Same underlying model as the 3D version, redrawn as three 2D-native panels instead of a camera-projected scene: a live gate(t) function graph (probability of correctly reporting T2 vs. milliseconds since T1), a scrolling RSVP letter strip that plays each trial's actual item sequence, and a bar chart of accuracy vs. lag.
t = lag × SOA
gate(t) = 0.95 if t < 100 ms (gate still open: lag-1 sparing)
gate(t) = 0.95 − D·exp(−(t − 300·k)² / (2·(120·k)²)) otherwise, clamped to [0.05, 0.95]
D = 0.55·k (dip depth)
k = T1-task-demand factor (0.7 low, 1.0 medium, 1.3 high)
Each trial samples a Bernoulli outcome with probability gate(t) — a Monte Carlo draw from the theoretical curve, exactly like a real subject's trial-by-trial report. The moving dot on the gate-curve graph tracks the live t since T1 during playback; the letter strip shows the actual RSVP sequence (distractors, T1 in cyan, T2 in pink) scrolling past at the chosen SOA.
- Lag slider — items between T1 and T2 (1 = immediately after T1).
- SOA slider — stream speed; converts lag into real elapsed time.
- T1 task demand — how attention-hungry the T1 judgement is.
- Run 1 Trial — plays the RSVP strip once and reports hit/miss.
- Run 50 Trials — samples the model 50× instantly to build up the bar chart.