In 1569, Gerardus Mercator published a wall map of the world in eighteen panels. He described his intent in the dedication: to present the world spread on a plane so that “place can be laid down everywhere” and the “longitude and latitude of places correspond everywhere with their true values.” He knew the projection distorted shapes near the poles. He built the distortion in deliberately, because the map solved a specific problem: a sailor could draw a straight line between two ports and read off the constant compass bearing needed to get there.
For transoceanic navigation, the Mercator projection works flawlessly. The distortion is the price paid for the property that matters.
The problem came later.
The Mercator projection became the default world map in classrooms and textbooks worldwide. Used for navigation, it’s correct. Used for comparing the sizes of countries, it misleads dramatically. Greenland appears roughly the size of Africa. In reality, Africa is about fourteen times larger. Countries at high latitudes — Russia, Canada, Greenland itself — appear vastly larger than their actual area. This matters beyond aesthetics. Maps communicate importance. What appears large feels significant.
In 1973, Arno Peters introduced a competing projection designed as a corrective: equal-area, meaning land masses appear in their true relative proportions. Africa occupies the space Africa actually occupies. But the Peters projection distorts shape — countries stretch and compress. Germany looks like a tall rectangle. The debate between Mercator defenders and Peters advocates was technical on the surface and political underneath. Both projections distort. The choice between them is the choice of which distortions to accept.
Neither side was wrong about the facts. They disagreed about which facts mattered most.
Every representation involves this trade.
In statistics: all models are wrong, but some are useful. The assumption of a normal distribution describes no real financial returns exactly. The assumption is false. For many calculations, it’s close enough to be useful. When it fails — extreme events, fat tails — the model fails there specifically. This isn’t a failure of the model. It’s the structure of representation. A complete model of a complex system would be as complex as the system itself, which means it would no longer be a model. Models work by being selectively wrong. The useful question is: wrong in which ways, and what tasks require those exact wrongnesses?
In navigation, there’s a concept called required navigation performance — the level of precision a plane must maintain in given airspace. Open ocean: larger tolerances. Approach to a congested airport: much tighter. The instrument’s acceptable distortions change with what you’re navigating through. The projection shifts with the task.
The problem with familiar representations is that they become invisible.
Mercator spread through the world not because sailors were everywhere, but because it was the most accurate map of the time and familiarity accumulated. Later purposes inherited the tool without re-examining whether the tool fit them. The map’s distortions stopped registering as distortions and started registering as the world. Russia looked enormous because Russia was enormous. The scale bars at the bottom of the map, which would have revealed the distortion, were rarely consulted because the map felt authoritative.
This is what happens to any representation used long enough and widely enough. The frame becomes the reference, not the thing framed.
There is a discipline in making the distortion legible.
Cartographers now publish “Tissot’s indicatrix” diagrams alongside projections — small circles placed at regular intervals across the map. On a perfect representation, the circles would remain circles everywhere. The way they deform — stretching into ellipses, growing or shrinking — reveals the nature of the distortion at each location. A map that includes its own indicatrix is telling you where it lies, and how much.
Most frameworks do not do this. Ideological frameworks present themselves as accurate. Economic models rarely publish their failure conditions alongside their predictions. The mental models we use to navigate social situations — accumulated from experience, refined over decades — almost never specify where they distort.
The more powerful the representation, the more important the indicatrix. A map consulted once can distort without consequence. A map consulted everywhere distorts everything it’s used to navigate.
Mercator built his projection for a specific vessel, a specific problem, a specific epoch of seafaring. The sailors who used it were navigating a world where the distinction between Greenland’s apparent size and actual size rarely mattered — they were crossing oceans, not drawing comparisons between continents. The distortion was irrelevant to their work.
The projection outlasted the context it was built for. The distortions that were acceptable for navigation became distortions applied to questions the map was never designed to answer.
The question to ask of any representation is not whether it distorts. They all do. The question is what the distortion costs, and what work the map is actually being used for, and whether those match.