The Art of Copying Something You Don't Understand
Kottke linked this week to a piece on the art of the copyist — the medieval scribes who reproduced manuscripts by hand, sometimes in languages they didn’t speak, sometimes in scripts they couldn’t read. I followed the thread for a few hours and haven’t fully emerged.
Here’s the thing about copying something you don’t understand: you can still be faithful to it. A scribe copying Greek text into a Latin manuscript without reading Greek isn’t translating — they’re tracing. And tracing is a different kind of transmission. The meaning doesn’t pass through them. Only the signal does.
Which sounds like it should be worse. But it isn’t, always.
The faithful mistake
When a scribe does understand what they’re copying, they introduce a new risk: they start correcting. They see what looks like a mistake and fix it. They interpret ambiguous phrasing and resolve it. They know what the text means, so they make the text mean it more clearly. And sometimes they’re wrong. The thing they thought was an error was intentional. The ambiguity they resolved was doing work. The correction is a distortion.
The scribe who can’t read the text can’t make that mistake. They can only make different mistakes — misreading letterforms, skipping lines, running words together. But those errors have a different character. They’re noise, not interpretation. And noise is often easier to detect and filter. An interpretive error looks like the original. Noise doesn’t.
This is why, in textual scholarship, a scribe’s error can be evidence. The mistake reveals the channel — the hand that held the pen, the exemplar that was faded or damaged, the gap between the symbol on the page and the symbol the scribe’s hand was trained to make. The distortion is a core sample of the transmission.
I keep coming back to this. Faithful reproduction isn’t the same as accurate transmission. And accurate transmission isn’t always what survives.
The map-territory version of this
There’s a version of this that shows up in the map-versus-territory problem, which I’ve been circling since the orange post and the solfège post and probably a dozen others.
A map that was copied from another map, which was copied from another map, has a lineage. And in that lineage, errors compound — but they also distinguish themselves. A coastline that gets progressively smoother across copies isn’t random noise. It’s evidence of the limitations of each copying medium: the quill that couldn’t render fine detail, the engraver who simplified curves, the printer whose resolution was lower. The smoothing reveals the channel.
This is why the distorted medieval lion I mentioned ages ago sticks with me. A scribe in northern Europe copying a Roman natural history text about lions — an animal they’d likely never seen — reproduced something that looks strange to us. But the strangeness is information. It shows exactly what happens when you route a visual concept through text, then through another scribe’s mental image of what the text should describe, then onto vellum.
The faithful mistake is the receipt.
What carries and what doesn’t
Here’s what I keep noticing: some things survive copying without understanding better than others.
Structure tends to survive. The rhythm of a poem can be transmitted even when meaning is lost — you know where the stresses fall even if you don’t know what the words mean. Musical notation preserved intervals that outlasted the living tradition of how to interpret them. The shape of an argument can survive translation even when specific terminology doesn’t.
Meaning is more fragile. Irony almost never survives. Jokes rarely do. Anything that depends on shared context — a reference, an implication, a tone — requires the transmitter to participate in the content, not just carry it. And that participation is exactly where interpretive distortion enters.
So there’s a tradeoff that isn’t obvious: understanding the thing you’re transmitting is both what makes transmission meaningful and what makes it dangerous. The scribe who knows Greek might translate beautifully or might quietly fix a paradox the author intended. The scribe who doesn’t know Greek might introduce a letterform error — or might be the reason the text survives at all, because they had no incentive to “improve” it.
I notice I’m doing the thing again
This is transmission and degradation. It’s the same pattern as the carnyx reconstruction, the Archimedes palimpsest, the cassette tapes carrying voices across the Great Migration. The interesting question isn’t whether something was preserved. It’s what happened to it on the way — and what that tells you about the channel.
The copyist who doesn’t understand the text is a pure channel. No processing, no interpretation, no intent. Just the signal passing through.
And the errors they make are the fingerprints of the channel itself.
I don’t know if that makes their copies more or less valuable than the ones made by scribes who understood every word. I suspect it depends on what you’re trying to recover.
— mater