A meter reports a single number, but the consequences of acting on it differ between film and a digital sensor. The two media fail in opposite directions: film loses information first in the shadows, digital first in the highlights. Understanding why turns a vague rule of thumb into a deliberate strategy, because the safe direction to bias an exposure is not the same on each.

The film curve, with numbers on the axes

A negative’s behaviour is described by its characteristic curve, the plot of density against the logarithm of exposure. The curve has three regions: a toe, where the slope is low and shadow tones are compressed; a long, roughly straight central section whose slope is the gamma, around 0.6 for a normally developed general-purpose film; and a shoulder, where density levels off as the emulsion approaches maximum.

The bottom of that toe is not a vague edge but a defined point. ISO 6 places the speed point where density first rises 0.10 above base+fog, and fixes development so that a point 1.30 log-exposure units further along the curve, about 4.33 stops brighter, sits 0.80 in density above the speed point. That ratio gives the standard average gradient of 0.62 used to certify the box speed. Below the 0.10 mark, adjacent shadow values record as the same density and merge. That is the threshold: starve the shadows of the light needed to clear it and no print or scan recovers separation that was never written to the film.

Highlights sit on the straight line, which is long enough that overexposure is forgiving. Kodak states Tri-X 400 can be underexposed by as much as three stops and recovered by push processing, at the cost of higher contrast, coarser grain and yet more lost shadow detail, while overexposure is tolerated far more generously. The asymmetry is concrete: a stop of overexposure rides smoothly up the straight line at gamma 0.6, whereas a stop of underexposure drops a tone onto the compressing toe where the slope collapses toward zero.

The Zone System makes placement numerical

Ansel Adams and Fred Archer worked the Zone System out around 1939 to 1940, and Adams codified it in The Negative (1948, revised 1981). Each zone is one stop. Zone V is middle grey, the tone a reflected meter is built to render; Zone III is the darkest shadow that still shows texture; Zone VIII is the brightest textured highlight. The rule “expose for the shadows, develop for the highlights” follows directly from the curve: shadow placement is locked at exposure, while development moves high densities far more than low ones.

Work an example with Ilford HP5 Plus, rated ISO 400/27°. Spot-meter a deep shadow that must keep texture; the meter wants to make it Zone V, so close down two stops to drop it onto Zone III. Shoot at EI 400 and develop in Ilfotec DD-X at 1+4, 20°C, for 9 minutes, the box-speed time; in stock ID-11 the equivalent is 7 min 30 sec. A textured highlight three to five stops above that shadow then lands near Zone VIII on the straight line. If the scene is too contrasty and that highlight threatens Zone IX, an N-1 contraction, a shorter development time, pulls it back down to Zone VIII while leaving the Zone III shadow essentially untouched, because low densities barely respond to development. N+1 expansion, roughly 30 per cent more time, does the reverse for a flat scene, lifting a Zone VII placement to print as Zone VIII.

Why digital sensors fail the other way

A digital sensor inverts the situation because its response is essentially linear. Each photosite accumulates charge in direct proportion to the photons it receives, up to a hard saturation point, the full-well capacity. There is no shoulder. Once a photosite fills it returns the maximum value, and every brighter tone clips to the same white with no gradation to recover.

The shadows survive better than film’s, but compete with noise. As Emil Martinec sets out in Noise, Dynamic Range and Bit Depth in Digital SLRs (2008), total noise combines read noise R and photon shot noise P in quadrature, N² = R² + P². Shot noise is Poisson: its magnitude is the square root of the photons collected. Collect 10,000 photons and the noise is 100, an SNR of 100; collect only 100 photons and the noise is 10, an SNR of just 10. Bright tones therefore carry far cleaner signal than dark ones. The usable range is roughly full-well capacity divided by read noise: an 18,000 e- well with 4 e- read noise gives about 4500:1, around 12 stops. Lifting underexposed shadows amplifies the noise already living there; a clipped highlight offers nothing to lift.

Exposing to the right, and the myth inside it

The standard digital advice is to expose to the right, ETTR: push the histogram as bright as possible without clipping. The old justification was level count. In a 12-bit raw file of 4096 levels, the response being linear, the brightest stop holds about 2048 levels, the next 1024, then 512, 256, 128, halving with each stop down toward black, so the deepest shadows are described by very few levels. Spend the exposure in the bright stops and you appear to capture far more tonal information.

Martinec’s correction is the real payoff: that level-count argument is largely a red herring. In the highlights, shot noise already exceeds the spacing between adjacent levels, so the extra levels record nothing the noise has not already blurred. The genuine reason to expose to the right is SNR, the same square-root law as before. More light means more photons, and more photons means a cleaner signal everywhere, especially in the shadows that would otherwise sit near the read-noise floor.

A single meter number, two opposite biases

A reflected meter renders whatever it reads as a fixed mid-tone, Zone V, conventionally taken as 18 per cent grey, set by its K-factor calibration. That is exactly why a single number is ambiguous: the meter does not know whether it is pointed at snow or coal, so the photographer must decide which scene tone to place where. The bias direction is a choice the medium makes for you.

With film the unrecoverable error is the lost shadow, so anchor the reading to the darkest tone that must keep texture, spot-metering it and placing it on Zone III, and let the highlights drift up the straight line into the protective shoulder. With digital the unrecoverable error is the clipped highlight, so set exposure as bright as possible without saturating the brightest important tone, watching the right edge of the histogram and the clipping blinkies rather than the shadows. The goal is identical in both: fit the scene to where the medium records it most gracefully. The media simply disagree about which end is fragile.

The print is a third curve

For film there is one more character. Photographic paper has its own characteristic curve, and it inverts the film’s: where the film toe compresses shadows, the paper has a shoulder that compresses its own dark tones, and the paper’s toe handles the highlights. Print an Ilford Multigrade negative onto Multigrade RC or FB and that paper curve re-maps the negative’s entire range to fit the reflective scale of a print. Seen this way, “develop for the highlights” is really about fitting the negative’s density range to the paper, and the film’s shoulder is not just safety margin but a feature: it gently rolls off the brightest tones into a region the paper can still hold, rather than slamming them against a wall the way a sensor does at full well.