One rotation of the microfilm reader's hand crank advances the reel roughly three-quarters of an inch, translating, depending on the density of the archive being searched, to somewhere between one and four newspaper pages. A researcher hunting a specific date in a specific year can expect forty to sixty rotations before the target surfaces, assuming the reel was wound correctly onto its takeup spool by whoever used the machine last, which is not guaranteed.
The screen is a pane of frosted glass lit from below, tilted at an angle that flatters no posture, throwing an image always slightly larger than life and always slightly out of true, kept in check by a focus knob that never quite settles on sharp. Text swims past too fast under an eager hand and must be walked back, half a rotation at a time, the machine's only real demand being a patience it has no mechanism to enforce.
Some readers still carry a faint chemical note from the acetate film stock loaded decades earlier, a smell that persists in institutional memory longer than the film itself will. The newer machines scan and project to a monitor and call this progress, and by most measures it is, though something goes missing in the trade: the crank offered proof of labor, a count of rotations between a researcher and the thing being sought, that a scroll wheel performs without registering the weight of the search at all.
Most branches keep one working reader for every four broken ones, a ratio that has held for years without anyone deciding it should.