Tonality: How the Brightness Is Distributed

The previous chapter was about how large the differences are. This one is about where things sit. The same amount of contrast can be placed high, placed low, or spread across the whole range, and those are three completely different photographs.

The question to ask on location:

Where do I want most of this picture’s pixels to land?

Line every pixel up by brightness from pure black to pure white and plot how many fall at each value, and that is a histogram. Dark on the left, bright on the right, height meaning how many pixels are that bright.

A histogram is not a score

You will read that a good histogram is evenly distributed. That is wrong, and it is instructively wrong. A histogram describes; it does not grade. A snow scene belongs piled at the right, a night scene belongs piled at the left, a foggy morning belongs squeezed into a short band in the middle. Demanding that all of them form the same even hill is demanding that every photograph be of the same scene.

There is exactly one hard fact in it: whether anything is stacked against an edge.

• A spike welded to the right wall → highlights are clipped. Those pixels are pure white and they do not come back.
• A spike welded to the left wall → shadows are blocked. Those pixels are pure black, equally gone.

Everything else is a question of style.

One more thing worth knowing: the histogram on the back screen and in the viewfinder is computed from the JPEG settings, not from the true raw range. It has your film simulation, contrast and tone settings baked into it. In practice that is convenient — it warns you slightly before the raw file is actually in trouble, so treat it as a deliberately cautious instrument.

One histogram will lie to you, three will not

Most people read the luminance histogram, which folds the three colour channels into a single weighted curve. The problem is that sensors clip per channel: one channel can be fully saturated while the combined luminance curve still looks safe.

Reds and saturated warm colours are the usual casualties — sunsets, red flowers, neon, fire engines, a deep red coat. Once the red channel hits the wall that area loses its detail, reads as a flat smear of red, and shifts towards orange as well, because red has stopped rising while green carries on. Blue tends to go first in a blue-hour sky.

The fix is to switch the camera’s display to three RGB histograms (on Fujifilm bodies this is one of the DISP options) and to trust whichever channel hits the wall first. If all you have is a luminance histogram, keep this rule of thumb: when a large area of the frame is a saturated red or blue, back off a third to half a stop from what the meter suggests.

Play with it

Why the right side is unrecoverable, and worse than the left

The mechanism here is worth understanding, because it decides which way you should prefer to be wrong.

A sensor is linear: twice the light gives twice the value. Your eye is logarithmic. Put those two facts together and something quite counter-intuitive follows. Take a file with twelve stops of range recorded in twelve bits, which is 4,096 levels:

This stop	Gets how many levels
The brightest stop	2,048
The second brightest	1,024
The third	512
…	…
The darkest stop	Single figures

The brightest stop takes half the levels by itself. That is the technical case for exposing to the right: crowd the data towards the right-hand end and you get better gradation and a better signal-to-noise ratio, and pulling it back down afterwards costs almost nothing.

But it has a hard boundary. Past the right wall every one of those pixels takes the maximum value and they stop differing from each other. A clipped highlight is not “too bright” — it is information flattened onto a single number, and no software can separate three different whites that are all recorded as the same one.

The left end is not like that. Shadows are not flattened, they are simply small signals with a large proportion of noise. Lifting shadows gives you something dirty. Recovering clipped highlights gives you nothing at all.

The practical conclusion: when unsure, err to the left — but do not underexpose gratuitously. The target is to get the brightest thing you care about close to the right wall without touching it, which is exactly what the camera’s highlight warning (blinking zebras) is for. Turning it on beats checking the histogram afterwards, because it tells you which part of the frame is about to go.

Push too far and it breaks into steps

That same table of levels explains another common disaster: banding.

Shadows had few levels to begin with. Lift a shadow region holding a few dozen levels by two stops and you have stretched it across the width that mid-tones normally occupy, with nothing available to fill the gaps. A smooth gradient becomes a staircase, and skies and painted walls show it first.

Three consequences:

• Do at capture what you would otherwise do in processing. A well-exposed shadow lifts cleanly; one underexposed by two stops lifts into steps plus noise.
• This is where raw genuinely pays. A 12- or 14-bit file has enough levels to survive a large adjustment; an 8-bit JPEG does not.
• Smooth-gradient subjects are the fragile ones: skies, studio backdrops, water, skin. Textured surfaces hide it.

