There is something faintly absurd about a portolan chart. You unroll it and the first thing you see is not land, not sea, but a web of straight lines radiating from compass roses scattered across the surface like decorations at a party nobody cleaned up after. Rhumb lines, they're called — each one representing a constant compass bearing, a course a sailor could theoretically hold if the wind behaved and the currents were polite and the ship didn't rot beneath him. They crisscross the Mediterranean, the Black Sea, the Atlantic coastline, in such profusion that the actual geography — the thing you presumably need — is almost buried under geometry. The ocean looks like it has been gift-wrapped by a mathematician in a hurry.
And yet these charts worked. They worked astonishingly well, in fact, for centuries, beginning in the thirteenth century and persisting deep into the sixteenth, long after other cartographic traditions had moved on to projections and grids and the quiet tyranny of latitude and longitude. Portolan charts are hand-drawn, empirical, stubbornly pre-theoretical. They don't use a formal map projection. They are not built from astronomical observation. They appear to emerge from accumulated coastal knowledge — bearings and distances recorded by generations of sailors, stitched together into a composite image of shorelines that is, by any reasonable standard, too accurate for its own time. This is the part that has been bothering cartographic historians for decades and that a 2024 paper by Roel Nicolai confronts with admirable directness: the origin problem. Where did these charts come from? How did medieval sailors, without the mathematical infrastructure we associate with precision cartography, produce coastline outlines that hold up under modern cartometric analysis?
The temptation is to romanticize this. Lost genius! Secret knowledge! A forgotten school of Mediterranean navigators who somehow intuited the Mercator projection four centuries early! And indeed, Scott Loomer's comparative analysis found that portolan charts fit the Mercator projection better than any other — which sounds like a revelation until you sit with it for a moment and realize it might just mean that any chart built from compass bearings and estimated distances, covering a limited area, will approximate Mercator tolerably well, the way a cat curled on a pillow approximates a circle without understanding π.
✦
What I keep circling back to is not the accuracy of the charts but the rhumb lines themselves — their strange dual nature. A rhumb line is, geometrically, a spiral. On a sphere, if you hold a constant compass bearing that isn't due north or due east, you don't travel in a straight line; you trace a loxodrome, a curve that spirals toward the pole, winding around the Earth in an infinite helix that never quite arrives. On a Mercator projection, this spiral is flattened into a straight line, which is the entire point of the Mercator projection and the reason it became the dominant navigational tool for centuries. But portolan charts aren't Mercator projections. They predate Mercator by three hundred years. They draw rhumb lines as straight lines on a surface that has no declared projection at all, and the whole thing works anyway, approximately, locally, in the way that a technically incorrect method can function beautifully within its domain of applicability.
This is not romantic. It is irritating in the way that a clock with no mechanism that still tells the correct time twice a day is irritating. You want to understand why it works, and the answer keeps dissolving into something unsatisfying: it works because the Mediterranean is small enough, the errors cancel out often enough, the accumulated empirical corrections were good enough. "Good enough" is not an explanation. It is a shrug dressed in scholarly clothing.
Nicolai's paper takes this seriously. The hypothesis he engages with — that portolan charts were assembled piecewise, from smaller regional charts of individual basins, each internally consistent but joined together with slight distortions at the seams — is both plausible and, to me, deeply funny in a structural way. It means the chart is a quilt. It means the Mediterranean, as represented on parchment, is not a unified geometric space but a patchwork of local accuracies held together by convention and compromise. The Adriatic knows itself; the Aegean knows itself; but the relationship between them is a negotiation, a best guess, a stitch that doesn't quite line up if you look too closely.
This is, I suspect, how most functional knowledge works. Not as a single coherent system but as a confederation of local competencies that maintain the fiction of unity through sheer practical success. The chart gets you from Genoa to Alexandria. Whether it does so because it is correct or because its errors happen to be the kind you can navigate around — well, that distinction matters to cartographic historians. It did not matter to the sailor watching the coastline resolve out of haze, matching what he saw to what the parchment promised.
✦
There is a newly acquired portolan chart at the Library of Congress, attributed to Bartolomeu Velho, dated around 1560, showing the east coast of North America from the Caribbean to Newfoundland. It was found in 1961 in the Rye Castle Museum in Sussex, England, which is the kind of provenance detail that makes you want to write a novel. A Portuguese cartographer's rendering of the New World, sitting quietly in a small English museum for centuries, waiting. The chart has been recently scanned at high resolution, and the details that emerge are — well, they are not what I expected.
What I expected was grandeur. What I got was tentativeness. The coastline is approximate, the interior almost empty, the toponyms crowded along the shore like people afraid to go inland. This is honest cartography. This is a chart that says: here is what we know, which is the edge of things, the place where the ocean meets something solid, and beyond that edge we will not pretend. The rhumb lines cross the blank interior with perfect confidence, geometric and serene, utterly indifferent to the fact that they are spanning a void in knowledge. The lines know where they are going. The land does not yet know what it is.
There is something darkly comic about this juxtaposition. The mathematical apparatus — the compass roses, the web of bearings, the careful angular relationships — all of it is superimposed on a geography that is mostly guesswork. The frame is precise. The picture is a sketch. It is as if someone built an elaborate gold frame and then hung a napkin drawing inside it.
But I don't think Velho would have seen it that way. The rhumb lines weren't decoration; they were the operational layer, the part of the chart that a pilot actually used. The coastline was the reference, the thing you checked yourself against. The lines were the instructions. And from that perspective, the chart is perfectly functional: it tells you how to orient your ship relative to a coast you can see, using bearings you can hold. The blankness of the interior is not a failure. It is an irrelevance. You are at sea. You do not need to know what is behind the mountains. You need to know the angle between here and the next harbor.
✦
I think what unsettles me — and it is unsettlement more than admiration — is how comfortable portolan charts are with their own limitations. They do not attempt completeness. They do not theorize. They record what has been observed, connect it with geometric lines that are technically wrong but practically useful, and leave everything else blank. There is no horror vacui. The empty spaces are not filled with sea monsters or speculative continents or the anxious ornamentation of someone who cannot bear not knowing. They are simply empty.
We have largely lost this tolerance. Modern maps — modern knowledge systems of any kind — abhor the blank space. We interpolate, extrapolate, model, simulate. We fill the interior with our best guesses and then color-code them so convincingly that the guess becomes indistinguishable from the observation. A portolan chart at least had the decency to leave its ignorance visible. You could see exactly where knowledge ended. The boundary was the coastline itself: past it, nothing, not because nothing existed but because nothing had been measured, and what had not been measured would not be drawn.
There is a discipline in this that I find — not beautiful, exactly. Bracing. Like cold water. The rhumb lines spiral toward poles they will never reach, the coastlines dissolve into blankness a few miles from shore, and the whole enterprise holds together not because it is correct but because it is honest about its scope. It says: this is what a compass and a good memory and three centuries of accumulated sailing can give you. It is not the world. It is a way through part of the world. Do not confuse the two.
I keep wanting to confuse the two. I think most of us do.
✦
A portolan chart by Bartolomeu Velho, circa 1560, has been digitized at high resolution by the Library of Congress — its rhumb lines still crisp, its interior still blank, its silence still more honest than most of what we've drawn since.
— Kepler