A reflected-light meter measures luminance, not reflectance. It has no way of knowing whether it is pointed at fresh snow, a black cat, or a face. To produce a single exposure recommendation, it must assume the scene averages to one fixed tone, and it renders whatever it reads as that tone. When the assumption is wrong, the meter is fooled: snow comes out grey, the black cat comes out grey. The grey card exists to give the meter exactly the average tone it expects, removing the subject from the calculation.

Why 18%, and what the meter actually does

The figure 18% is a fact about perception, not about light. Visual lightness tracks roughly the cube root of luminance, the relationship formalised in the CIE L* function: a relative luminance of about 0.18 maps to L* of roughly 50, the exact midpoint between black (L* = 0) and white (L* = 100). A surface that reflects 18 to 20% of the light falling on it therefore reads as the middle grey between the two. A 50% reflectance, by contrast, looks distinctly light grey. Kodak built its Neutral Test Card, Publication R-27, around this: the grey side reflects 18% across the visible spectrum, the reverse white side reflects 90%, and both are matte so neither throws a specular highlight into the meter. Spectral neutrality matters because panchromatic film responds across the whole visible band, and a card with a colour cast would meter differently under tungsten than under daylight, and differently again behind a deep yellow or red filter.

A reflected meter renders whatever it reads as Zone V, the middle grey of the Zone System that Ansel Adams and Fred Archer formalised at the Art Center School around 1939-40 and that Adams set out in The Negative (1948; revised 1981). Meter a grey card directly and you place the card on Zone V. From that anchor the rest of the exposure follows: meter the darkest area in which you still want texture, stop down two stops to drop it onto Zone III, and you have fixed the shadow placement for the negative. With a film like Ilford FP4 Plus or Kodak Tri-X 400 that placement, paired with development, governs where every other tone lands.

How K maps to reflectance

Reflected meters are not calibrated to 18% at all. The governing standard is ISO 2720:1974, which fixes calibration through a reflected-light constant K. The exposure equation is

N² / t = L·S / K

where N is the f-number, t the shutter time in seconds, L the luminance, and S the ISO speed. The standard recommends K in the range 10.6 to 13.4 with luminance in cd/m², and it is explicit that K and the incident constant C “shall be chosen by statistical analysis of the results of a large number of tests” of photographs judged acceptable by many observers, across many subjects and luminances. It is empirical, not derived from any ideal grey.

In practice Canon, Nikon and Sekonic settle on K = 12.5; Minolta, Pentax and Kenko on K = 14. To turn K into an implied reflectance you need the incident side as well, because reflectance is R = π·K / C, where C is the incident calibration constant. ISO 2720:1974 puts the hemispherical (dome) C in the range 320 to 540 lux, with Minolta near 320 and Sekonic near 340. Plug K = 12.5 and C = 330 into R = π·K / C and you get about 11.9%, call it 12%. Run K = 14 through the same relation and you get about 13.7%, closer to the 18% card but still short of it. The reflectance a meter “expects” depends on the assumed scene illuminance through C, not on K alone.

The half stop, worked

Take a Sekonic (K = 12.5) and an 18% R-27 card in even, open shade, EV 12 at ISO 400. The meter reads f/8 at 1/250. But the card is 18% and the meter is built for about 12.5%, and 18 / 12.5 = 1.44; log₂(1.44) = 0.53. The card is 0.53 stop brighter than the tone the meter assumes, so the meter, doing its job, recommends about half a stop less light and renders the card half a stop dark. Open up: f/8 at 1/180 (or hold the time and open to roughly f/6.7) and the card lands on true middle grey, with the rest of the scene exposed to match.

Hand the same card to a Pentax or Minolta meter at K = 14 and the implied reflectance is around 13.7%; the gap to 18% shrinks to roughly four-tenths of a stop, less than half the offset a K = 12.5 meter carries. The discrepancy is real, but its size depends on which meter you hold.

Where the half stop does not come from

It is tempting to credit this correction to Kodak, and the attribution is wrong. The R-27 instruction sheet prescribes no half-stop adjustment for a normal subject. Its adjustment guidance is, verbatim: “with either side of the card, decrease the calculated exposure by 1/2 to 1 stop if the subject is unusually light. If the subject is unusually dark, increase the calculated exposure by 1/2 to 1 stop.” For a subject of normal reflectance you use the card reading directly. The corrective half stop above comes from the K-versus-reflectance mismatch, an artefact of meter calibration, not from anything on the sheet.

The sheet is precise about geometry, and it is worth following exactly. Hold the card straight up and down, close to and in front of the subject, facing halfway between the camera and the main light — not square to the camera, which under-reads, and not square to the light, which over-reads. Meter no more than six inches away, and keep your own shadow and the meter’s off the card. In dim light the sheet has a separate trick: the white side reflects five times the grey, so divide the film-speed setting by five, read off the 90% side, and use the indicated exposure for a normal subject. Because the card reading is in effect a measurement of incident illumination, the grey card is the conceptual twin of an incident meter; the R-27 sheet says as much, describing its readings as “measurements of incident illumination.”

When it actually matters

The half stop bites only when your latitude is tight. A negative B&W film such as HP5 Plus comfortably absorbs roughly +2 to -1 stop around the metered point, so a half-stop calibration offset disappears into the shoulder and the printing. Reversal stock and precise Zone work give you perhaps half a stop to a stop of usable latitude, and there the offset shows. The honest value of the card is not its exact percentage but its repeatability: it converts a guess about an unknown scene into a measurement of the light alone, and whatever fixed offset your meter carries can be characterised once and dialled in every time after. For the deeper exposure-and-development chain that hangs off this anchor, Adams’s The Negative and Lambrecht and Woodhouse’s Way Beyond Monochrome are the standard references.