PSY 424 Notes -Classical conditioning 3 (Rescorla-Wagner Model).pdf

Slide 1/2 Blocking (Kamin's Experiment) Experimental Setup:  Stimuli Used:  CS1: Light  CS2: Noise  US: Shock Control Group:  Conditioning:  8 trials of CS1 + CS2 → US (Light and noise p...

Slide 1/2 Blocking (Kamin's Experiment) Experimental Setup:  Stimuli Used:  CS1: Light  CS2: Noise  US: Shock Control Group:  Conditioning:  8 trials of CS1 + CS2 → US (Light and noise paired with shock).  Both CS1 (light) and CS2 (noise) paired with the US from the start.  Outcome:  Conditioned Emotional Response (CER) to both CS1 and CS2 (fear/suppression to light and noise). Experimental Group:  Conditioning 1. Phase 1 (Pretraining):  16 trials of CS1 → US (Light paired with shock).  CS1 (light) fully predicts the shock (US). 2. Phase 2 (Compound Conditioning):  8 trials of CS1 + CS2 → US (Light and noise paired with shock).  CS2 (noise) is introduced in the compound with CS1 (light).  Outcome:  CER only to CS1 (light).  No CER to CS2 (noise) — Blocking occurs. Key point CS1 blocks CS2: Since CS1 already fully predicts the US (shock), the introduction of CS2 provides no new information about the US. Therefore, CS2 fails to acquire an association with the US (no CER to noise). Follow-up Manipulation: Increasing US Intensity: Increase the shock intensity (US) in the compound phase.  CS2 acquired some CER (now elicits suppression/fear).   The increase in US intensity "unblocked" CS2, allowing it to acquire an association with the US. Takeaways A. Blocking: Prior learning about CS1 (light) prevents CS2 (noise) from acquiring an association with the US. B. Prediction Error: CS2 is not learned about because it provides no new predictive value when CS1 already predicts the shock. C. CER (Conditioned Emotional Response): The emotional response (like fear) used to measure learning in the experiment. D. Unblocking: Increasing the intensity of the US can overcome blocking, allowing CS2 to acquire a CER. Slides 3/4 Rescorla-Wagner Model Key Concepts Prediction Error: Learning occurs based on the difference between what is expected (predicted) and what actually happens. Associative Strength: The strength of the connection between a conditioned stimulus (CS) and the unconditioned stimulus (US). Learning is Limited: Each CS can acquire a maximum amount of associative strength (e.g. 100) based on the US (denoted as λ). Equation ΔV=α(λ−V) ΔV: Change in associative strength (learning) for the CS in each trial. α: Rate of learning λ: Maximum associative strength the US can support (the "ceiling" of learning). V: Current associative strength (how much the CS already predicts the US). (λ - V): Prediction Error (difference between what is expected and what occurs). Explanation of the Equation: (λ - V): Determines how much learning occurs. If the US is surprising (high prediction error), there is more learning. If the CS already predicts the US well, there is less learning. α : Influence the rate of learning. More noticeable stimuli or stronger US values lead to faster learning. Compound Conditioning in the Rescorla-Wagner Model Compound Stimuli: When two or more conditioned stimuli (CS1 and CS2) are presented together and paired with the US, the total associative strength is shared between them. Modified Equation for Compound Conditioning: For two stimuli (CS1 and CS2) presented together: ΔVA=αA( λ− [VA+VB] ) ΔVB=αB( λ− [VA+VB] ) See calculation practice examples below VA: Associative strength of CS1, VB: Associative strength of CS2.  Each stimulus (CS1 and CS2) receives part of the total associative strength based on its salience (α). More salient or noticeable stimuli will acquire more associative strength.  Key Effects in Compound Conditioning:  Overshadowing: In compound conditioning, the more salient CS (e.g., a loud tone) will dominate learning, overshadowing the less salient CS (e.g., a dim light).  Outcome: The stronger CS will acquire more associative strength, and the weaker one will acquire less.  Blocking: If a CS (CS1) has already fully predicted the US, introducing a second CS (CS2) in the compound conditioning will result in little to no learning for CS2, as there is no prediction error left.  Outcome: CS1 blocks CS2 from acquiring associative strength. Summary  Learning is driven by prediction error: Larger errors lead to greater changes in associative strength.  Compound stimuli share associative strength: The total amount of learning is distributed between the stimuli based on salience.  Blocking and overshadowing: Learning can be blocked or overshadowed depending on the prior associations and the salience of the stimuli. Slides 5 - 9 Prediction Error /ACC Ensemble Study Prediction Error: Occurs when actual outcomes differ from expected ones, driving learning and behavior change. Anterior Cingulate Cortex (ACC):  Important for detecting prediction errors and guiding decisions.  Neurons in the ACC (neural ensembles) respond to unexpected outcomes. Study Findings:  Neural ensembles in ACC become active during both positive prediction errors (unexpected rewards) and negative prediction errors (missed rewards).  Different groups of neurons respond to different types of errors, helping to update future behavior.  Prediction Error as a Learning Signal: Prediction error signals in the ACC help the brain update models of the world, which is essential for reinforcement learning and decision- making. Takeaway: ACC is crucial for learning from mistakes by tracking and updating expectations when outcomes are surprising.