Flexible categorization in perceptual decision making.

Area of Science:

• Cognitive Neuroscience
• Computational Neuroscience
• Decision Science

Background:

• Perceptual decisions involve accumulating sensory evidence.
• Drift diffusion models and attractor network models are common frameworks for studying this process.
• The qualitative equivalence of these models remains an open question.

Purpose of the Study:

• To compare the dynamics of drift diffusion models and attractor network models in perceptual decision-making.
• To investigate the impact of stimulus fluctuations and duration on decision state transitions.
• To identify unique predictions of attractor models validated by psychophysical data.

Main Methods:

• Simulations of attractor network models and drift diffusion models.
• Analysis of decision state transitions under varying stimulus conditions.
• Psychophysical experiments to validate model predictions.

Main Results:

• Attractor models, unlike drift diffusion models, show increased state transitions with stimulus fluctuations/duration.
• A crossover from primacy to recency weighting of evidence was observed.
• A flexible categorization regime with decision reversal asymmetry and non-monotonic psychometric curves was identified in attractor models.

Conclusions:

• Attractor models offer a richer explanation for perceptual decision dynamics, including evidence weighting flexibility.
• Decision reversals and their correction are key features of perceptual decision-making.
• Findings highlight the distinct predictive power of attractor network dynamics compared to drift diffusion models.