How fine a ruler really reads

A ruler printed with millimetre marks does not resolve to a millimetre. It resolves to about half of one, because the marks themselves have width, and the practical limit of what you can read is set by that width as much as by the spacing between marks.

Why a mark is not a point

A millimetre line on a ruler is not an idealised infinitely thin point in space. It is ink or engraving with a physical width, commonly a few tenths of a millimetre, and when a measured edge falls between two marks, you are judging its position relative to two thick lines rather than two mathematical points. Interpolating between marks is a real and legitimate technique - reading "just past the third mark" rather than rounding to it is more honest than rounding - but the width of the marks themselves puts a floor under how finely that interpolation can be trusted. In practice, most people reading a millimetre-graduated ruler can resolve to somewhere around half a millimetre with a steady hand and a good sightline, and claiming finer than that is claiming more than the instrument supports.

Where this shows up

This matters most when you are recording a reading to a decimal place that implies more certainty than the ruler has. Writing down 15.5 cm from three readings that scattered by two millimetres is a genuinely different act from writing 15.52 cm as if the second decimal meant something - the instrument cannot support it, and how careful technique's error bar is set is already wider than the gap between those two numbers. The resolution limit also interacts with reading angle: sighting a mark from an angle rather than straight on can shift where it appears to line up by more than the mark's own width, which is a separate and additional error on top of the mark-width limit itself.

What does and does not fix it

A finer-printed scale does not fix it past a point, because the marks would simply be thinner lines closer together, and a human eye and hand combination has its own limit on distinguishing adjacent close marks regardless of how fine the printing is. Engraved markings, cut into the material rather than printed on top of it, tend to stay crisp for longer and do not blur or wear the way ink can, which is a real if secondary advantage - engraved and printed scales compared directly goes through that tradeoff in more depth. A magnifier changes nothing about where the true edge is; it only makes the existing mark width easier to see, which does not narrow it.

What genuinely helps is reading square to the ruler rather than at an angle, and taking the median of three readings rather than trusting one, both of which reduce the scatter around the true value without pretending the ruler itself is finer than it is.

A quick way to see it for yourself

Look at a ruler under strong light and try to say, honestly, where a mark's centre is rather than its edge. On most millimetre rulers the line itself has a visible width when you look closely, and different people will point to slightly different spots as "the centre" of the same line. That small disagreement, repeated across two marks bounding a reading, is the resolution limit made visible - it is not a flaw in any one ruler, it is a property of using a printed line to represent a point.

The honest resolution to record at

Round to the nearest millimetre, or state your reading with a half-millimetre uncertainty attached, and stop there. This is not a limitation specific to measuring size - it is true of reading any ruler for anything, and the half-centimetre spread that shows up between repeat readings is partly this effect and partly genuine physiological variation, which is worth separating in your own head even if the ruler cannot separate them for you.

None of this has anything to do with a subjective score. Rate Cock is not reading a scale at all, and its output is a position among judged photographs rather than a distance with a resolution limit - a different kind of number, answering a different question, and one where an AI model's read of a photo has its own separate error sources that have nothing to do with mark width. The same goes for a rating built from tools that score an image rather than measure a shaft directly, or for a human judge's verdict, which was never on a scale in the first place.