The sixty-degree cone of vision
Open any book on perspective drawing and somewhere near the front there is a rule: keep the subject within a cone of about 60° from the station point, or the drawing will look distorted.
It is stated with confidence, it is stated everywhere, and it is almost never stated with a reason. This site’s business is measuring the taught constructions rather than repeating them, so: what is the rule about, and is it true?
The thing the rule is nominally about
The phenomenon the rule is pointing at is real and easy to see. A sphere near the edge of a wide frame is drawn as an ellipse, stretched along the radius from the picture’s centre by — 1.31 at 40° off axis, 1.41 at 45°. Faces at the corners of a wide group photograph look broadened; columns at the edge of a wide interior look fat.
The rule says: stop before this gets bad. Sixty degrees across means thirty degrees off axis at the corner, where , and fifteen per cent is judged tolerable.
That much is arithmetic and the figure confirms it: the drawn ellipse’s aspect is 1.309 against a predicted 1.305, the small difference being the sphere’s own angular size rather than a point.
And the thing that is not there
Now the other measurement, and it is the one that changes what the rule means.
Take that same ellipse — the marks on the paper, nothing else — put an eye at the point the picture is correct from, and ask what shape the ellipse subtends.
It subtends a circle. The angular aspect is 1.000000000000, and the spread of the angular radii over 720 points of the ellipse is 4e-13 degrees.
So the stretch is real on the paper and gone in the eye. That is not a coincidence; it is the definition of a correct projection. The sphere subtended a circular cone in the world, the picture is a section of that cone by a plane, and viewing the section from the cone’s apex reconstitutes the cone. The obliquity that stretched the ellipse on the paper is exactly the obliquity at which the eye now views it.
The marginal stretch is not an error. It is what the projection has to do, and it is undone completely by standing in the right place.
Which makes the rule a statement about the reader
If the stretch vanishes from the station point, the rule cannot be about the drawing. So what is it about?
It is about where the reader will be, and the arithmetic is short. A picture whose field of view is is correct from
of its own width. For 60° that is 0.866 widths; for 90°, 0.500; for 40°, 1.374.
A reader holding a book at forty centimetres is at 2.89 times the distance a 60° illustration printed 160 mm wide is correct from. So they are not seeing the marginal stretch — it has vanished, wherever they stand, in the sense that the ellipse is a correct section of a cone. What they are seeing is a scene 2.89 times too deep, which is a completely different error and one the cone rule does not mention.
What the rule is actually doing
Read that way, the rule makes sense as a heuristic, and it is worth reconstructing the sense.
The reader will stand at whatever distance is comfortable, which is roughly a fixed number of centimetres regardless of what is on the page. The picture’s correct distance is a fixed fraction of its width, and the fraction depends on the field of view. So narrowing the field of view moves the correct distance outward toward where the reader actually is.
A 24° illustration 160 mm wide is correct from 37.6 cm, which is essentially reading distance, and reads almost true. A 60° one is correct from 13.9 cm and reads 2.89× too deep. A 90° one is correct from 8 cm and reads five times too deep.
So “keep it under 60°” is a compromise: narrow enough that the depth exaggeration is bounded, wide enough that a scene fits on a page. It is genuinely good advice and the reason it is good advice is not the one given.
The version the books state — the picture will look distorted — is pointing at the marginal stretch, which is not the problem, and it survives because the two failures are correlated. Wide pictures have visible marginal stretch and they are read from much too far back, so a rule that limits the first also limits the second, and nobody has needed to work out which one they were avoiding.
Where the rule fails, and where it costs
Two consequences of getting the mechanism right.
A wide picture viewed correctly is fine. A 90° photograph shown large and looked at from half its width away is a correct projection and reads as one; the marginal stretch subtends circles. That is why a cinema screen at the front rows, an immersive projection, or a large print approached closely can carry a much wider field than the rule allows without any of the trouble the rule predicts.
A narrow picture viewed wrongly is not. A 20° photograph on a phone held at 30 cm is being read from about a third of its correct distance, and depicts a scene a third as deep. That is the “compression” people describe in telephoto pictures, and the cone rule has nothing to say about it because it is below the threshold the rule watches.
So the rule has a failure at each end. It forbids something that is fine if shown properly, and permits something that is wrong for a reason it does not track. Both failures come from a rule about the picture standing in for a rule about the reader.
The measurement, and why it is a fit
There is a methodological point in the figure worth extracting, because the obvious way to make this measurement is circular.
The tempting version: take the sphere’s silhouette directions, image them, and then ask what angle each image point subtends at the station point. That returns the direction it came from — by construction, since imaging a direction and then reading the direction back is the identity — and the check restates its own derivation.
So the measurement is a fit instead. The drawn ellipse is taken as marks on paper; the direction from the station point to each mark is computed; and the circular cone that best fits that set of directions is found by solving a smallest-eigenvector problem, the same solver the conic fit uses. If the drawn ellipse really is the image of a circular cone, the fit is exact and the residual spread is zero.
It is: 4e-13 degrees over 720 directions.
That is the kind of care this site has had to learn twice. A cross-ratio taken over four consecutive divisions gives the equal-steps-by-eye method a perfect score, because four points equally spaced in the picture have the same cross-ratio as four equally spaced in the world. A conformality test taken along a surface’s own coordinate directions gives the cylinder a perfect score, because those two directions happen to stay perpendicular. Both were necessary conditions evaluated at the one input where they cannot fail. This would have been the third.
