FL

Chapter 6 Lecture notes

Overview of Classical Conditioning Theories

  • Discussion involves standard non-programmable calculators for simple math that can be done by hand.

Neural Mechanisms of Learning

  • Focus on brain mechanism underlying learning patterns in Pavlovian conditioning.

Competing Theories

  • SS Learning (Stimulus-Stimulus Learning):

    • Belief that conditioned stimulus (CS) takes on properties of the unconditioned stimulus (US).

    • Same brain areas activated by US as by CS.

    • Example: Pigeons eating the light predicting food; rats responding to predictors of food.

  • SR Learning (Stimulus-Response Learning):

    • States that once the CS predicts the US, the US loses its importance.

    • CS alone produces conditioned response (CR) independent of the US.

    • Example: Rats exhibiting behavior toward a stranger rat based on prediction of food, treating it kindly despite no prior interaction.

    • Conditional responses depend on the identity of CS rather than just the food.

Evidence for Theories

SS Learning Evidence

  • Sign tracking in animals: behavior toward predictors of food (light) indicates the CS evokes a memory of US.

SR Learning Evidence

  • Studies showing behavior dependent on CS (e.g., kindness towards a strange rat).

  • Significant variances in response to CS based on prior experiences.

Experiment Proposal

  • Two hungry rat groups tested against each other with a predictive CS (light) for food.

Phase 1

  • Presentation of the CS (light) that predicts the US (food) for both groups.

Phase 2

  • One group (Group 1) has the US (food) devalued (overfed); the other group (Group 2) remains food-deprived.

  • Hypothesis:

    • If SR learning is correct, both groups should respond similarly to the light since US is not central to their response.

    • If SS learning is correct, responses should differ, as Group 1 should show a reduced response to the CS since they do not want the US anymore.

Test Phase

  • Observing the rats' CR to the light without the US in the test phase.

  • Expected Results:

    • Group 1 shows diminished response if SS learning holds (US mattering leads to different CR).

    • Group 2 shows greater response if SR learning holds (indicating an automatic response).

Conclusions on Results

  • Different responses from Group 1 indicate the US matters for their conditional response suggesting SS learning.

  • Additionally, both theories have relevance and utility for classical conditioning explanations.

Classical Conditioning Models

Overview of Multiple Models

  • No model is comprehensive; various models aim to explain different features of associative learning.

US Modulation vs. CS Modulation Approaches

  • US modulation: Focus on changes to the US's predictability.

  • CS modulation: Focus on changes in response to the predictors.

Rescorla-Wagner Model

Key Features

  • A mathematical model that calculates associative strength based on surprisingness.

    • Learning occurs when there's predictability, and this predictability is measured by a difference in predictability (b) and outcome (lambda).

    • Surprise is high in the beginning.

Assumptions of Rescorla-Wagner Model

  1. Learning is possible: Organisms can learn given pairing of CS and US.

  2. Curvilinear learning: Learning occurs in a non-linear manner as experiences accumulate.

  3. Surprise matters: Two kinds: positive (unexpected reward) and negative surprises (less than expected outcome).

  4. Totality of stimuli for prediction: All available stimuli influence predictions.

Key Variables

  • Associative strength (b) represents predictability.

  • US is represented by the Greek letter lambda (λ), typically equating to expected outcomes (value of 1).

  • A decrease in surprise correlates with an increase in learning. Delta (Δ) signifies change in associative strength across trials.

Learning Mechanism

Calculating Learning from Trials

  • Utilizing the equation Δb = k(λ - b)

  • Where k is the influence of parameters, such as species and type of conditioning.

Example Calculation

  • For the first trial:

    • $ ext{Delta } b = k imes ( ext{lambda} - ext{b})$ where b for the first trial is zero.

    • If λ = 1, then $ ext{Delta } b = k imes (1 - 0)$ yielding maximum surprisingness.

Importance of Surprise

  • First trial yields the most learning opportunity, inferred loss of surprise over successive trials.

Application of Model

  • Predicts real-world learning and how changes incrementally modify responses over trials.

Blocking Phenomenon

  • Definition: Existing learning structures/cues block new learning from occurring.

  • Example with conditioned tone predicting food is brought into context with a new light stimulus.

  • Adding a new light does not lead to new learning if existing cues fully predict the outcome; it remains unnoticed as a contributor in predicting the US.

  • Unblocking: To teach the new CS effectively, the US needs to increase in quantity or quality, enhancing surprise.

Conclusion on Theories

  • Both SS and SR learning concepts assist in understanding variations in classical conditioning.

    • Research continues to confirm and refine these models, none alone defining the entirety of classical learning mechanisms.