Perception as evidence accumulation and Bayesian inference: insights from masked priming

Area of Science:

• Cognitive Psychology
• Computational Neuroscience
• Psycholinguistics

Background:

• Perception is often modeled as Bayesian inference, integrating noisy sensory evidence.
• Masked priming paradigms in visual word recognition present a challenge for understanding evidence accumulation.
• Existing models struggle to explain the flexibility of masked priming effects.

Purpose of the Study:

• To formalize how evidence from primes and targets is combined within a Bayesian framework.
• To test a specific algorithm for evidence combination against visual word recognition data.
• To investigate the impact of task demands on masked priming patterns.

Main Methods:

• Comparing two algorithms for evidence combination in masked priming.
• Integrating the successful algorithm into the Bayesian Reader model.
• Conducting three experiments manipulating tasks (lexical decision, same-different judgment) in masked priming.

Main Results:

• Only one of the two algorithms for evidence combination aligned with existing literature.
• The Bayesian Reader model, incorporating this algorithm, accurately predicted experimental outcomes.
• Masked priming patterns were shown to be task-dependent, not fixed.
• Task switching from lexical decision to same-different judgment significantly altered priming effects.

Conclusions:

• The Bayesian framework provides a robust explanation for masked priming phenomena.
• Masked priming is not solely determined by prime-target relations but is modulated by task context.
• This research has implications for understanding unconscious cognition and visual masking processes.