• Contexts:

  • Context A: Lights on, white-noise radio, lemon scent.

  • Context B: Lights off, quiet, almond scent.

• Groups (both $n=??$; exact number not provided):

  • Same – extra unsignaled shocks delivered in same Context A.

  • Different – extra unsignaled shocks delivered in different Context B.

Phase Structure

1. Phase $1$ (Acquisition for both groups)

   • $\text{Tone} \rightarrow \text{Shock}$ pairings in Context A.

2. Phase $2$ (Unsignaled shocks)

   • Same: Random shocks in Context A.

   • Different: Identical number of random shocks in Context B.

3. Test (Context A)

   • Present Tone alone; record time to $25$ licks.

Theoretical Predictions

• Contingency Theory:

  • Total CS–US and US-alone counts match ➔ No group difference.

• Comparator Hypothesis:

  • Same: Context A gains extra associative strength in Phase $2$, potentially exceeding $V_{Tone}$ ➔ Weak CR.

  • Different: Extra strength accrues to Context B (irrelevant at test) ➔ $V<em>{Tone} > V</em>{Context\,A}$ ➔ Strong CR.

Results

• Same: Completed $25$ licks quickly (low fear).

• Different: Took markedly longer to finish $25$ licks (high fear).

• Matches comparator prediction; contradicts contingency model.

Matzel et al. $1987$

Design Overview

• Same lick-suppression apparatus.

• Groups:

  • EXT (Extinction)

  • No-EXT (No Extinction)

• All training/testing uses Context A (unless otherwise noted).

Phase Structure

1. Phase $1$ (Context pre-exposure to shock)

   • Both groups: Unsignaled shocks in Context A (Context A becomes excitatory).

2. Phase $2$ (Context manipulation)

   • EXT: Multiple sessions in Context A without shock (extinction of Context A–US link).

   • No-EXT: Remain in home cage (no change to Context A link).

3. Phase $3$ (CS Acquisition)

   • Both groups: $\text{Tone} \rightarrow \text{Shock}$ pairings in Context A.

4. Test (Context B: novel)

   • Present Tone; measure time to $25$ licks.

Theoretical Predictions

• Contingency Theory: No differences (identical CS–US counts).

• Comparator Hypothesis:

  • EXT: Context A’s associative strength reduced in Phase $2$ ➔ $V_{Tone}$ can dominate comparison ➔ Strong CR.

  • No-EXT: Context A retains high strength ➔ $V<em>{Context\,A} \ge V</em>{Tone}$ ➔ Weak CR when tested elsewhere.

Results

• EXT: Long latency to finish $25$ licks (robust freezing).

• No-EXT: Rapid completion (minimal freezing).

• Again, data support comparator account over contingency theory.

Why These Findings Matter

• Provide empirical evidence that contextual information competes with discrete CSs.

• Show that performance can mask learning: identical CS–US histories can yield different behavior depending on contextual strength.

• Imply animals possess:

  • Attention: ability to shift focus across cues.

  • Memory for place: storing multi-modal features of contexts.

  • Comparative evaluation: deciding which cue better predicts the US at the moment of testing.

• Pushes field from purely associative rules toward cognitive models of animal learning.

Ethical & Methodological Notes

• Water deprivation and foot-shock are standard but invasive; highlight necessity of:

  • Minimizing deprivation duration.

  • Using lowest effective shock intensity.

  • Institutional Animal Care and Use Committee (IACUC) approval.

• Comparator-style designs often require multiple contexts ➔ researchers must carefully equate salience (lighting, odor, sound) to isolate the variable of interest.

Connections to Earlier & Future Lectures

• Builds directly on conditioned emotional response procedures (lever pressing ➔ lick-suppression as faster metric).

• Sets stage for upcoming discussion of critical variables:

  • US factors (intensity, biological relevance).

  • CS factors (modality, salience, novelty).

  • Temporal factors (inter-stimulus interval, trace vs. delay conditioning).

• Links to broader themes in animal cognition history: transition from behaviorism to cognitive approaches.

Take-Home Summary

• Contingency theory cannot explain learning/performance dissociations when context is manipulated.

• Comparator hypothesis introduces cognitive mechanisms—attention and relative comparison—that successfully predict empirical outcomes.

• Context is not a passive backdrop; it actively competes with explicit CSs for behavioral control.

• Performance (CR) reflects relative associative strength, not absolute CS–US pairing count.

• Future research explores how stimulus, US, and temporal properties modulate these comparative processes.