The rule of thirds works as a beginner's nudge away from dead-centre framing, but not as a law of good composition. When Amirshahi and colleagues measured 727 paintings and a set of highly rated photographs in 2014, their subjects sat on the thirds points no more than in photos that ignored the rule, and beauty ratings did not track the computed thirds score at all. Preference experiments from Stephen Palmer's lab find viewers usually like a single front-facing subject best in the centre, and an off-centre subject best when it faces into the frame.
- 1797: John Thomas Smith coined the rule for sky and land masses, not subject placement.
- 727 paintings measured in 2014 showed no thirds advantage over rule-ignoring photos.
- ρ = 0.17: beauty ratings barely track perceived thirds placement.
- Centre bias: viewers prefer a single front-facing subject in the middle.
- Inward bias: a side-facing subject looks best off-centre, facing into the frame.
- 4.9% of the frame is all that separates the thirds lines from the phi grid.
Every camera app now draws a 3 × 3 grid over the viewfinder, and most photography courses teach the rule of thirds in week one. So it is fair to ask where the rule came from, whether anybody has tested it, and what the tests say a photographer or designer should do instead. This post answers each in turn, with the numbers.
What is the rule of thirds?
The rule of thirds says you should divide the frame into three equal parts horizontally and vertically. The textbook version then gives three instructions:
- place the main subject on one of the four lines, or on one of the four intersections, often called power points;
- put a horizon on the upper or lower line rather than across the middle;
- set a face or a lone tree near an intersection.
On a 6000 × 4000 pixel photo from a 24-megapixel camera, the lines sit at 2000 and 4000 pixels across and at 1333 and 2667 pixels down. On a 1080 × 1920 phone story, they are at 360 and 720 across and 640 and 1280 down. The grid is simple to compute, simple to draw, and simple to teach, which explains much of its spread.
Where did the rule of thirds come from?
In 1797 a London engraver, John Thomas Smith, published Remarks on Rural Scenery, a book of twenty etchings of cottages drawn from nature. In its notes he picked up a remark by Sir Joshua Reynolds from 1783 about balancing dark and light in a painting, a remark that gave no numbers. Smith gave it numbers. "Two distinct, equal lights, should never appear in the same picture," he wrote, and he carried the idea over to landscape: the sky should fill about two-thirds or about one-third of the picture, never half. A ratio of about two to one, he said, was "a much better and more harmonizing proportion". He called it the rule of thirds. Nowhere did he draw a grid or mark four power points. Within fifty years the colour theorist George Field was already warning that the rule universalised a particular case.
Wikipedia: Rule of thirds, history section quoting Smith (1797) and Field (1845)The phrase first appears in John Thomas Smith's 1797 book Remarks on Rural Scenery, a collection of cottage etchings with notes on the picturesque. Smith was extending an idea from Sir Joshua Reynolds's 1783 annotations to Du Fresnoy's Art of Painting, which said that a painting should not split light and dark, or warm and cool, into equal halves.
Smith's rule was about proportions of masses, not about where to put a subject. He argued that a landscape's sky should take roughly two-thirds or one-third of the picture rather than half, and that a long wall or line should be broken at a similar ratio, because two equal, competing parts look undecided.
In 1845 the colour theorist George Field repeated the rule in Chromatics and added that it did not supply a general law. The four "power points" are a later photographic addition that Smith never described.

Does the rule of thirds make photos better? The measurement study
Not measurably. The most direct test is a 2014 study from the University of Jena that compared computer-measured thirds placement with human judgements and beauty ratings across six image sets.
The team computed saliency maps (a model of where the eye is drawn) and measured how much saliency fell in boxes 16% of the frame wide around the four thirds intersections. Image sets: 679 photos labelled as following the rule and 403 labelled as not following it (from Mai et al., 2011), 606 near-random snapshots, 200 simple object photos, 200 highly rated photos from Photo.net, and 727 Western paintings. Thirty participants rated 200 photos for thirds placement and for beauty. Computed thirds scores agreed moderately with human judgements (Spearman ρ = 0.31 for rule-following photos). Beauty ratings correlated only weakly with perceived thirds placement (ρ = 0.17) and not at all with the computed score. Highly rated photos and paintings scored about as low as photos that ignore the rule. Limit: saliency models are only an estimate of where viewers look.