The centre that is not the centre
The fit produces something extra that is worth noticing.
The axis of the fitted cone is not the direction to the drawn ellipse’s own centre. At 40° off axis, in the figure’s own magnified drawing, they are 3.42 px apart. That is this site’s oldest result, seen from the viewing side: the image of the centre and the centre of the image are different points.
Which matters here for a practical reason. Anybody measuring the marginal stretch by drawing a box around the ellipse and taking its axis ratio is measuring about the box’s centre, and the box’s centre is not the sphere’s direction. The measurement is close enough for a 1.31 : 1 claim and it is not the same quantity, and the difference grows as the field widens.
Following the rule’s own arithmetic through
It is worth doing the sum the books do not, because the answer is concrete enough to check against a shelf of them.
A textbook illustration is typically reproduced about 120 mm wide on a page held at 35 to 40 cm. If it is drawn at the recommended 60°, it is correct from 10.4 cm — and the reader is at three and a half times that.
Meanwhile the same books show the construction: a plan with the station point marked, the cone of vision drawn at 60°, and the picture plane crossing it. In those plans the station point sits at 0.87 of the picture’s width from the picture plane, which is exactly right and is the number the rule encodes.
So the books contain the correct number twice — once in the rule and once in every construction plan — and never mention that the reader of the printed page is not at it. The construction is drawn to be understood rather than to be stood at, which is fair; what is missing is the sentence saying that the printed reproduction has its own, quite different, correct distance.
That gap is the reason this site computes the viewing distance on every figure rather than only in the essays about viewing. A number that appears in a construction and never in a caption is a number the reader cannot use.
The other taught rules this joins
This site’s wrong field is a small collection of confidently taught constructions put through a measurement, and it is worth saying where this one lands among them.
The two-point cube turns out to depict a cube exactly when the two by-eye placements are symmetric, and something else entirely when they are not — eight points of asymmetry gives a box 1.39 times shallower than wide. The construction is fine and the step it leaves to judgement is what decides the answer.
The by-eye depth methods — equal steps, halving, tapering — misplace a post by more than a quarter of a metre each, and the natural test for them scores all three perfectly because it is a necessary condition evaluated where it cannot fail.
The constant-ratio workshop rule for a pavement’s bands drifts, and by the sixth band claims more than twice the depth it should.
The cone of vision is different from all three, and better. It is not wrong. Its reason is wrong, and the rule survives because the reason and the real mechanism happen to point the same way over the range where anyone uses it. That is the most interesting failure mode a taught rule can have, because it is the one that will mislead exactly when the conditions change — a wide picture shown large, a narrow picture held close — and it will do so with three centuries of authority behind it.
What to do instead of the rule
If the rule is really about the reader, the honest replacement is to say so, and it fits in a sentence.
State the distance the picture is correct from. It is one number, computable from the field of view and the width the picture is shown at, and it costs a line. Every figure on this site carries it, which is the site’s one non-negotiable piece of furniture, and it turns a rule of thumb into a fact the reader can act on.
With that number available, the illustrator’s decision changes shape. A wide field is not forbidden; it is a commitment to showing the picture large or to accepting a stated depth exaggeration. A narrow field is not automatically safe; it comes with its own factor, in the other direction, at ordinary reading distance.
And the sixty-degree cone becomes what it always was: the field of view whose correct viewing distance is 0.87 widths, chosen because at book sizes that is not too far from where a reader’s head happens to be.
What a wide picture does cost
Nothing above says a wide field is free, and it would be a poor essay that left that impression. Three costs are real and none of them is the one the rule names.
The picture surface runs away. A flat picture’s half-width is , which is unbounded: a 170° flat picture is impossible at any finite size, and a 140° one is enormous relative to what it shows. That is a hard limit on the flat surface, not a matter of taste, and it is why very wide views are cast onto cylinders or spheres instead.
Resolution goes to the edges. A flat projection spreads the outer parts of the field over more paper than the middle, which is the area scale growing as . On a sensor that means the corners get more pixels per solid angle than the centre — a strange way to spend a fixed number of them, and the reason a very wide rectilinear lens produces a picture whose subject occupies a small central patch.
And the correct viewing distance becomes unreachable. A 90° picture 160 mm wide is correct from 8 cm, which is inside the near point of most eyes. There is no position from which such a picture can be read correctly at that size, so the depth exaggeration is not merely likely but unavoidable.
Those three are the honest case against a very wide picture, and only the third has anything to do with the reader’s position. None of them is what a book means by “it will look distorted”.
A note on the rule’s provenance
One historical remark, offered as context rather than as scholarship.
The 60° figure is often justified by an appeal to the field of sharp vision — the fovea, the cone of attention, the angle over which the eye takes in detail without moving. Those are real quantities and they are all much narrower than 60°: foveal vision is a couple of degrees, and the region of comfortable attention perhaps 20°.
So the rule’s usual justification does not survive its own numbers either. Whatever 60° is, it is not the angle of sharp vision, and the fit between the two is loose enough that the appeal is decorative.
What 60° is — a viewing distance of 0.87 widths, which is roughly where somebody looking at a whole sheet of paper ends up — is a better explanation, is checkable, and has the advantage of predicting when the rule should be relaxed. Which is the test this site applies to any taught construction: not whether it works, but whether the reason given for it is the reason it works.