Made for One Position

There is a marble disc set into the floor of Sant'Ignazio in Rome. Stand on it and the church has a dome. Step off it and the dome shears apart, because there is no dome. There is a flat canvas thirteen metres across, painted by Andrea Pozzo in 1685 because the money for an actual dome had run out in 1642 and never came back.

The disc is the interesting object, not the painting. Somebody had to decide where the correct place to stand was, and then mark it in stone, permanently, so that four hundred years of visitors would know that the work has one viewport and this is it.

Every screen has that problem and none of them have the disc.

Andrea Pozzo, the nave ceiling of Sant'Ignazio, Rome, about 1685 to 1694. Photograph by Amchupa, CC BY-SA 4.0, via Wikimedia Commons.

Four ways to build a picture that only works from one place

Pozzo, and the honest fake

There are two floor markers in Sant'Ignazio, not one: a second, further east, is the optimum point for the fake dome specifically. The painted architecture is calculated from those points and from nowhere else. Stand under the crossing and, as one guide puts it, the dome simply ceases to exist.

Two details make this better than it first looks. The first is that the fake dome was a budget decision, openly. The Jesuits could not afford the structure, so they bought the appearance of one. The second is that what hangs there now is not Pozzo's canvas at all. His burned. Francesco Manno reproduced it faithfully in 1823, working from Pozzo's own drawings. The illusion survived its own material.

Masaccio, and the eye level of a standing adult

Masaccio, Holy Trinity, about 1426 to 1428, Santa Maria Novella, Florence. 667 × 317 cm. Public domain, via Wikimedia Commons.

Two and a half centuries before Pozzo, Masaccio put the vanishing point of the Holy Trinity at the eye level of a person standing in the church, with the horizon running across the floor between the two donors. The fresco is nearly seven metres tall and the illusion is calibrated to a body at a specific height on a specific floor.

I looked for a published viewing height in metres and could not find one, which is itself worth saying. The qualitative claim is everywhere and the measurement is nowhere. Brunelleschi is usually named as the source of the geometry, and there is no concrete evidence of his direct involvement either. Some of the most repeated facts in perspective's origin story are inherited rather than checked.

Borromini, and the corridor that lies about its own length

Francesco Borromini, the forced-perspective colonnade, Palazzo Spada, Rome, 1653. Photograph by OrbitalNomad, CC0, via Wikimedia Commons.

Borromini built a colonnade for Cardinal Spada in 1653 with diminishing rows of columns and a rising floor, so that about eight metres of real corridor reads as something between thirty and thirty-seven, and a sculpture at the far end that is well under a metre reads as life-size.

The numbers deserve a flag. Sources give the apparent length as thirty metres and as thirty-seven; the statue as under fifty centimetres and as sixty. The Italian state museum's own pages are closed to automated reading, which is where the authoritative figures presumably live. I am giving you the spread rather than picking one and sounding confident.

Borromini did not do the geometry alone. The calculations were worked out with an Augustinian priest and mathematician, Giovanni Maria da Bitonto.

The Ames room, and the removal of every other cue

The Ames room at the Cité des Sciences, Paris. Photograph by mosso, CC BY 2.0, via Wikimedia Commons.

Adelbert Ames Jr. patented the distorted room in 1940. It is an irregular hexahedron with one corner much further away than the other, built so that from one point it projects the retinal image of an ordinary rectangular room, and two people standing in opposite corners appear to be wildly different sizes.

The peephole is the design. It does three things at once: it fixes the viewer at the single correct position, it restricts vision to one eye so stereo depth is unavailable, and it prevents the head movement that would produce motion parallax and give the game away.

The room does not create an illusion. It removes every channel of evidence except the one that has been rigged.

That is the purest statement of this whole theme, and it is the one I find most uncomfortable to sit with, because removing channels of evidence is a thing interfaces do constantly and usually without intending to.

And the one that argues with itself

Hans Holbein the Younger, The Ambassadors, 1533. Oil on oak, 207 × 209.5 cm. The National Gallery, London, NG1314. Public domain, via Wikimedia Commons.

