In How to Draw by Scott Robertson, they say that to draw an ellipse within a plane, you find the minor axis by drawing the line perpendicular to the plane. You do this by using another vanishing point and drawing through the plane's center point.

enter image description here

The problem is that when I draw an ellipse that touches all four points, I can't seem to get the minor axis to cross the plane's center like in the book. Does this only work when the plane is a perfect square? How do you draw an ellipse in a non-square plane? Also, won't the minor axis change depending on where the second vanishing point is? When I use other vanishing points the ellipse doesn't look right.

enter image description here

I'm confused about how to draw ellipses in perspective. Is there a big difference between this method of simply dividing the plane several times to know where the ellipse should be?

3 Answers 3


You have a misconception. The ellipse is a way to build a perspective image of a rectangular box. The ellipse is the same perspective view of a circle on a face of that box. You should build the perspective image of a rectangular face of a box around the ellipse, you should not try to crunch an ellipse into an already drawn 4-gon. That's because you can draw any ellipse inside a 4-gon and it will be a perfect perspective image of another ellipse.

In case your ellipse should be a perspective image of a circle on a square the question is valid. See NOTE1 and read my story to learn how the "Chord method" works.

The useful story would tell how one could construct a consistent cube if the horizon line and the perspective image of a circle on one of the faces is given. The story is a kind of advanced eyeballing for practical drawing on paper, it's not exact math. It doesn't start from the measures of the cube and imaging parameters, it starts from the horizon line and one ellipse.

In the next image we have the blue horizon line and an ellipse which presents a circle on one face of a cube. For simplicity we assume that the ellipse is actually a circle on a vertical face. In addition we assume 2 point perspective is good enough, so vertical lines are imaged as vertical:

enter image description here

The theory says that the minor axis of the ellipse points to the vanishing point of those horizontal edges of the cube which are perpendicular with the face of the ellipse (see NOTE2. We find the vanishing point (V) by extending the minor axis of the ellipse. Its easy in Illustrator which places nodes at the ends of the ellipse axles:

enter image description here

As already said we assume the ellipse is on a vertical face of the cube and we use 2 point perspective. The next image has the great eyeballing part:

enter image description here

The green lines are nothing special. They are all vertical lines (as we assumed vertical lines can be drawn as vertical lines) One goes through the midpoint of the ellipse and the other two are tangent lines. The great trick is to find the other vanishing point. We draw from an arbitary point A of the ellipse curve a straight line towards the horizon line so that the half segments ABD and CBD look visually as big. The line meets the horizon line at the wanted vanishing point U.

In drawing guidance texts this is the "Chord Method". The chord is the line AC.

Now we can draw a part of the edges of the face squares from the vanishing points and crossings:

enter image description here

The black lines are eyeballed tangents from U and the orange lines are drawn from V to the crossings.

We need still 2 lines to get a cube. One possibility is to use again an ellipse. We can eyeball one with vertical minor axis. It's the image of the circle inside the horizontal top square:

enter image description here

An eyeballed tangent from U and a vertical line from the crossing completes the cube.

enter image description here

I guess eyeballing sounds bad. I used Illustrator which should have precise snapping and I didn't utilize it except in the crossings. But: Chord method is designed for persons who draw with pen and paper. Some of them can also have a ruler and an ellipse drawing tool.

NOTE1: Chord method can also be used reversely to decide the lengths of the of the axles of the right ellipse (which is the image of a circle) if the vanishing points are already given.

You can find a longer description of the chord method from here: https://www.idsa.org/sites/default/files/2002_Randy%20Bartlett.pdf

Beware the "from math taken" but exact convention to define the direction of the planar surfaces by the directions of the surface normals. There the floor of an usual house is vertical and the walls are horizontal.

NOTE2: There are later comments which say that this rule is not a general perspective imaging fact. I do not know enough of perspective imaging math to say how accurate the rule is. So, take it as a practical way to find one of the possible images of possible fitting cubes, that one which produces an even looking result.


You are right, and Scott Robertson is wrong. It is a very common misconception that the minor axis always lies along the "axle" perpendicular to the plane of the ellipse. (And just because he repeats it in capital letters in his book, that doesn't make it true!) Incidentally, Marshall Vandruff similarly gets it wrong in his 1994 course, although he has excellent advice overall. The "Chord Method" posted by @user287001 in the other answer is also wrong. The only reason it appears to work in that case is that the objects in question are photographed in the centre of vision.

The "axle" principle should be considered more of a rule of thumb, and it's actually a pretty good approximation when the object is within the cone of vision. In the example you gave, note that there's some significant distortion happening because you're outside the cone of vision, and this shows exactly where the rule of thumb no longer applies.

Here's a picture taken from Wikipedia's article on Columns:

Columns of an ancient ruin shown in perspective

I've overlaid perspective circles on the tops of the columns; according to the "rule", the minor axes of the ellipses should be pointing straight down along the length of the columns, but they clearly do not. Note that the ellipses on the left side of the photo, being closer to the centre of vision, do point downward.

  • About the chord method: It's not wrong, it finds just that cube seen from that station point where the minor axis of the ellipse is directed to one of the vanishing points of the cube edges.
    – user82991
    Oct 16, 2021 at 12:19
  • @user287001 The Chord Method starts by aligning the minor axis of the ellipse with a vanishing point, which is a common mistake. It is only a reasonable approximation when the ellipse is near the centre of vision.
    – Théophile
    Oct 16, 2021 at 14:38
  • The station point nor the circle were NOT given, only the imaging plane, an ellipse (as the image of an unknown circle) and the horizon line. were there. Your claims make sense if you can show that there exists no such circle and no such station point which allow the continued minor axis meet the vanishing point of the normals of the circle.
    – user82991
    Oct 16, 2021 at 15:18
  • @user287001 I had supposed that the vanishing point indicated in the diagram was the centre of vision, in which case we can of course reconstruct the station point. But that's just a detail; the main point is that in general, the normals through perspective circles do not line up with the minor axis of the corresponding ellipse. I am curious what you make of the picture of the columns: how would the Chord Method apply in that case?
    – Théophile
    Oct 18, 2021 at 2:35
  • 1
    For anyone interested, this is a method to construct the minor and major axis. math.stackexchange.com/questions/3820517/…
    – cdosborn
    Oct 7, 2023 at 5:50

I've learned a little bit since posting this question. About the minor axis intersecting the square's center point, in Marshall Vandruff's 1994 class on perspective, he mentions that while the minor axis will come close to intersecting the center point, it does not necessarily do so.

On the point of the minor axis changing depending on where the vanishing point is placed, I have learned that the minor axis ellipse technique only works for perfect circles in perspective. So if the ellipse does not fit into the rectangle, it is because the rectangle is not a perfect square. When you move two vanishing points together, it changes the FOV and therefore changes what a perfect square looks like in perspective, which in turn changes the minor axis of your ellipse.

As for how to place an elongated ellipse (not a perfect circle) in a rectangle (not a perfect square) in perspective, I imagine the best method would be to simply divide the rectangle into several parts, or to first use the technique to place a perfect circle and then stretch the result.

  • My recommendation is not to try to precisely find the axis at all (it's really too complicated, except in special cases), but instead to find several points that lie on the ellipse as well as the tangents through them. Then you can essentially connect the dots.
    – Théophile
    Oct 16, 2021 at 4:07

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.