Human change detection follows a statistical-integration model as estimation precision improves with observation

Evidence

The dependence of performance on change time suggests a dynamical integration mechanism to play a role in the decision process. While the dependence on change time indicates an ongoing estimation of the initial statistics, it does not provide much insight into the estimation of the post-change statistics. We hypothesized that reaction times here could be informative. Reaction times had previously been demonstrated to depend on task difficulty (Kiani et al. 2014 ), and difficulty in the present paradigm correlates inversely with the availability of a greater difference in change size, i.e. statistics (Fig. 2a ). — Dijk, Pim van, Physiology, Psychoacoustics and…, 3333 (#expands, via linked claim)

Categorization

• XREF: connects to: Auditory change detection integrates statistical evidence accumulated over time — Dijk, Pim van, Physiology, Psychoacoustics and…, 3285; Statistically-driven change detection may improve when attention is withdrawn.; Detection of statistical change requires temporal integration of the marginal distribution (+5 more)

Related

• Detection of statistical change requires temporal integration of the marginal distribution (similarity 0.96)
• Change-detection sensitivity rises with observation time but stays low for small changes. (similarity 0.94)
• Auditory change detection integrates statistical evidence accumulated over time (similarity 0.94)
• Reaction times reveal how listeners estimate post-change statistics in integration tasks. (similarity 0.90)
• Temporal integration may sample informative snapshots rather than integrating all energy (similarity 0.89)
• Larger change sizes steepen and left-shift hit-rate curves, shortening detection time constants. (similarity 0.89)
• Listeners detect statistical changes in ongoing sound based on time available for evidence integration (similarity 0.88)
• Time available for integrating evidence is the main predictor of change detection success (similarity 0.87)
• Longer listening time to initial texture statistics improves change detection up to an asymptote (similarity 0.87)
• Detecting statistical changes in sound improves with listening time and change size (similarity 0.85)
• Longer pre-change sampling durations speed auditory change detection. (similarity 0.85)
• Larger statistical changes are detected sooner and with less listening time. (similarity 0.82)
• Change detection in stochastic textures depends on statistical prediction rather than spectrotemporal predictability (similarity 0.82)
• Larger statistical changes produce faster and more certain auditory detection (similarity 0.82)
• Change detection in statistical sounds relies on predicting statistics rather than exact occurrences (similarity 0.82)

Broader Connections

• Detection of statistical change requires temporal integration of the marginal distribution #expands (0.85)
• Longer pre-change observation improves initial-statistic estimation and speeds detection. #expands (0.85)
• Change-detection sensitivity rises with observation time but stays low for small changes. #expands (0.80)
• Longer listening time to initial texture statistics improves change detection up to an asymptote #corroborates (0.80)
• Reaction times reveal how listeners estimate post-change statistics in integration tasks. #expands (0.80) — Dijk, Pim van, Physiology, Psychoacoustics and…, 3333
• Statistically-driven change detection may improve when attention is withdrawn. #corroborates (0.78)
• Auditory texture change detection may paradoxically improve when attention is withdrawn #corroborates (0.78)