Posterior Probability Matching and Human Perceptual Decision Making.

Area of Science:

• Cognitive psychology
• Perceptual science
• Decision making

Background:

• Probability matching is a decision-making theory used in cognitive models.
• Posterior probability matching (PPM) is increasingly applied to perceptual models.
• Consistency of PPM with signal detection theory (SDT) in vision and hearing is unclear.

Purpose of the Study:

• To test if posterior probability matching models align with established signal detection theory literature.
• To evaluate PPM models' consistency and efficiency in perceptual tasks.
• To compare PPM model performance with human observers and ideal observers.

Main Methods:

• Utilized a two-pass experiment to assess response consistency across repeated identical trials.
• Compared PPM model performance against a theoretical ideal observer in a discrimination task.
• Applied tools from signal detection theory to analyze PPM models.

Main Results:

• PPM models exhibit high response inconsistency at low performance levels.
• Practiced human observers demonstrate greater trial-to-trial consistency than PPM models predict.
• PPM models are inefficient at low-to-moderate performance levels, with human observers often exceeding PPM capabilities.

Conclusions:

• Findings support classic signal detection models and challenge broad classes of PPM models for expert perceptual performance.
• PPM models are ruled out for explaining expert performance in tasks like contrast discrimination and symmetry detection.
• The study suggests inexperienced observers might exhibit PPM behavior and offers methods to test this.