Distributional learning exhibits hysteresis, with prior distributions influencing judgments after distributions change¶

Claim · All claims
Evidence · 1 study · causal r2 · q3
Currency · newest study 2025 · not yet checked against its sources · not reviewed by a person

Subclaims¶

q3 i? After a mid-study shift from a unimodal to an equiprobable distribution, the formerly high-probability frequency retains a detection and RT advantage ('stubborn predictions'), and formerly low-probability tones remain poorly detected despite probability increases. → Luthra 2025

Evidence¶

Luthra 2025¶

Luthra, S., Luor, A., Tierney, A. T., Dick, F., & Holt, L. L. (2025). Statistical learning dynamically shapes auditory perception. npj Science of Learning. https://doi.org/10.1038/s41539-025-00328-z

q3 · i? · causal · r2

In Exp 6b, after a mid-study shift to equiprobability, decision RTs for 1000 Hz remained as fast as when it was 80% probable (d = 0.025), and overall frequency modulation persisted (p = 8.306e-5, η² = 0.068). In Exp 5b, 1000 Hz detection declined after the shift (d = 0.496) but formerly low-probability tones showed no detection change despite a more than 3-fold probability increase (p = 1).

"Even though 1000 Hz is now 20% probable, RTs are not significantly different than in thefirst experiment half when it was 80% probable (linear contrast, p = 0.720, Cohen's d = 0.025)"

Discussion¶

Related Claims¶

• Distributional learning modulates tone detection in noise, with better detection of high-probability frequencies — related
• Perceptual sensitivity and decisions rapidly update in volatile statistical contexts — reports the opposite
• Graded proximity effects: equiprobable rare tones near a distribution mode are detected better than distant ones — related
• Distributional learning across a task-irrelevant dimension slows duration decisions for low-probability frequencies — related
• Massed exposure to a single frequency produces lasting suppression of detection at distant frequencies — related