Once you've chosen a task design (2IFC), you still need to decide how to choose stimulus intensities on each trial. Two classic methods: Method of Constant Stimuli (thorough but slow) and Staircases (fast but fragile).

Method of Constant Stimuli:
• Pre-select a fixed set of stimulus intensities
• Randomly present each many times (~20+ repetitions per level)
• Plot proportion correct at each level → fit a smooth curve (cumulative Gaussian)
• Read threshold off the fitted curve

✅ Very reliable — hard to go wrong
❌ Inefficient — needs ~300 trials; many trials at uninformative levels (too easy or too hard)

Staircase methods:
• Start with an easy (high intensity) stimulus → subject gets it right
• Make it harder (lower intensity) until they make a mistake
• Then make it easier → harder → easier, 'walking up and down' around the threshold
• Step size decreases over time, homing in on threshold

✅ Fast — only ~30 trials needed
❌ A few early errors (button press mistakes) can derail the whole staircase

Best practice: average multiple interleaved staircases to get a stable estimate.

JND vs Threshold:
The detection threshold is really just a JND from zero — the smallest stimulus you can detect at all. For discrimination tasks (which weight is heavier? which light is brighter?), the JND is measured relative to a reference stimulus.

PSE vs JND in discrimination:
• PSE = stimulus level where subject is at 50% (what feels equal)
• JND = how far you need to go from PSE to reach threshold performance (75% or 84%)
• PSE tells you about bias; JND tells you about sensitivity

🎯 The big picture: neuroscience aims to relate all of human experience to neural activity. But you need precise measurement of both sides. Neural recordings give you the biology. Psychophysics gives you the perception. Without rigorous psychophysics, neural data has no behavioral anchor.