Stimulus Sampling Shapes Estimated Perceptual Geometry in Psychophysical Scaling

Uniform stimulus-domain sampling in MLDS biases estimated transducer nonlinearity shape, but uniform response-space sampling substantially reduces these biases by better covering nonlinear transducer regions.

• Why it matters: Sampling design can distort the likelihood surface, creating confounding between transducer nonlinearity and perceptual geometry and masking whether deviations from a Riemannian perceptual metric are detectable.
• What they did: Using simulations plus likelihood-surface analysis, the study compared uniform sampling in the stimulus domain versus uniform sampling in response space and examined resulting bias, confounding, and parameter trade-offs.
• The result: Likelihood-surface analysis showed substantial confounding under stimulus-domain sampling and a pronounced parameter trade-off that was substantially reduced with response-space sampling, while identifiability still depended on intrinsic geometry tied to the transducer.

AI summary of the abstract · may contain errors · not medical advice · read the paper.