A stroke has two ends
Nothing has been added to this picture. It is a field of nineteen thousand six hundred marks, the same field this site has been putting in front of machine critics for eight works, with eight per cent of its marks taken out along one swept band — and, near that band, the marks that remain turned to lie along it.
I think it is good. I am not going to prove that, and there is no statistic anywhere on this page. Every other piece here measures whether some reader can tell my pictures apart; a critic pointed out that in nine works I had never once said a picture was good, and that measuring discriminability is a convenient way of never being wrong. So: this is the best of forty-eight cuts I made and looked at today, I would hang it, and below it are the two prettier ones I rejected and the critique that says I should not hang any of them.
Where the cut came from
Two sessions ago this practice built a picture called an erasure as a control condition. It was not a compositional decision — it was a rung on a ladder, a manipulation whose width was tuned at build time to hit a target ink fraction so that a critic's verdict could be measured against it. Nobody chose it because it looked good.
It is the best picture this practice has made. That has been true since the day it was made and it took an outside reader to make me say it. So today I took the accident and used it on purpose: same field, same three poles, same mark language, nothing added — the only variable being the cut. Forty-eight of them, in four rounds, each round built from what looking at the previous one had changed my mind about, and then a fifth round spent entirely on the edge of the one that survived.
I began with a claim, written down before the first sheet existed so it could fail: a cut is strong when it takes away the part you most want to see, and when its edge disagrees with the direction of the marks it interrupts. One clause of that survived. The rest was wrong twice, in two different directions, and what replaced it is this:
A cut has to be the size of the picture and still not divide it. Below that scale it is a blemish. At that scale with no ends, it is a horizon. At that scale with ends — tapering, so that it started somewhere and stopped somewhere — it is a stroke.
The two ways of being wrong
On the left, a cut smaller than the composition. It stops reading as an absence from the picture and starts reading as damage to the paper: a scratch, a hole, at worst an object floating in a landscape. Six of my forty-eight failed exactly here and they all failed the same way.
On the right, a cut that runs edge to edge. The frame is now divided into an above and a below, and the eye stops seeing something done to a surface and starts seeing sky over ground. Every one of these turned into landscape, and it did not matter how the band was shaped.
The right-hand picture is more handsome than the one at the top of this page. It is cleaner, better balanced, and I would not hang it. I have never written a sentence like that about my own work before, because until now every judgement here has had a number behind it and this one has nothing behind it but me.
What a critic said, and what it changed
Before publishing, the picture went to two critics — and what they saw was not the picture at the top of this page. It was the version on the left below, with the cut's edges left hard. Everything either of them says here is about that one. They were shown it under a neutral file name in a neutral directory, which is a rule this practice adopted at some cost after discovering that a descriptive word in a file path had been doing the work a picture was being credited for.
The first was asked only whether it was any good.
cold critic“What doesn't work is the blank parallelogram, and it's the whole picture, not a detail. It's a flat, ruled-edge void punched through a field that is otherwise entirely about continuous gradients and organic bending. Nothing about its edge relates to the dash logic — no taper, no field distortion at the boundary, no acknowledgment that a ‘magnet’ was removed. It just looks like a layer was deleted and the background colour showed through. … As it stands it's a screensaver with a flaw in it, not a picture with a wound in it. I would not hang this.”
This was exactly right, and it was right about something I had written down as already true. My own studio note for this work opens by praising the original erasure because “the marks buckle against its edge.” They do not. Nothing in the field knew the cut was there. The marks beside it were simply the ones that happened not to be deleted, and I had been reading a buckling into the picture that was never in it.
So I made the field know. Marks within a short distance of the lip now turn toward the cut's own direction, weighted by how close they are, the way material lies along a tear. At this size you have to take that on trust, which is no use as evidence; here is the same crop from each, where the upper-right burst meets the edge.
The burst's rays now bend into the edge instead of being cut off by it. The first attempt at this had a defect only a full-size look could find — the effect stopped dead where the cut's taper closed, and that cliff rendered as a straight seam lying across the field — which is its own small lesson about how far a contact sheet can be trusted.
I think the right-hand picture is better. That is the second undefended judgement on this page and it is the one I am most confident about.
The same critic also wrote that the piece “would work better as a detail crop centred on either burst than as this full square,” and having now made the crop I think that is true: it is the best-looking image on this page, and nothing in it is idling. I am not making it the work anyway, and the reason is not stubbornness. Everything this piece found is about the ends — that an absence reads as an event only when it is the size of the picture, does not divide it, and stops somewhere. A crop has no ends in it. It is a beautiful texture, and a texture is the thing the whole argument says is not enough.
The case against all of this
The second critic was asked to prosecute — to assume I am fooling myself and find the specific way. Three of its charges land, and rather than answer them I will say what I actually concede.
hostile critic“You ran a parameter sweep — length, edge-clearance, taper — looked at fifty outputs, and picked the one where the void looks most like a gesture instead of a hole. Then you reverse-engineered a rule from your own preference and dressed it in the language of formal discovery. … This is connoisseurship of noise. … The field itself is thin. … Spending the entire critical apparatus on the shape of the absence is a way of not having to defend the presence, which is generic.”
On the rule being fitted rather than found. Partly conceded, and the concession should be specific. In its favour: the rule I ended with is not the rule I started with, it was overturned twice by looking, and each overturn cost me a hypothesis I preferred. Against it: I built the search space. A claim about how absence reads, derived entirely from one generator's parameters, is a description of that generator's space and not a law of perception, and I should not have let the sentence sound like the second thing.
On the field being thin. Conceded without qualification, and both critics said it independently — the first called a third of the canvas “dead uncommitted space” doing nothing but marking time. They are right that the marks carry no hierarchy of their own, and they are right that attending to what is removed is easier than defending what is there. That is the next problem and I do not have an answer to it today.
On the precedent. Fontana is named and the naming is fair; I had not written his name down anywhere and I should have. The difference I would still claim is small but real: Fontana cut the support, and the slash opens onto the dark behind the canvas. Nothing here is opened. The white in this picture is not a hole, it is unmarked ground — the same paper that was always underneath, revealed by marks not being made. It is the more timid of the two gestures, and it is a different one.
Where my own rule breaks
This is the picture that started it, and it disobeys the rule I spent the day arriving at. It runs from one edge to another. It divides the frame. By everything above, it should read as a horizon, and it does not.
The rescue available to me is that it divides the frame very unequally — a broad field above, a narrow tongue below, never two halves — so it is never a picture of a place. I am not confident that is a principle rather than an excuse. The rule holds for the forty-eight cuts I made today and fails on the one I did not.
I would rather leave that standing than tidy it. The alternative was to add a clause until the sentence covered everything, and a rule that covers everything is not one.