Accuracy-precision trade-off in visual orientation constancy.

Area of Science:

• Neuroscience
• Perception Psychology
• Computational Neuroscience

Background:

• Maintaining visual orientation constancy during lateral head tilt is crucial for spatial awareness.
• Previous studies identified opposing bias effects (A-effect and E-effect) in subjective visual vertical (SVV) performance across different tilt angles.
• The underlying mechanisms for these SVV error patterns remain incompletely understood.

Purpose of the Study:

• To investigate if a Bayesian spatial-perception model can explain the observed SVV error patterns (A- and E-effects).
• To test the hypothesis that SVV errors result from a precision-accuracy trade-off in combining sensory information about head orientation.
• To evaluate the model's ability to predict human performance in spatial orientation tasks.

Main Methods:

• Employed a psychometric approach to quantify the accuracy and precision of the subjective visual vertical (SVV).
• Tested eight human subjects under nine different lateral tilt angles, ranging from -120 to 120 degrees.
• Fitted the collected SVV data to a Bayesian spatial-perception model.

Main Results:

• Both SVV accuracy and precision significantly decreased as the magnitude of the lateral tilt angle increased.
• The observed pattern of worsening accuracy and precision with tilt angle was well-described by the predictions of the Bayesian model.
• The model successfully accounted for both undercompensation (A-effect) and overcompensation (E-effect) biases.

Conclusions:

• Spatial vision performance, particularly in tasks like the SVV, adheres to the principles of Bayes' optimal observer theory.
• Systematic errors in spatial orientation are not arbitrary but reflect an optimal computational strategy balancing precision and accuracy.
• The Bayesian model provides a robust framework for understanding visual orientation constancy and its limitations under head tilt.