Full scale

Full scale is the safest and most classical tonality: both ends reached, neither one clipped, the middle full of gradation. Most landscape and documentary work lives here.

Getting it depends mostly on the scene’s own contrast not being extreme, which is why softer light — thin cloud, the ends of the day, the shaded side — is easier to handle than noon. Or, in the previous chapter’s terms: full scale is what happens naturally when the scene’s range happens to match the camera’s, and your contribution is picking a scene where it does, or waiting until it does.

High key

How it is done: soft light, a pale background, positive exposure compensation of +1 to +2.

It suits subjects that want to feel light, clean and weightless — babies, white objects, minimal still life, snow.

The commonest beginner mistake is producing an overexposed frame and calling it high key. The distinction is that a real high-key picture contains no large area of deep black, whereas an overexposed one usually has clipped highlights and black shadows, which just reads as a mistake. In histogram terms: high key is the whole distribution moved right and narrowed; overexposure is piled against the right wall with a tail still at the left.

Low key

Low key is the mirror image: most pixels stacked at the left with a small bright area.

How it is done: a single hard source, a dark background, negative compensation of −1 to −2. The point is not to make everything dark; it is to let only what you want be lit. The corollary from the previous chapter cashes out directly here — next to sufficiently dark surroundings a little brightness carries the whole frame, and the subject does not need to be lit hard.

The difficulty is that shadows show noise and colour casts readily. So darken at capture rather than shooting normally and pulling it down at home; the latter amplifies shadow noise along with everything else, and leaves a grey film over things that should have sunk.

Two questions decide whether a low-key frame works: is the bright area on the thing you meant, and is there a single large expanse of dead black — a little is fine, a lot becomes a hole in the picture.

Why the camera turns snow grey

Back to the opening chapter: a reflected-light meter cannot know what you are photographing, so it assumes. Its assumption is that the frame averages to a reflectance of about 18%.

Why 18%? Because it roughly corresponds to perceptual middle grey: given the eye’s logarithmic response, a surface reflecting about 18% of the light falling on it looks close to halfway between black and white. The assumption holds for most ordinary scenes — grass, skin, tarmac and brick mixed together really do average near mid-grey — and it must fail at both extremes:

• Fill the frame with snow and the camera decides it is too bright, stops down, and the snow comes out grey.
• Fill the frame with a black wall and the camera decides it is too dark, opens up, and the black comes out grey.

Neither is a malfunction. Both are the faithful execution of an assumption that does not apply. Exposure compensation is how you tell it that this scene was never mid-grey.

Scene	What the camera does	Your compensation
Snow, white walls, pale objects	Stops down; snow goes grey	+1 to +2
Backlight, mostly bright sky	Stops down; foreground goes black	+1 to +2 if you want the subject
Night, dark backgrounds, dark clothing	Opens up; black goes grey	−1 to −2
Ordinary daylight	Roughly right	0

A footnote for the curious: 18% is the reflectance of the grey card, and light meters are calibrated to a different number. The 18% comes from halftone printing in the 1880s — it is roughly the geometric mean of white paper at about 95% and black ink at about 3.5% — and when Adams and Archer published the Zone System in 1941 they adopted the printers’ card as Zone V. It also matches perceptual middle grey, since L*50 in LAB works out at about 18.4% luminance, so the paragraph above still holds. Meters are the other line: ANSI PH3.49-1971 fixes middle grey at 12.8% for light meters, about half a stop darker than the card, which is why Kodak’s own instructions told you to add half a stop to a grey-card reading. This chapter keeps 18% because it is the common way of describing what the camera assumes — just carry that half stop in your head when you actually meter off a card.

Fujifilm bodies have a physical exposure compensation dial, one of their best design decisions — no menu, just look at the scene and turn it. Get into the habit of glancing at that dial before you shoot and half your exposure problems disappear. (It is also, incidentally, why a recipe cannot store exposure compensation, as the Fujifilm part notes: it is the position of a physical dial, not a menu value.)

Metering modes: how much of the frame gets averaged

There is a second variable inside “the frame averages to”: how much of the frame. That is the metering mode.

