Yas- CM2

Blocking

• Reminder of Blocking:

  • Initial learning phase: A+ (Stimulus A followed by a US)
  • Blocking phase: AB+ (Stimuli A and B together, followed by a US)
  • Test: Present B alone. The typical finding is that there's little to no learning about B.

• Prediction Error: Rescorla-Wagner Model

  • Explains blocking by stating that the US is already predicted by A during the AB+ trials.
  • Since there's no surprise (no prediction error), no association is formed between B and the US.

• Alternative 'Attentional' Account of Blocking

  • Suggests that attention to cues (α) changes with learning.
  • During AB+ trials, attention to A is high (because A already predicts the US), while attention to B is reduced. This reduced attention to B leads to less learning about B.

Experiment Example (Shock Experiment)

• Task: Participants are presented with colored squares, some followed by an electric shock.

• The task is to determine which colors predict the shock.

• If two colors that individually predict shock are presented together, the participant receives a double shock.

• Example trials:

  • Orange -> Shock!
  • Orange Purple -> Shock!
  • Brown White -> Shock!

• Test Phase: Participants are asked which colors will be followed by a shock if presented alone.

• Standard Blocking Result: Brown > Purple (Brown is more likely to be associated with shock than Purple).

Associative Links

• Blocking is often explained as a failure to form an associative link between Cue B and the US.

  • Orange -> Shock!
  • Purple (no link formed)

Prediction Error in Blocking

• Design: A+ then AB+
• By the end of A+ trials, A fully predicts the US. Therefore, there's no prediction error.
• During AB+ trials, the US is not surprising because A is already predicting it. Thus, B is not processed, and no B-US association is learned.
• Learning about B is 'blocked' by A.

Importance of Blocking

• Before blocking, it was believed that associations form simply by pairing stimuli (Hebbian learning).
• Blocking demonstrates that learning isn't just about pairing.
• It is central to contemporary learning theory.

Formalizing Prediction Error

• Surprise is the difference between what you predict will happen and what actually happens.

• Rescorla-Wagner Model (1972)
  ΔV = α(λ – ΣV)

  • ΔV = Learning: Change in associative strength on a given trial.
  • λ = Total associative strength supported by the outcome.
  • ΣV = The expected outcome given all cues present.
  • α = Attention to cues (salience).

• Key point: Learning happens only when there is a prediction error.

Alternatives to Rescorla-Wagner: Attention to CSs

• Signals of important events are attended to particularly well.
• The salience of the CS (α) is not fixed.

Mackintosh (1975): Blocking due to Changes in α

• A+ then AB+
• The most predictive cues increase in salience/α (e.g., cue A).
• Less predictive cues decrease in α (e.g., cue B on AB+ trials).
• Not much learning about cue B.

Measuring Attention: Eye-Tracking

• Beesley and Le Pelley (2011) used eye-tracking in a learning task.
• Participants viewed stimuli (chemicals) and outcomes (symptoms in Mr. X) on a screen.
• The study tracked whether participants looked at the word 'Addexium' (the CS).

Experimental Design (Beesley & Le Pelley, 2011)

• Stage 1: A-o1, C-o1, E-o2, G-o2, I-o1, L-o1, O-o2, R-o2
• Stage 2: AB-o1, CD-o1, EF-o2, GH-o2, JK-o1, MN-o1, PQ-o2, ST-o2
• Stage 3: BK-o3, DN-o4, FQ-o3, HT-o4
• Test: BT?, DQ?, FN?, HK?, BN?, DK?, FT?, HQ?

Eye Gaze Results in Stage 2 (AB+/JK+ trials)

• Attention to cue A is high because it's a good outcome predictor.
• Attention to