Reduced summation with common features in causal judgments.

Area of Science:

• Cognitive Psychology
• Learning Science
• Behavioral Neuroscience

Background:

• Causal judgment tasks investigate how humans learn associations between events.
• Elemental theories, like the Rescorla-Wagner Model (RWM), predict summation effects when learned patterns are combined.
• Configural theories propose alternative mechanisms for pattern association and outcome prediction.

Purpose of the Study:

• To test predictions of elemental and configural theories of causal learning.
• To investigate the impact of shared features on summation effects in pattern learning.
• To evaluate the explanatory power of the Rescorla-Wagner Model and Pearce's configural model.

Main Methods:

• Human participants completed causal judgment tasks across three experiments.
• Training involved associating visual patterns with an outcome.
• Testing assessed outcome likelihood with novel pattern combinations, including those with shared features.

Main Results:

• Summation effects were observed when novel patterns combined previously learned associations.
• A common feature between training patterns abolished, rather than reduced, summation.
• Results were inconsistent with the Rescorla-Wagner Model and a configural alternative.

Conclusions:

• Existing elemental and configural models inadequately explain the observed summation failures.
• Causal learning may involve flexible strategies encompassing both elemental and configural processing.
• The Replaced Elements Model provides a better account of summation and summation failure phenomena.