Reduced prediction updating shapes serial dependence in autistic traits

Area of Science:

• Cognitive Neuroscience
• Perception
• Autism Research

Background:

• Serial dependence, the impact of prior experience on current perception, is typically studied in static settings.
• This study explores serial dependence as feedback-based internal model updating, examining its variation with autistic traits.
• An immersive virtual reality penalty-kick task was used to investigate prediction errors and updating mechanisms.

Purpose of the Study:

• To investigate if serial dependence reflects passive carryover or active feedback-based updating of internal models.
• To determine how autistic traits influence the process of updating internal models based on prediction errors.
• To differentiate between general perceptual biases and specific impairments in prediction updating related to autistic traits.

Main Methods:

• Participants performed a virtual reality penalty-kick task where the ball disappeared mid-flight.
• Trial-by-trial prediction errors were introduced by laterally displacing the ball upon reappearance.
• Autistic traits and football expertise were assessed to correlate with prediction updating strategies.

Main Results:

• Individuals with higher autistic traits exhibited larger prediction deviations and more rigid reliance on priors, indicating less flexible updating.
• Unlike neurotypical individuals, those with higher autistic traits did not reduce reliance on priors during maximal distortions.
• Classical stimulus and response history biases were unaffected by autistic traits, suggesting a specific impairment in prediction updating.

Conclusions:

• Serial dependence in dynamic tasks involves flexible or rigid deployment of predictions based on feedback.
• Elevated autistic traits are associated with a specific rigidity in prediction weighting, not a general perceptual bias.
• Domain experts demonstrate contrasting stabilization strategies, utilizing robust mappings resistant to perturbations.