Wei Ji Ma

Drift-like dynamics and information flow across the cortical hierarchy during working memory

The paper investigates how working memory information flows across the cortical hierarchy using fMRI decoding, finding drift-like dynamics and top-down information flow from higher to lower visual areas.

Neural mechanisms of resource allocation in working memory

The paper identifies a neural mechanism for resource allocation in working memory by showing how frontal cortex feedback modulates visual cortex activity based on item priority.

Reward prediction error neurons implement an efficient code for reward

The paper shows that reward prediction error neurons encode rewards efficiently, aligning with principles from sensory neuroscience and reinforcement learning.

Normalization by orientation-tuned surround in human V1-V3

The paper shows that a tuned normalization model better explains fMRI responses to curved and straight stimuli in V1-V3 than a classical divisive normalization model.

Point-estimating observer models for latent cause detection

The paper investigates how the brain infers latent causes from spatial data by comparing human behavior to Bayesian models and proposing simpler 'point-estimating' models that better fit observed behavior.

Joint representation of working memory and uncertainty in human cortex

This study shows that the brain represents visual working memory content and uncertainty as probabilistic distributions in retinotopic and mid-dorsal stream cortical areas using fMRI data.

Uncertainty is Maintained and Used in Working Memory

The paper investigates whether people maintain and use uncertainty in working memory, finding that uncertainty is both maintained and used optimally in decision-making.

The Role of Sensory Uncertainty in Simple Contour Integration

The paper investigates how sensory uncertainty affects contour integration by manipulating retinal eccentricity and finds that humans use a heuristic rule that accounts for uncertainty but deviates slightly from Bayesian optimality.

The psychophysics of visual search with heterogeneous distractors

This paper examines visual search with heterogeneous distractors using simple stimuli and Bayesian modeling, finding that minimum target-distractor feature difference better predicts performance than mean difference or distractor variance.

Confidence reports in decision-making with multiple alternatives violate the Bayesian confidence hypothesis

This study challenges the Bayesian confidence hypothesis by showing that confidence reports in three-alternative decisions depend on the difference between the best and next-best options' posterior probabilities, not just the best option's probability.

Humans incorporate trial-to-trial working memory uncertainty into rewarded decisions

The paper shows humans use trial-to-trial working memory uncertainty in rewarded decisions, integrating it with prior beliefs and rewards.

A neural basis of probabilistic computation in visual cortex

The paper investigates how uncertainty is encoded in primate visual cortex during a visual categorization task and shows that population activity encodes likelihood functions that predict decisions better than point estimates.

Human online adaptation to changes in prior probability

The paper investigates how humans adapt to changes in prior probability during decision-making, finding an exponential averaging model with a bias towards equal priors.

Sources of suboptimality in a minimalistic explore-exploit task

The paper examines human behavior in a simplified explore-exploit task and finds deviations from optimal strategies, revealing a suboptimal switching rule and sequence-level variability.

A Diverse Range of Factors Affect the Nature of Neural Representations Underlying Short-Term Memory

This study trains recurrent neural networks on short-term memory tasks to show that sequential and persistent memory solutions emerge as a spectrum based on task and circuit conditions.

Comparing Bayesian and non-Bayesian accounts of human confidence reports

This study compares Bayesian and non-Bayesian models of human confidence reports using perceptual tasks and finds that while sensory uncertainty is considered, behavior aligns better with heuristic models.

Strategic allocation of working memory resource

The paper investigates how humans allocate visual working memory resources based on task importance, finding strategic prioritization and error minimization.

Variable precision in visual perception

The paper investigates variable precision in visual perception tasks and finds weak evidence for variable precision, primarily in experiments with variable distractors.

Limitations of proposed signatures of Bayesian confidence

The paper examines two proposed signatures of Bayesian confidence, showing they have hidden assumptions and are not necessary or sufficient for Bayesian confidence.

Humans incorporate attention-dependent uncertainty into perceptual decisions and confidence

This study shows humans adjust perceptual decisions and confidence based on attention-dependent uncertainty, using a Bayesian framework.

A low-level perceptual correlate of behavioral and clinical deficits in ADHD

This study dissociates perceptual and executive deficits in ADHD using a task that varies low-level stimulus features and measures task-irrelevant motor output, finding higher perceptual variability and TIMO in ADHD participants.

A resource-rational theory of set size effects in visual working memory

The paper proposes a normative theory explaining set size effects in visual working memory by modeling resource allocation as a trade-off between behavioral and neural costs.

Bayesian Comparison of Explicit and Implicit Causal Inference Strategies in Multisensory Heading Perception

The paper compares Bayesian and non-Bayesian causal inference strategies in multisensory heading perception using computational modeling and behavioral experiments.

Problematic usage of the Zhang and Luck mixture model

The paper identifies issues with the Zhang and Luck mixture model for analyzing delayed-estimation data and proposes five solutions to address model mismatch.