Read the paper (free)See the data as a table
| Mean saliency in thirds boxes | |
|---|---|
| Photos chosen to follow the rule | 0.102 |
| Photos that ignore the rule | 0.067 |
| Portrait paintings | 0.068 |
| Complex scenes with people (paintings) | 0.061 |
| Abstract paintings | 0.059 |
Source: Amirshahi et al., Art & Perception 2014 · GBVS saliency, maximum within four boxes 16% of the frame wide. Only the photo set selected for following the rule stands out; paintings sit at the level of rule-ignoring photos.
What the paintings showed
The painting results are the useful detail. Portraits scored slightly higher than other paintings (0.068 against 0.059 for abstract work), because painters tend to put faces in the upper part of the canvas, near the top third line. That is a placement habit for heads, not evidence of a thirds system. Across all 727 paintings, the saliency peak sat near the centre, shifted a little up and to the left.
What the beauty ratings showed
In the beauty test, the photos selected for following the rule were rated a little higher (0.59 against 0.54 on a 0 to 1 scale, effect size d = 0.36). But those photos also differed in many other ways, and within each group the thirds score did not predict which photos people liked.
"The rule does not supply a general law, but universalises a particular." George Field, Chromatics (1845), as quoted on Wikipedia
Do people prefer subjects in the centre or on the thirds?
For a single subject, most people prefer the centre. Palmer, Gardner and Wickens's 2008 experiments in Spatial Vision found a centre bias for front-facing objects and an inward bias for side-facing ones: a right-facing object was preferred left of centre, so that it looks into the frame.
The direct test of the thirds grid
A 2014 Vision Sciences Society poster from Palmer's Berkeley lab tested the rule of thirds head-on. Participants chose between two pictures of the same fish, dog or eagle at every pair of positions on a grid built from the thirds lines and the centre lines.
The centre was the most preferred position for every object and facing direction, and the bias was strongest for symmetrical, forward-facing objects. Side-facing objects were pushed away from the centre, toward the side opposite their facing direction, which is the inward bias again.
Five experiments with 15 to 36 participants each, using pictures of two or more objects in a rectangular frame. People adjusted object positions to the most and least pleasing places, ranked images and made forced choices. Central configurations were preferred when objects were shown in groups, and compositions that faced into the frame were preferred (the inward bias). But balance around the centre explained only 3.6% to 6.4% of the variance in rankings; how close related objects sat to each other mattered far more, and overlapping placements were almost never chosen. The most-preferred placements replicated almost perfectly between experiments (r = 0.99). Limit: simple clip-art style objects on plain backgrounds, mostly student participants.
Read the paper (free)Is the golden ratio grid better than the rule of thirds?
No. The golden-ratio or "phi" grid puts its lines at 38.2% and 61.8% instead of 33.3% and 66.7%, a shift of 4.9% of the frame. On a 6000-pixel-wide photo that is 294 pixels, and no study has shown viewers prefer one set of lines over the other. The golden-ratio claims themselves mostly rest on retrofitted diagrams, as our post on the golden ratio in design shows.
| Guide | Where it puts the subject | What the evidence says |
|---|---|---|
| Centre | 50% across | Preferred for single, front-facing, symmetrical subjects (Palmer et al., 2008; VSS 2014) |
| Rule of thirds | 33.3% or 66.7% | No advantage in highly rated photos or 727 paintings (Amirshahi et al., 2014) |
| Phi grid | 38.2% or 61.8% | No preference study shows it beats thirds or centre |
| Inward placement | Off-centre, facing in | Preferred for side-facing subjects (inward bias) |
| Grouping | Related objects close together | Explains far more of preference than balance (Leyssen et al., 2012) |
The "diagonal method" that comes up in photography forums proposes placing key points on 45-degree lines from the corners. It has the same weakness: it is a drawing on top of finished pictures, not a tested predictor. Any grid with enough lines will hit something in a good picture.

So why does the rule of thirds seem to help beginners?
