Two Archive Scans of Hey Diddle Diddle: A Resolution and Line-Fidelity Test

The two files sat side by side on the studio monitor at 4:12pm on a rainy Thursday, both labelled as scans of the same nineteenth-century Hey Diddle Diddle plate, both offered as high-resolution, both downloaded from public archives within the same hour. One was 600 DPI. The other was 1200 DPI. On paper the second file should have been the clear winner for anyone drawing a nursery chart from it. In practice, once we zoomed to 400 percent and looked at the line the cat had been drawn with, the comparison stopped being about numbers and became something quieter and more interesting.

The Two Files, on the Same Screen at 400%

We put them into two adjacent windows, matched the zoom, matched the pan, and locked the scroll. The 600 DPI file sat on the left. The 1200 DPI file sat on the right. Both had come from institutional archives that photograph or scan children's book plates and publish them for anyone to use. Both were TIFFs. Both were over a hundred megabytes. Both had the same plate identifier printed in the corner. From two feet back the images were visibly the same picture: cat, fiddle, cow, moon, dish, spoon, the whole rhyme laid out in the plate compositor's original arrangement.

Then we started zooming. At 100 percent the two files looked, honestly, indistinguishable to anyone not already inclined to squint. At 200 percent the second file's edges started to feel a little cleaner, but the difference was closer to a preference than a fact. At 400 percent the two files stopped looking like the same picture. The higher-resolution file was picking up things the lower one had smoothed away — the tremor of the engraver's hand along the cat's back, the tiny stipple that gave the cow its volume, the interruptions in the moon's outline where the plate had held ink unevenly. The lower-resolution file had been kind to those imperfections. The higher-resolution one had not been kind; it had been honest. Those are different jobs.

This is where the comparison shifts. Anyone recommending a scan by its DPI number alone is recommending the wrong thing. DPI is a container size. What the scan actually contains is a separate question.

Where Resolution Numbers Stop Meaning Anything

The standard advice, when someone is looking for public-domain source material for a project, is to go for the highest resolution available. This advice appears in blog after blog, forum after forum, always with the same confident tone: bigger file, better quality, done. It is the advice of people who have not opened the files. Resolution is a measurement of how many samples the scanner took per inch. It says nothing about what those samples captured, how the scanner was lit, whether the original had been laid perfectly flat, whether the archive did any post-processing before publishing, or whether the scanner was even capable of resolving detail at the density it was sampling.

Our two files illustrated all of this at once. The 600 DPI file had been produced by an archive that clearly used soft, diffuse lighting. Its shadows were gentle, its blacks were slightly greyed, and the paper texture was legible without being noisy. The 1200 DPI file had been produced under harder, more direct light. Its shadows were crisp, its blacks were true black, and the paper texture came through as a fine noise that had to be visually filtered out to read the illustration underneath. Neither approach was wrong. They were answering different questions. The 600 DPI file was answering "what does this plate look like as an image?" The 1200 DPI file was answering "what does this piece of paper physically look like right now?"

For a studio drawing a nursery chart, that distinction is the entire game. We are not reproducing the paper. We are reading the drawing off the paper. A scan optimised for archival fidelity to the object is not the same as a scan optimised for readability of the image the object contains. Recommending one metric — DPI — as a stand-in for both jobs is recommending the wrong X, and it is recommended constantly because it is the easiest number to quote.

Hey Diddle Diddle The print from this article · from €29.95 View the print →

The Line the Cat Was Drawn With

Everything downstream of a nursery chart lives or dies on line quality. The line is what the child's eye follows. It is what a two-year-old traces with a finger from the bed. It is what has to survive being redrawn at a different scale, in a different palette, on a printer with a different halftone screen. When we take an archive scan and use it as reference for original vector work, we are not tracing pixels — we are reading intent. We are trying to see what the original engraver was trying to do with each stroke, so that a fresh, honest drawing can be built from that reading. The scan is a translator. Bad translators lose the tone of voice.

The cat in our two files is a small figure, maybe an inch and a half tall on the plate. In the 600 DPI file, the line describing the cat's spine was continuous, smooth, and about three pixels wide at that magnification. In the 1200 DPI file, the same line was six pixels wide, but it was not one line — it was a bundle. You could see three separate engraver's strokes overlapping, with the tiny gap between them where the plate had held less ink. That gap is meaningful information. It tells us the direction the tool moved. It tells us how the engraver built up density. It tells us where they hesitated. The 600 DPI file had smoothed the bundle into a single confident stroke that the engraver never actually drew.

If we had drawn our vector cat from the 600 DPI file, we would have produced a cat with a slightly false confidence — an animal whose spine had been drawn in one gesture by someone who never existed. From the 1200 DPI file we could see the real hand, hesitations included, and choose which of those hesitations to honour in a fresh drawing and which to quietly translate out. That choice was only available to us because the higher-density scan had refused to be helpful. It had preserved the mess. The mess was the source.