Holbein's double portrait is the case that does not resolve, and that is why it is last.

Two men stand either side of a shelf loaded with instruments: a celestial globe, a terrestrial globe, a polyhedral sundial, a torquetum, a lute, a case of flutes, a Lutheran hymnal, a merchants' arithmetic book. Everything on that shelf is a device for measuring the world accurately. Across the floor at their feet lies a grey smear that resolves, from one position, into a human skull.

The painting therefore contains two mutually exclusive viewing positions. From the front, you get the instruments and the men, and the skull is an unreadable stain. From the position where the skull is correct, everything else is unreadable. There is no place to stand from which the picture is entirely legible, and that is not a defect. That is the content.

The National Gallery's own public page says the skull can only be seen properly if you look up at the painting from the bottom right corner. The scholarly literature gives a two-ended answer, high on the right or low on the left, which is geometrically what an anamorphic stretch produces. The Gallery's 2024 scholarly catalogue entry does not address viewing position at all. Three institutional voices, one of them the same institution twice.

Two smaller things the Gallery confirms and the popular telling skips. Dinteville's hat carries a small badge featuring a skull, so death was already in the picture before the anamorphosis. And the silver crucifix half-hidden behind the curtain at the top left sits, in paint-layer terms, on top of the finished green damask, so it may have been added late. Two memento mori, both smuggled in at the edges, one of them possibly an afterthought.

Video. "Holbein's Ambassadors: theories, questions and what you need to know," The National Gallery.

What it means to have one correct viewport

Every case above shares a property: the work was calculated for a viewing condition, and the viewing condition is not stored in the work. Pozzo's church solves it with a disc in the floor. Ames solves it with a peephole. Holbein does not solve it, on purpose.

Digital work has the same problem, several times over, and almost never marks the floor.

Table 1. Five works that know their viewing condition, and one medium that mostly does not.

Three practical consequences, in order of how sure I am.

• Design from a stated viewing condition, and state it. Every layout is calculated for something: a viewport width, a pixel density, a distance from the eye, an ambient light level, a colour profile, an attention state. Most of those are decided implicitly by whatever machine the designer happened to be using. Pozzo picked a spot and marked it. The equivalent is writing the target condition into the file, so the next person knows which one was privileged and which were sacrificed.

• Count the cues you are removing. The Ames room works by subtraction. So does a product screenshot with the error states cropped out, a chart with a truncated axis, a demo video at four times speed. None of those add a false claim. They remove the channels through which a false claim could be caught.

• Sometimes the right answer is Holbein's. Two views, mutually exclusive, both true, neither complete. That is an unusual thing to ship deliberately, and there are cases that deserve it. A dashboard that will not reconcile two definitions of the same metric might be more honest showing both and naming the conflict than picking one and looking authoritative.

A detail-by-detail walk through The Ambassadors, every instrument on that shelf and what it reads, is on my Korean blog: 베이니월드.

Sources

Object record, dimensions, inscriptions, the crucifix paint-layer observation and the hat badge: The National Gallery, NG1314 and its 2024 scholarly catalogue entry. Sant'Ignazio's fake dome, its thirteen-metre canvas and the 1823 reproduction by Manno: Turismo Roma. Palazzo Spada and Giovanni Maria da Bitonto: VisitItaly, with figures that other sources dispute. The Ames room's peephole, monocular restriction and suppression of parallax: the standard account of Ames's 1940 patent.

Note on images: the National Gallery licenses its own image files under CC BY-NC-ND, which forbids derivatives, so the reproduction above is the public-domain Google Art Project file rather than the Gallery's. Their descriptive text, by contrast, is CC BY, which is why I have been able to quote it directly.

Not verified and therefore not asserted: a measured viewing height for the Holy Trinity, the exact apparent length of the Spada colonnade, and the widely repeated claim that The Ambassadors was hung above a staircase so that visitors met the skull on the way up. No museum source supports the last one.