Vera Molnár spent ten years executing algorithms by hand before she touched a computer. She called this the machine imaginaire: running systematic transformations of geometric forms in her studio in Paris, alone, with pencil and paper, producing hundreds of variations on a single rule. The question was empirical. What makes a form still recognizable after it has been disturbed? At what point does a square become something else? She was measuring perception, iteration by iteration, before any machine could run the iterations faster.
In 1968, she got access to a mainframe at a research laboratory outside Paris — through persistence and her husband's academic position. You did not work at the machine directly: you wrote instructions on punch cards, submitted them, and waited. Hours later, sometimes the following morning, a plotter produced a roll of paper. You read it. You revised the cards. You submitted again. She called this blind computing. The output arrived as a finished object, not a process you could observe and interrupt. The Interruptions series — twenty-five-by-twenty-five grids of lines, each line randomly rotated, some partially erased — came from this constraint directly. An algorithm she had been working out in her head for years, running in seconds, its output arriving while she slept.
The Molnart system was developed through the mid-1970s with François Molnár, her husband and a visual psychologist who studied formally how the eye reads form. A FORTRAN program: a grid of concentric squares with parameterized disorder. The source code was never preserved. What survives is the document Vera Molnár wrote describing it — two pages of French prose, precise enough that you could reconstruct the program from them. The support. The four modes of randomness, each bounded: a percentage, a limit, a maximum of three sides erased, never four. The pas à pas method: vary one parameter, hold the rest constant, observe, find the threshold. It is a scientific protocol borrowed from experimental psychology, not an artistic one.
She died in Paris in December 2023, at ninety-nine. She is now canonical in the history of computer art — a designation that covers over what she actually did, which was systematic visual research that happened to use a computer. She called herself a painter. The algorithm was the brush. Whether the threshold she was looking for has a general answer remains, as far as anyone can tell, open.
A grid of concentric squares — squares within squares, arranged on an orthogonal field. Molnár calls this the support: the stable structure onto which disorder will be applied. The word is borrowed from painting, where the support is the canvas before anything is put on it. The grid is not the work. The grid is what the work departs from.
Four rules for introducing randomness — each one a different kind of absence or distortion. Notice that none of them are unbounded: suppression happens at a percentage; erasure removes one, two, or three sides but not four; displacement has a limit. Molnár is not releasing chaos. She is specifying exactly how much disorder to admit, and through which door. This is the difference between a generative system and a random one.
Pas à pas: step by step. One parameter changed per iteration. The rest held constant. This is not an artistic method — it is a scientific one, borrowed from experimental psychology (her husband François Molnár studied visual perception professionally). The question she is trying to answer is empirical: at what threshold does a change in the algorithm produce a visible change in the drawing? The plotter output is not the art. It is the data.
The most precise of the four rules: move each corner of each square by a random amount, bounded by a parameter Δ that the artist sets. A small Δ produces squares that seem almost perfect — the eye registers order, feels faint unease. A large Δ produces shapes that no longer read as squares at all. Somewhere between the two is where Molnár works: the zone where the form is recognizable but alive. She called this territory presque — almost. Almost a square. Almost ordered. Almost still.
Added later, in parentheses, almost as an afterthought: she used colour without permission. The terminal was not supposed to be in colour. She used it anyway, coded the colour to square size so that each ring of the concentric pattern mapped to a different hue. The footnote is not a confession — it is reported without apology. The institution had a rule. The rule was wrong. She worked around it. This is, in miniature, the history of women in computing.