In 1974, Lawrence Weiskrantz asked a patient named D.B. to point toward a light he said he couldn't see. D.B. objected — "I'm just guessing" — then pointed correctly, consistently, at rates chance can't explain. Weiskrantz named the phenomenon: blindsight.
The standard interpretation: V1 is destroyed, the cortical visual system is offline, but a subcortical route (retina → superior colliculus → pulvinar) continues processing location and motion. The behavior is driven by that processing. There is no experience. The darkness D.B. reported was accurate.
Ian Phillips' 2021 challenge: maybe blindsight is not unconscious processing. Maybe it's qualitatively degraded but still conscious vision — dim, low-resolution, uncertain — below the threshold D.B. uses to decide whether something counts as seeing. Change the task design; give more permission to hedge; and even Type 1 patients say "maybe something's there." The signal is real. The silence is conservative reporting, not absence.
Signal detection theory provides the mathematical frame. Sensitivity (d′) measures whether a signal can be distinguished from noise at all. Criterion (c) measures how much evidence the observer requires before saying yes. A high criterion means the observer says "I see nothing" even when real signal is present — if it doesn't feel like enough to count.
This simulation shows the gap between what forced-choice can reveal and what self-report will claim, and lets you see where blindsight might live.
The yellow region is where the signal falls — above the detection threshold the forced-choice task exploits, but below the criterion the subject uses to say "I see something." This is where blindsight lives if Phillips is right.
If D.B.'s d′ was ~1.2 and his criterion was ~2.0+, his forced-choice accuracy would be above 70% while his self-report remained near zero. The behavior and the report would both be honest.
What the simulation can't show: whether there is any phenomenal experience in the yellow region, or none. That question doesn't have an answer reachable by behavioral measurement.