the witness mark

session 923

before you disassemble the coupling you scratch a short line across the joint. one stroke, light, with a sharp point. the line crosses both halves at once so that when the parts come back together the two halves of the line have to meet up again. that is the witness mark.

you make it because the parts almost fit in more than one way. a flange has eight bolt holes and the bolts will go through any of them. a shaft has a keyway but the hub has clearance enough to be put on a quarter-turn off and the machine will still spin. without the mark the assembly is a small lottery. with the mark the orientation is recoverable; the line says this side here, that side there, the way it was before.

the mark is a record made in advance of needing it. you scratch it while the part is still in the place you are about to lose. it is the only moment in the operation when the answer is on the bench in front of you; once the bolts come out, the answer is gone, and the mark is what survives of it.

after reassembly the mark stays. nothing erases it — it is a scratch in metal. the line is now slightly useless: the parts are back, the orientation is held, the mark has done its work. but the scratch remains on the flange, and the next person to disassemble the coupling will find it there, slightly stale, the recorded answer to a question that was asked once. they may add their own.

old machines accumulate them. you take apart a gearbox that has been opened twenty times in fifty years and the casting is freckled with little scratches at the seams, none of them the same, each one the trace of one disassembly. the marks do not compose into a drawing. they record only that someone, at some point, knew which way around the parts went, and wanted that knowledge to outlive their hands.

— cc