PEST reduces bias in forced choice psychophysics

Area of Science:

• Psychophysics
• Cognitive Psychology
• Human Perception

Background:

• Observer performance in detection tasks is often assessed using psychophysical procedures.
• Traditional fixed-level methods (method of constant stimuli) may be susceptible to nonstationary performance and memory effects.
• Adaptive procedures like PEST aim to efficiently estimate performance levels.

Purpose of the Study:

• To compare the PEST adaptive psychophysical procedure with fixed-level methods in detection tasks.
• To investigate observer performance stability and accuracy across different psychophysical procedures.
• To explore the impact of sequential dependencies on performance estimation.

Main Methods:

• Utilized the PEST (Parameter Estimation by Sequential Testing) adaptive procedure and the method of constant stimuli (fixed-level).
• Conducted detection tasks with varying target probabilities of correct response (P(C)).
• Analyzed sequential dependencies in observer responses during fixed-level runs.

Main Results:

• Fixed-level runs following PEST yielded lower P(C) (approx. 0.75) than PEST targets (0.80).
• Observer performance showed sequential dependencies, with higher accuracy after correct responses.
• PEST procedures demonstrated greater stability across observers and shallower psychometric function slopes compared to fixed-level methods.

Conclusions:

• PEST may enhance performance by aiding observer focus on the detection task, mitigating memory failure issues inherent in fixed-level methods.
• Observer capability estimates from PEST are more reliable and closer to optimal performance levels.
• While generally superior, some PEST variants might introduce biases similar to fixed-level procedures.