Seeing: How a Picture Is Built
Composition: Sending the Eye Where You Want It
Composition is not a list of rules. It is a set of tools for controlling where a viewer looks first, and where they go next.
Nobody takes in a photograph all at once. The eye lands somewhere, jumps somewhere else, jumps again, and then decides whether to keep going. Composition is the arrangement of that route.
So when you are judging a photograph, do not ask whether it obeys the rule of thirds. Ask three questions:
Where did I look first? Where was I taken next? Is there anywhere I got stuck, or slid out of the frame?
The rest of this chapter explains why eyes move the way they do, and which part of that route each compositional tool actually changes.
The eye moves in jumps
Understanding composition starts with a piece of physiology: your eye does not sweep smoothly across a picture.
High-resolution vision exists only in a small patch at the centre of the retina, the fovea, covering something like one to two degrees of your visual field — about the size of your thumbnail at arm’s length. Everything outside it is coarse, good enough for movement, brightness and rough shape. Seeing therefore consists of flicking that small sharp patch from place to place in a series of very fast jumps called saccades, resting two to three hundred milliseconds at each stop, with the travel between them edited out of your experience.
The impression of taking a picture in at a glance is a reconstruction. What actually happened was four to six fixations, and composition determines where those fixations land and in what order.
So where does the first one go? Roughly in this order of priority:
- The strongest luminance contrast wins. This is why a blown-out patch of sky in a corner is so destructive: it has taken the first fixation before you have even recognised what the picture is of.
- Faces beat almost everything. Dedicated machinery in the brain handles faces, and a legible face will be the first or second stop even when it is small, dark and off to one side.
- Text has a similar privilege. Signs, street names, lettering on clothing — if it can be read it will be read, and reading is an expensive, lingering fixation. Text you do not want read is noise.
- Lines decide where the next jump goes. Having landed, the eye tends to continue along whatever continuous structure is most obvious.
- Wherever a person in the picture is looking, viewers look too. Gaze following is a social reflex, and it is strong enough to work from a profile, or from the back of a turning head.
The first fixation also has a default: the middle of the frame. With nothing else to go on, the first glance lands near the geometric centre. Hold on to that, because the next section depends on it.
One qualification, from the eye-tracking work that established all this. When Alfred Yarbus recorded viewers looking at a single painting in the 1960s, the paths they took changed completely depending on the question he had just asked them — estimate the family’s wealth and they toured the furniture; guess the ages and they toured the faces. Composition sets the defaults, but a viewer who arrives with a question of their own will overrule you. That is an argument for making the intended subject unambiguous, not merely well placed.
Leading lines: the most direct tool
Any line — a road, a river, a railing, a shadow, a row of trees — drags the eye along it. It is the most useful and most forgiving tool in composition, because it barely ever fails: put a clear line pointing at your subject in the frame, and viewers will follow it.
Why so reliable? It exploits two things from the previous section at once. A line has strong edge contrast, so it is easy to find; and saccades tend to continue along continuous structure, so the line answers the question where next on the viewer’s behalf.
Only one thing needs watching: the line has to point at something. A line that runs out of the frame with nothing at the end sends viewers out with it, and they generally do not come back.
This frame demonstrates the pairing completely. The line moves the eye; brightness holds it. That patch of light at the end is the highest-contrast region in the picture, so arriving there is arriving somewhere. Remove it and the same two ranks of trees become a slide that tips the viewer out of the frame.
Which reduces to one habit on location: find the line, then look at what is at the end of it. If the answer is nothing, move until it collides with something, or give up on that line.
There is also a kind of line that is not there at all. Three similar objects in a row, the tops of a queue of heads, two people looking at each other across a street — no physical line exists, and the eye connects them anyway. Implied lines behave exactly like real ones and are far more common: railings and roads are scarce, but things that can be joined up are everywhere. Test them the same way — see where the imaginary line points, because it will lead viewers out of the frame just as efficiently.
The rule of thirds: separating the first glance from the subject
Divide the frame with two horizontal and two vertical lines and put the subject on a line or an intersection. It is the most-quoted rule in composition and the most commonly mistaken for an aesthetic law.
It does exactly one thing: it moves the subject off the spot the first glance lands on.
Return to the last point of the mechanism section: with no other cues, the first fixation goes near the centre. If the subject is in the centre, the first stop is also the last stop — the eye arrived, the business concluded, there is no route, and the picture reads as static. Move the subject onto a third and the first glance lands on empty space, after which a jump is required to reach the subject. That jump is the smallest possible route. The extra fifth of a second is the entire content of “it holds the attention better”.
Once the mechanism is clear, two things follow:
- There is no magic in those particular lines. A subject on a third and a subject at a quarter behave much the same; what matters is not being dead centre. So do not shuffle the frame around to align things with a grid.
- When to ignore it now has an answer too. Centre the subject when stillness is the point: symmetrical architecture, a formal portrait, a sun on the axis. Those pictures want the first stop to be the last stop.
One side effect gets overlooked: moving the subject off centre empties two thirds of the frame. That space has to be worth looking at, or be deliberately empty — otherwise you have swapped a static subject for a large boring region.
Symmetry: stability as the content
Symmetry is the inverse of the thirds: it manufactures complete stability on purpose. Done well it produces order and sometimes solemnity. Done badly it is merely inert.
The difference is usually precision. Symmetrical frames are brutally sensitive to level and placement — two degrees off looks like a mistake, because the viewer is comparing the halves point by point and will catch any disagreement. That also explains why symmetry rewards a long look: it turns the viewer into a proofreader, and proofreading is slow.
So use the electronic level when shooting symmetry — the X-T5 and X-E4 both have one, and it is worth adding to the DISP cycle — and put the axis on the centre of the frame rather than approximately there.