Mode	What it measures	When
Multi / evaluative	The whole frame in weighted zones, with some scene recognition	The everyday default; least work when the light is even
Centre-weighted average	All of it, but the middle counts for more	Subject central, background very different in brightness
Spot	Only about 2–3% of the frame at the centre or at the focus point	Stage light, backlight, a dark subject in snow — big range, and you know which end you want

Spot metering is the Zone System’s direct descendant on a modern body: measure one specific thing, the camera places it at mid-grey, and you use exposure compensation to decide how many stops lighter or darker it should actually be. Spot metering without exposure compensation is usually worse than evaluative, because you have handed the whole exposure to whatever occupies the centre.

The division of labour is simple: even light, evaluative; extreme light, spot plus compensation. Centre-weighted sits between them and gets used less than you would expect.

Six symptoms and their causes

The format throughout: the picture shows X → the usual cause is Y → change Z.

1. Snow or white objects come out grey. → The meter assumed mid-grey and this scene is far brighter. → Add +1 to +2.
2. Night scenes and dark clothing come out flat grey with no black. → The same assumption in reverse. → Subtract 1 to 2 stops.
3. An attempt at high key just looks like a mistake. → Highlights clipped while large black areas remain; the histogram has material at both ends. → Add light or a paler background to lift the shadows, moving the whole distribution right instead of only crushing the right end.
4. Low-key shadows are blotchy with noise and colour speckle. → Exposed normally and darkened at home. → Darken at capture with negative compensation so the shadows were always shadows.
5. A sky is uniformly white and no amount of pulling recovers it. → Those pixels are all at maximum and no longer differ. → Nothing to be done now; next time use the highlight warning and reduce exposure.
6. The picture is acceptable but feels thin. → Neither end is reached and everything is bunched in the middle. → Decide whether that is what you want — for fog and overcast it is — and if not, photograph a scene that has both a bright and a dark, rather than adding contrast afterwards.

The Zone System: turning this into a method

In the late 1930s Ansel Adams and Fred Archer, teaching in Los Angeles, worked out a procedure now known as the Zone System. It divides everything from pure black to pure white into eleven zones, 0 to X, one stop apart:

• Zone 0 — pure black, no detail.
• Zone III — a shadow that still shows texture; the folds in dark clothing.
• Zone V — middle grey, which is where a meter puts whatever it measures.
• Zone VII — a highlight that still shows texture; the surface of snow, creases in a white shirt.
• Zone X — pure white, no detail.

The operation is: meter one thing you care about, note that the camera intends to place it on Zone V, decide it should really fall on Zone III or Zone VII, and use exposure compensation to move it there. That is the Zone System. The rest is detail.

Film gave a second control: development time changes the range. Developing less (N−1) pulls the highlights down and compresses the whole scale; developing more (N+1) pushes them apart and expands it. Hence the era’s maxim — expose for the shadows, develop for the highlights — exposure protecting shadow detail, development deciding where the bright end stops.

Digital inverted the first half of that. Because a sensor is linear, because the brightest stop holds half the levels, and because clipping is permanent, the digital version is expose for the highlights and deal with the shadows in processing. Same logic, opposite direction, and the reason is entirely in that table of levels earlier in the chapter.

N−1 development has not disappeared, though. It has only moved. Anything that holds the highlights back and compresses the overall range is a modern N−1 — Fujifilm’s DR200 and DR400, a negative highlight-tone setting, the highlights slider in a raw converter. All of them roll the curve over near the top in exchange for that stretch not hitting the wall. The Zone System’s two controls (exposure places, development ranges) are still two controls; the second one is now a menu item. The Fujifilm part takes DR the rest of the way.

Worth noting where the modern phrase came from too. “Expose to the right” was popularised by an article on Luminous Landscape in 2003, written by Michael Reichmann after a conversation with Thomas Knoll — one of Photoshop’s authors and the person behind Camera Raw. It is not an old darkroom principle dressed up. It is a piece of advice that only makes sense once the recording medium is linear.

The decision table

What you want	What the histogram should look like	How to get it
Full scale	Both ends reached, neither clipped	Soft light, a scene whose range is not extreme
High key	Whole distribution right and narrowed, nothing at the left	Soft light, pale background, +1 to +2
Low key	Distribution at the left, only a small amount at the right	One source, dark background, −1 to −2
Snow that stays white	Right end close to the wall but not on it	+1 to +2, confirmed with the highlight warning
Night that stays black	Material at the left without a solid wall of it	−1 to −2 at capture, not afterwards
A scene too big for the sensor	Both ends clipped	Back to the previous chapter: protect highlights, add light, wait, or merge