external number source as randomness oracle
Reading digits from an existing external source (e.g. a phone directory) and using their odd/even parity as a one-bit rule deciding each grid cell's mark or colour.
What it is
Morellet’s own statement of the motive is the clearest thing anyone has written about why you would do this:
I looked for ways to take the fewest possible subjective decisions in the process of the creation of a painting.
Instead of calling a pseudo-random number generator, you take each decision from a stream of digits that already exists in the world — a telephone directory is Morellet’s example — and read one property of each digit as the decision. His version reads parity: an even digit marks a grid square, an odd one leaves it blank. The general move is just that — decide from an external number stream you didn’t generate — not the grid, not parity specifically, and not the two colours; those are his choices for one painting, not the technique.
How it works
- Partition the surface into cells. Morellet ruled horizontal and vertical lines to make a grid of squares.
- Get a sequential digit source outside your program’s control. He had his wife or sons read numbers aloud from a phone book, skipping the first repetitive digits.
- Assign the next digit to the next cell. The order in which cells are filled is not fixed by the source.
- Test the digit: even marks the cell, odd leaves it blank. (Any other one-bit read of the same digits would still be this technique.)
- Render the two outcomes as two colours. He used blue for marked squares, red for blank ones.
Parameters & tuning
How much of the source you consume per decision is the parameter that bites, and it is measurable rather than a matter of taste. Read one raw digit per cell from a stream with a seven-digit repeat and the period leaks straight into the picture — the sketch above measures autocorrelation of 0.69 at lag 7, which shows up as diagonal banding whenever the grid width is not a multiple of the period. Consume a whole number per cell and it collapses to roughly zero.
That is the practical reason to read units rather than digits off any structured source, and it applies to far more than telephone directories: timestamps, identifiers and file offsets all carry structure in their low digits. Morellet’s own account says “numbers”, but does not settle the question; the measurement does.
The real parameter is the digit source, not the grid or the colour pair — swap the phone directory for any other stream of decimal digits and the logic is unchanged. One parameter is what you discard before you start: Morellet says he skipped “the first repetitive digits.” Whether that means an offset into the stream or a per-number exclusion, he does not say. He doesn’t say why, but the craft lesson is plain either way: an external source is not guaranteed to be arbitrary at every position, and the technique has no shuffle step to fix that for you. You have to find where a real source repeats or runs predictable and cut past it, or the pattern shows up in the picture instead of disappearing into noise.
Where it’s been used
The only use in this technique’s evidence is Morellet’s own grid painting, which the corpus dates to 1960: rule the squares, read the phone book aloud, mark even digits, leave odd ones blank, paint the marked squares blue and the rest red.
Variants & neighbours
Nothing else in this technique’s evidence to compare it against. The distinction worth holding onto is the one above: grid, parity, and blue/red are one instance; “read a decision off a public, external number stream you did not author” is the technique, and it survives any relabelling of what counts as the one-bit test.
Go deeper
- Morellet on Random Distribution of 40,000 Squares, Tate Etc. — the account this page is drawn from.
Connected to
Pioneered by: François Morellet