Frames within frames: cutting out the subject with composition
Wrapping the subject in something already in the scene — a doorway, a window, branches, an arch — draws a second frame inside the photograph. It does two things: it concentrates the eye, and it adds depth.
It works differently from the tools above. Leading lines and the thirds answer where to go; a frame within a frame answers where not to go. A dark, closed border shuts the surroundings out so the eye has nowhere else to be.
That makes it valuable in exactly the situations where nothing else helps. When the background cannot be cleaned up, find something to frame the subject with and you have cut it out with composition instead. Three conditions in practice:
- The inner frame must be darker than the subject, or the frame itself takes the first fixation. The rock above is a backlit shadow side, which is not a coincidence.
- It should enclose on three sides or more. A single edge is not a frame, it is a pillar.
- Keep the subject clear of the frame, because touching turns the two into one shape.
Negative space: room to breathe
Photographing the subject too large is the commonest compositional fault there is. The cause is that your attention was already on the subject, so subjectively it loomed. The camera has no idea what you were attending to and records what is there.
The cure is procedural: take the frame, step back two paces, take it again, and compare at home. You will choose the more distant one more often than you expect.
In terms of the route, negative space clears the board: it reduces the number of possible fixations to one. It is the most extreme form of eye control available, which is also why it reads as quiet, lonely and vast — not because of the subject matter, but because the eye genuinely has nowhere else to go.
Six symptoms and their causes
The format throughout: the picture shows X → the usual cause is Y → change Z.
- The picture feels off and you cannot say why. → Something in a corner is competing for the first fixation. → Check the four corners in order and move anything too bright, too busy or readable out of the frame.
- Viewers say they do not know where to look. → Two or more fixations of equal weight with no line joining them. → Move closer so one grows, or change angle so one is hidden.
- The eye slides out of the frame and does not return. → A leading line points at the edge with nothing at the end. → Shift position until the line meets a subject, or block its exit.
- A symmetrical picture looks like a slip. → A degree or two off, or the axis not on centre. → Electronic level on, and align the axis with the centre line.
- The subject is on a third and the picture is still dull. → The thirds only guarantee the subject is off centre; the remaining two thirds are still empty and uninteresting. → Give that space content — foreground, shadow, some environmental cue — or commit to it as negative space.
- A portrait subject looks out of the frame and viewers follow them out. → Gaze following has escorted the audience off the picture. → Turn the subject’s gaze inward, or leave space on the side they are looking towards.
The last one is the most frequently violated and the easiest to fix: leave room on the side someone is looking towards. On location it costs half a step.
Where these rules came from
“The rule of thirds” sounds like a mathematical truth. Its actual origin is decidedly casual.
The phrase first appears in 1797, in Remarks on Rural Scenery by the English painter and engraver John Thomas Smith. He was not discussing where to put a subject. He was discussing the proportion of light to dark areas in a landscape, proposing roughly two parts to one, and he called that idea the rule of thirds. It is a recommendation about tonal balance. It has nothing to do with the intersections of a grid.
The grid attached itself later, and so did something worse. Through the second half of the nineteenth century, following Adolf Zeising’s writing in the 1850s, the golden ratio was retrospectively discovered in the Parthenon, in Renaissance painting and in the human body — mostly by choosing which measurements to take. Photography inherited that habit wholesale, which is why compositional advice still arrives with φ, Fibonacci spirals and an implication of natural law. There is no evidence that any of the photographers whose work is held up as proof were using it.
That is worth remembering not as a correction of terminology but as a reminder: these tools are descriptive, not prescriptive. Somebody noticed that a thing tended to work and gave it a name. The pictures came first.
Frames within frames have a cleaner pedigree. Seventeenth-century Dutch painting had a word for it — doorkijkje, a small view seen through a doorway. Pieter de Hooch built interior after interior on it: a dim room in front, a door in the middle, a bright courtyard or street beyond. The structure is identical to framing a distant peak with an arch, and so is the mechanism — dark surround, bright content, isolated subject.
Diagnose first, then choose a tool
Do not carry five rules around and apply them to everything. Run the sequence the other way:
- Who is this picture about?
- Where will a viewer’s eye land first? (If that is not the subject, you have a problem.)
- What is competing for attention — a bright corner, a busy background, a half-cut figure, readable text?
- Which tool solves that?
| The problem is | Tool | Why it works |
|---|---|---|
| The eye does not know where to go | Leading line | Answers “where next” on the viewer’s behalf |
| The picture is static | Rule of thirds | Separates the first glance from the subject, forcing a jump |
| The background is noisy and cannot be cleared | Frame within a frame | A dark border shuts out competing fixations |
| Too many fixations fighting each other | Negative space | Reduces the possible stops to one |
| The frame feels loose and disordered | Symmetry | Turns browsing into proofreading |
| The subject is too small to read | Walk closer | Changes perspective and relationships, not just size |
When you may break them
Every rule exists to solve a problem, which gives “may I break it” a mechanical answer: when that problem is not present in your picture.
- Centre the subject when stillness and frontality are the point — symmetrical architecture, a formal portrait, someone looking straight down the lens. The thirds solve “too static”, and you do not want that solved.
- Let a line run out of the frame when leaving is the subject — a figure walking away, a boat heading out. Pointing at the border stops being a leak and becomes the content.
- Push the subject hard against an edge when unease or urgency is what you are after. As the previous chapter argued, a decisive cut carries tension of its own.
- Fill the frame with no negative space at all when the subject genuinely is too much — a market, a crowd, a stack of signs.
All four share a structure: first say what the tool solves, then say why you do not need it solved. Without that, it is not breaking a rule, it is not having noticed one — and the two are indistinguishable in the picture. The difference shows up in whether you can do it again on purpose.