Because it fixes a real beginner mistake: putting everything in the middle by default, including horizons, and leaving no room in the direction the subject faces. A centred horizon splits the frame into two equal halves, the very thing Reynolds and Smith warned against, and an off-centre subject facing into empty space reads as having somewhere to go.
The rule is a crude way to get those benefits, and it is still useful as a first habit. Palmer's lab says as much: the rule may help by pushing people toward deliberate, off-centre placements.
The trouble starts when it becomes a checklist item, because then strong central compositions get pushed sideways for no reason:
- symmetrical architecture;
- a portrait looking straight at the lens;
- a temple gopuram;
- a road vanishing into the distance.

Where should you put the subject? A test you can run
Use this to check what you just read. Enter your frame size and see where the thirds lines, the phi lines and the 16% tolerance box from the Jena study land, so you can see how small the differences really are.
Thirds, phi and centre positions for your frame
On a 6000 × 4000 frame the left third line sits at 2000 px and the phi line at 2292 px, 292 px apart, while the scoring box is 960 px wide. The two grids overlap almost completely at the scale people actually perceive.
What should you do instead of the rule of thirds?
Compose by what the picture is about, then check balance. These steps follow from the studies above:
- Decide the subject's direction. If it faces or moves sideways, give it space in front (inward bias). If it faces you and is symmetrical, try the centre first.
- Split masses unequally. Sky against land, light against dark, warm against cool: avoid exact halves unless symmetry is the point. That is Smith's original rule, and it is sound.
- Group related things. Leyssen's results show closeness and relationship drive preference more than balance. Two people talking belong together; don't spread them to two power points.
- Control the background before the grid. Lens choice and depth of field change composition more than a few percent of placement; see focal length and perspective compression and depth of field.
- Crop afterwards. Shoot a little wide and decide in the edit, when you can compare centre, thirds and in-between side by side. Our RAW and dynamic range guide covers keeping the file flexible for that.
For the history of how Indian painters handled space before any of these rules reached them, read the story of Raja Ravi Varma's press. More of my design notes live on the Kala page.
What's new
- Google's Camera Coach on the Pixel 10 Pro uses Gemini to suggest framing and composition before you shoot Google blog, Sep 2025
- Carousel dataset of 277 human-labelled images targets automatic multi-crop composition arXiv / VCIP 2025, Nov 2025
- Palmer lab summary: how the rule of thirds fails in preference tests UC Berkeley, ongoing
Questions people ask
Is the rule of thirds a myth?
The rule exists and is old, but its claim to make pictures better is not supported. Measured across hundreds of good photos and paintings, subjects are not concentrated on the thirds points, and viewers often prefer centred subjects. Treat it as a reminder to place things deliberately, not as a law.
When should you break the rule of thirds?
Whenever the subject is symmetrical or faces the camera: architecture shot head-on, formal portraits, reflections, roads or corridors that vanish to a point. Preference experiments show the centre wins for these. Also break it when two related subjects need to be close together.
Should the horizon always be on a third line?
The useful part is avoiding an exact half, which Smith warned against in 1797. Whether the sky takes a third, a quarter or four-fifths depends on whether the sky or the land is the subject. Many strong landscapes put the horizon very low or very high.
Does the rule of thirds apply to graphic design and layouts?
Only loosely. Layout depends on a grid built from content, margins and type size, not on fixed thirds. Our typographic grid guide explains how to build one from the text up.
Sources
- Amirshahi, S. A. et al.: Evaluating the Rule of Thirds in Photographs and Paintings, Art & Perception 2, 163–182 (2014)
- Leyssen, M. H. R. et al.: Aesthetic preference for spatial composition in multiobject pictures, i-Perception (2012)
- Palmer Lab, UC Berkeley: How the "Rule of Thirds" is wrong (summaries of Palmer, Gardner and Wickens 2008 and VSS posters)
- Smith, John Thomas: Remarks on Rural Scenery (Internet Archive scan) (1797)
- Field, George: Chromatics (Internet Archive scan) (1845)
- Wikipedia: Rule of thirds (history and references)
- Google: How to use Camera Coach on Pixel 10 Pro (2025)
- Loya, R. et al.: Carousel, a high-resolution dataset for multi-target automatic image cropping (2025)
- Photography Stack Exchange: What is the rule of thirds?