What a Nursery Wall Actually Needs From an Archive Scan

Here is where the analysis stops being about scanning and starts being about walls. A nursery chart hangs, most often, somewhere between four and five feet off the floor — high enough for an adult to read while carrying a child, low enough for the child to see when they are standing on the floor themselves. The reading distance for a small person under three feet tall, looking up at a chart above the changing table, is roughly two to three feet. At that distance, on a printed chart in the twelve-to-eighteen-inch size range, almost none of the sub-pixel detail from either scan is going to be visible. The child will see a shape, a colour, a friendly presence. Neither the engraver's hesitation nor the archive's noise will register.

So why does the source scan matter? Because the drawing process compresses. When a studio draws a fresh vector chart from an old plate, every honest decision made in the reference stage propagates outward into the finished piece. If the reference was smoothed and confident, the vector redraw will be smoother still, and by the time it lands on a nursery wall it will feel like clip art — pleasant, professional, and empty. If the reference was faithful to the original hand, the vector redraw carries a residue of that hand even after it has been simplified into pastels and clean shapes. That residue is what makes a chart feel drawn rather than generated. It is the difference between a wall that a small child returns to and a wall that becomes background within a week.

The industry-standard recommendation for source material — "pick the highest DPI available and get on with it" — leaves this whole layer unexamined. Our test suggested a better working rule: pick the scan that shows you the most information you would rather not have, then decide as an artist which of that information to keep and which to translate away. Ease is the wrong criterion. Honesty is the right one. This is the same principle we apply when we draw a chart like Dinosaurs to Scale from commonly cited palaeontology length estimates rather than from a rounded, tidy version of the numbers — the awkward figures are the ones that keep the drawing from lying. If any of this thinking makes you want to see the finished work rather than the source material behind it, the current collection lives at see the Hey Diddle Diddle print.

Notes on What This Test Did Not Cover

This piece compared two scans of one plate. It did not compare scanners; the archives publish very little about the hardware they use, and the meaningful variable is the finished file, not the machine. It did not address colour management. Both files were greyscale and both were viewed on the same calibrated monitor, so colour did not enter the comparison — but for coloured plates the same DPI number can hide much larger differences in colour handling than in line rendering, and that is its own investigation.

It did not test the scans through a printing pipeline. What we saw at 400 percent on a screen is not what a chart printed at eighteen inches wide would show; the printer's own halftone, the paper stock, and the ink coverage all filter aggressively, and some of the fidelity we prized in the 1200 DPI file would collapse under a coarse screen. A separate test, done at the print stage rather than at the reference stage, would produce different conclusions.

It did not cover copyright. Both source files were published as public-domain reproductions of pre-1900 plates by institutions that state as much on their download pages, but the rules governing whether a scan itself acquires new rights vary by jurisdiction, and that is a legal argument the studio is not qualified to make. And it did not cover the many public-domain nursery rhyme plates that are only available as low-resolution derivatives, where the choice is not between two archives but between using what exists or drawing without a reference at all. Each of those is a separate piece.

FAQ

Does the highest-DPI scan always give the best result for redrawing?

No. In our test the 1200 DPI file was more useful for a studio redraw because it preserved information — engraver's hesitations, plate noise, stroke bundles — that the 600 DPI file had smoothed away. But "more information" is not the same as "prettier". For someone who wanted a finished-looking reference to print directly, the 600 DPI scan was more immediately pleasant. The right file depends on whether the scan is a destination or a source.

Is a 600 DPI archive scan too low for professional nursery-chart work?

Not inherently. 600 DPI is more than enough spatial detail for any nursery chart printed at eighteen inches or smaller. The limitation we found was not resolution but processing — the lower-density scan had been softened in ways that hid the underlying line quality. A well-lit, unprocessed 600 DPI scan can be more useful than a harshly processed 1200 DPI one. The number on its own does not tell you which situation you are in.

Why not just work from a modern reproduction of the nursery rhyme?

Modern reproductions are usually re-illustrations, not reprints. Working from one means drawing a drawing of a drawing, with each layer averaging out the character of the original. The whole point of going to archive scans is to reach past the copies and read the historical hand directly, so that a fresh drawing carries something of the original's tone rather than the last illustrator's aesthetic.

How can you tell if a scan has been processed before the archive published it?

Look at the shadow edges and the flat areas of paper. Unprocessed scans have gentle shadow transitions and slightly noisy paper because that is what paper looks like under a light. Aggressively processed scans have very clean shadow edges and unnaturally smooth paper because a filter has been applied to reduce noise, and the same filter almost always eats fine line detail as collateral damage.

Does any of this matter for a chart hung above a crib?

It matters upstream, not downstream. The child looking up from the crib will not see the engraver's hand or the archive's noise. But the chart they are looking at was built from decisions the studio made while reading a source, and those decisions determine whether the drawing on the wall feels alive or feels generic. The source discipline shows up in the wall six steps later.

Should we always use two scans and compare them?

When two exist, yes. Where a plate has only been scanned once, the comparison collapses to reading that one scan carefully — asking what it preserves and what it may have flattened — rather than trusting its metadata. The habit worth keeping is the zoom to 400 percent and the question "what is this file trying to tell me?", regardless of how many files are available.