Stereograph: Twin Photos Shot a Couple of Inches Apart

Lay the two halves of Met 1980.1056.3 over each other, and the nearest pillar shifts 29 pixels. Line the halves up on the buildings far down Broadway, where the shift is zero, and everything nearer the camera slides sideways by its own amount. That sliding is the point of a stereograph: two photographs of one scene, taken from two points a couple of inches apart and mounted side by side on one card. In a stereoscope each eye sees one half, and the small differences between them read as depth.

At a glance the two halves look like copies of one print. They are not. The museum's record, as given on Wikimedia Commons, dates this card to about 1860 and credits it to Edward Anthony and Henry T. Anthony; it is two albumen silver prints, the record's name for this kind of photograph on paper, mounted under arched tops.

A name and a method from 1859

The name stereograph appears in an essay in the Atlantic Monthly of June 1859, titled “The Stereoscope and the Stereograph.” The magazine printed it without an author's name. On page 743 the essay calls the pair “the double-eyed or twin pictures, or Stereograph, if we may coin a name.” A page later it says “the stereograph, as we have called the double picture designed for the stereoscope.”

So the essay proposed the name, and said so. Whether anyone had used the word earlier is a separate question that this page does not try to settle.

The same page gives the method in one sentence. “A first picture of an object is taken,—then the instrument is moved a couple of inches or a little more, the distance between the human eyes, and a second picture is taken.” The next sentence offers the better way: “two pictures are taken at once in a double camera.” Either way, the two halves start from two different points.

Why two points a couple of inches apart give depth

Open the diagram full size

Hold up a finger and close one eye, then the other. The finger jumps against the wall behind it, while the wall stays put. The 1859 essay puts the same fact plainly on page 742: the two eyes see different pictures “for the obvious reason that they look from points two or three inches apart.”

A camera moved by that distance records the same jump. Far objects land in almost the same place in both pictures; near objects land in different places, and the nearer they are, the bigger the difference. The diagram shows one near post in front of a far wall: from the left lens it covers one part of the wall, from the right lens another.

David Brewster's 1856 book, The Stereoscope, gives numbers for the same idea. He takes the distance between the eyes as 2½ inches. For pictures made for his lens viewer, he writes that “the distance between similar points in the two pictures” varies “from 2.30 inches to 2.57, according to the different distances of the foreground and the remotest object in the picture.” In other words, the spacing on the card itself carries the depth.

Laying one half over the other

Here the left half is printed in violet and the right half in amber, then laid one over the other. Where the halves agree, the colors cancel into one plain picture. Where they disagree, an edge splits into a violet copy and an amber copy.

The far buildings sit together, while the carved figures on the corner building and the fluted pilaster beside them split wide. A pilaster is a flat pillar set into a wall, and this one is the nearest of the measured points.

The shifts come from a measurement on the Commons scan of the card, which is 3,871 pixels wide. For each box drawn on the left half, a program cut the box into small squares, 48 pixels on a side. For each square it searched the right half for the best match and kept only strong matches. The middle value of those offsets is the number for that box.

The far buildings matched at an offset of 1,934 pixels, and every other shift is counted against that offset. The scan gives no physical size for the card, so the numbers stay in pixels and are not turned into inches. Squares of other sizes move some figures by about a pixel, so read each one as a whole-pixel estimate.

How much each part of the street moves

Open the diagram full size

The order matches the layout of the street: distance grows from the corner building to the walkers and out to the far end, and the shift shrinks the same way. The lamp's figure rests on only 3 strong matches, because the post is thin, so treat it as rougher than the rest.

On this card the whole spread is 29 pixels out of 1,934, about 1.5 percent. Brewster's range for his own pictures, 2.30 to 2.57 inches, spans about 10.5 percent. He was writing about other pictures, made at other distances, so the two figures show the same pattern rather than the same amount.

The 1859 essay warns about one trap. “It is common to find an object in one of the twin pictures which we miss in the other,” it says on page 745, because “the person or the vehicle” moved between the two photographs. A moving walker would add a shift of its own. On this card the median for the walkers matches the lamp beside them, but that box holds several walkers and its matches spread from 1,921 to 1,927 pixels, so this measurement cannot say whether any one of them moved.

Download the shift measurements for this card (CSV) · source and license notes

Seeing the two halves in turn

You can watch the depth without a viewer by showing the halves one after the other. The loop below holds the left half for 3 frames, then the right half for 3 frames, at 3 frames per second. Each half stays up for 1 second, so the loop runs 2 seconds and the picture changes once a second.

Broadway in the rain, left half and right half in turn

Every frame, side by side

1. 1left half
2. 2left half
3. 3left half
4. 4right half
5. 5right half
6. 6right half

Both crops are lined up on the far buildings, so the end of the street stays still. The walkers move a little and the carved figures on the corner swing farther. If you see no movement at all, look at the right edge first, where the shift is largest. The right half is lighter in this scan, so the picture brightens when the right half comes up and darkens when the left returns; that change is one of tone, not depth.

Lining the crops up on a near object instead would flip the effect: the corner would hold still and the far street would slide. The depth is the same either way. Only the point you hold still changes.

The cards of The Horse in Motion turn this around. There, a line of cameras spaced along the track catches many moments of one horse, one picture each; here, two points a couple of inches apart catch one scene. Spinning drawings in a zoetrope makes motion from flat pictures; a stereograph makes depth from flat pictures.

Stereograph or stereoscope?

The stereograph is the card; the stereoscope is the viewer. The 1859 essay keeps the two apart in its title and its text. It asks how to “make one picture out of two,” when the matching parts are “separated by a distance of two or three inches.”

One answer is to cross your eyes, which the essay calls “tedious, painful, and to some impossible.” The other is a viewer with two half lenses joined at their thin edges. The essay names it “a squinting magnifier.” With the right half you see the right picture and with the left half the left picture, and the two run together into one.

Who devised the viewer is a crowded question. The 1859 essay credits the instrument to Professor Wheatstone and says he described it in 1838. Brewster's 1856 book names two men working on binocular pictures at that time, Wheatstone and an Edinburgh mathematics teacher named Elliot.

Brewster also describes the lens viewer he built himself, which he calls the lenticular stereoscope. This page leaves those claims side by side, since none of the sources read here settles which came first.