A psychometric function plots the probability of a correct (or 'yes') response against stimulus intensity. Its position tells you the threshold — how sensitive the observer is. But naively measuring it can be misleading.

The problem with Yes/No tasks:
Imagine you flash a dim light and ask 'Did you see it?' (yes or no). Two observers could have identical visual sensitivity but give very different psychometric functions — simply because one says 'yes' more liberally (lower decision criterion) than the other.

The cautious observer's curve shifts right. The liberal observer's curve shifts left. You can't tell if the difference is due to sensitivity (they actually see better) or criterion (they're just more willing to say yes).

✅ The fix: Two-Interval Forced Choice (2IFC). Present two intervals — one with the stimulus, one blank — and ask 'which interval had the light?' The observer must choose, so there's no decision criterion. Performance at zero signal is always exactly 50% by chance.

2IFC advantages:
• Performance at zero signal is locked at 50% (chance)
• Performance approaches 100% at high signal strength
• Threshold = stimulus level at some defined performance (e.g., 75% or 84% correct)
• 75% is halfway between chance (50%) and perfect (100%)
• 84% is one standard deviation above 50% on a cumulative Gaussian — commonly used in research

A lower threshold (the curve is shifted left) means the observer is genuinely MORE sensitive.

Simultaneous contrast — an example of PSE:
Two grey discs are physically the same luminance, but the one on a dark background looks brighter. The Point of Subjective Equivalence (PSE) is where two stimuli feel equal — even if they're physically different. PSE ≠ physical equality reveals a perceptual bias in the brain.

For this illusion, the brain seems to encode contrast (luminance relative to background) rather than absolute luminance. The neural correlate may be center-surround cells in the retina.