How To Draw Optical Illusions

Getting started with impossible shapes

Optical illusion drawing is mostly about tricking your brain into accepting something that can't exist in three dimensions. You draw lines that suggest depth, intersection, and perspective, but you deliberately break the rules of geometry at key points so the viewer's eyes keep trying to resolve the contradiction. The foundation is the impossible triangle, also called the Penrose triangle. Start with a regular triangle on paper. Then instead of connecting the corners directly, extend each corner into a beam or bar. The trick is that each bar appears to connect to the next one in a loop, but if you try to measure it, one connection is always slightly wrong. When I first tried this, I spent two hours erasing and redrawing because I couldn't get the angles to work. The problem was that I was drawing each corner separately and then trying to join them. I ended up constructing the entire shape as one continuous outline, measuring angles with a protractor, and only then cutting out the internal lines. That gave me a version where the impossibility felt convincing. Another common starting point is the impossible staircase, the Penrose steps. You draw a series of connected rectangular blocks that form a U-shape, but the heights don't add up consistently across the corners. Each section by itself looks like normal perspective. At the junction where two sections meet, one step appears to be rising while simultaneously appearing to fall. The workaround I use now is to build the drawing in three distinct layers: first sketch the overall path without any depth cues, then add shading to one side, then introduce the contradictory depth markers. Building it backwards that way makes it easier to spot where the geometry actually breaks.

The mechanics behind why this works

Your brain uses a set of assumptions when interpreting line drawings as three-dimensional space. Parallel lines converge toward vanishing points. Light comes from above, so shadows fall downward. Adjacent surfaces share consistent orientation. Optical illusion drawings violate one or more of these assumptions at a single junction point, and the brain spends several seconds cycling through possible interpretations before giving up. The most important technical skill here is perspective inconsistency. You are not drawing randomly. You are drawing with proper one-point or two-point perspective locally, then introducing a deliberate conflict at the merge point. A proper perspective grid helps enormously. I draw a faint grid before I start any serious piece, which keeps the local geometry honest and makes the intentional breaks much more visible when I review them.

Practical steps for a basic impossible cube

I typically begin with a standard isometric cube. Draw a hexagon, then connect every other vertex to create three visible faces. That gives you a solid three-dimensional box. Now take each outer edge and extend it past the cube in a direction that conflicts with the existing faces. For example, extend one edge straight outward as if it continues the top face, while extending another edge as if it continues the front face, and then connect those extensions to a new vertex that does not align with either plane. The result is a cube that has three visible corners meeting at points where only two should be possible. Shading matters more than people expect. Without it, the image reads as a confusing set of lines. With directional shading applied to specific faces, the brain locks into a 3D interpretation and the impossibility becomes striking. I usually shade the top faces darker, the side faces medium, and leave the front-facing planes light. This creates a light source from above that reinforces the perspective illusion.

Common mistakes beginners make

The biggest issue is that people try to draw the impossible part first. They start with the contradictory junction and then try to build outward. That almost never works because there is no consistent geometry to anchor from. The second biggest problem is over-shading. Heavy shading on every surface creates visual noise that distracts from the structural contradiction. The third is inconsistent vanishing points. If one part of your drawing uses a one-point perspective and another part subtly shifts to two-point, the brain detects the mismatch and the illusion collapses before the impossibility even registers. A more refined version of the impossible triangle uses a continuous gradient along each beam. The gradient should shift direction at each corner so that it appears consistent with a single light source, but when you trace the gradient around the full loop, it becomes impossible. I recommend using a soft pencil or charcoal for this because the midtones blend better than hard graphite lines. The transition between light and shadow at each corner is where the brain accepts the drawing as coherent, so those transitions need to be smooth and confident. Smudged or hesitant gradients undermine the effect immediately. This approach usually takes about twenty minutes for a clean finished piece if you work from a prepared grid. Without a grid, expect forty-five minutes or more because you will spend most of the time adjusting junction points.

When this method doesn't work

Impossible shape drawings rely on flat presentation. If you photograph the drawing from an angle, the whole thing falls apart. The same applies if you view it through a curved mirror or projection surface. The illusion only holds under direct frontal viewing conditions, which limits where you can effectively display these pieces. Some artists work around this by creating ambiguous figures rather than strictly impossible ones, since those survive minor perspective distortions. An ambiguous figure allows multiple valid interpretations instead of one contradiction, which is a different technique entirely but often more practical for reproduction. If you want to experiment further, digital tools like Procreate or Krita let you duplicate layers and compare multiple perspective grids simultaneously, which speeds up the trial-and-error phase considerably. The grid layer method alone cuts debugging time roughly in half compared to working freehand on paper.