Lookalike comparisons pairs: what the side-by-side does

Putting two pictures side by side is not neutral display. Lookalike comparisons pairs are an instrument, and the instrument has a known effect on the viewer.

The pair invites the eye to find agreement, supplies the frame in which agreement counts, and removes the option of concluding nothing. Understanding that is more useful than any particular pair.

What to take away

The device that makes the point

Nineteenth century parlors held an instrument built on exactly this principle. Charles Wheatstone first presented the stereoscope in 1838 to the Royal Society in London, using a pair of mirrors, and because practical photography was not yet available, the first pictures were drawings. David Brewster suggested lenses in 1849, and the lens-based card viewer became the parlor object.

That viewer presents two photographs of the same object, taken from slightly different angles, one to each eye, and the viewer does not see two pictures. The apparatus and the visual system fuse them into a single scene with depth that neither photograph contains.

That is the whole argument in a piece of furniture. Two images, presented together in a controlled arrangement, produce a perception that is in neither of them. A side-by-side comparison is a cruder version of the same trick, and the thing it manufactures is agreement rather than depth.

Why the eye compares wholes

Face processing does not proceed feature by feature. The system works on the arrangement of a face taken together. That is why an upside-down face is hard to read, which psychologists call the face inversion effect, and why a small spacing change can make a familiar person unrecognizable.

The Thatcher effect shows the same thing from the other side. Changes to facial features are difficult to detect when a face is upside down, even though the same changes are obvious upright. Peter Thompson, a psychology professor at the University of York, first demonstrated it in 1980 on a photograph of Margaret Thatcher.

With two faces at once, the viewer compares two configurations, and configuration is dominated by outline, hair mass, coloring and general proportion.

Those four are also the four things a layout most easily equalizes. Matched crop equalizes outline. Matched scale equalizes proportion. Consistent grading equalizes coloring. Similar framing equalizes hair mass. The layout has adjusted every variable the eye weights most heavily, and left the small stuff, which the eye weights least.

The question arrives before the evidence

A pair is not a question mark. It is a proposition with two illustrations attached. By the time anyone looks at either face, the claim has been made and the viewer's job has been reduced to agreeing or disagreeing with it.

Being given a category or a name in advance shapes what is noticed next, and a caption supplies one efficiently. So the honest test is whether you would have paired these two unprompted, out of everyone you know. The answer is almost always no, which locates the resemblance in the format.

What the layout adds on top of the camera work is the mechanism this page is about.

A named pair that survives the format

The Red Hot Chili Peppers drummer Chad Smith is widely known for his resemblance to the actor Will Ferrell, and he has acknowledged it by wearing shirts reading "I Am Not Will Ferrell" in live performances. The two have used the resemblance in joint appearances, including a 2014 drum battle on The Tonight Show Starring Jimmy Fallon.

That pair is useful precisely because it does not depend on a matched layout. Viewers saw both men on one stage, in the same light, moving and playing. Run the same pair through a cropped, graded side-by-side and the format adds its own agreement on top; the stage appearance is the evidence, the layout is only the illustration.

The one-person control

The cheapest way to see the format working is to run it on somebody you can verify. Take two photographs of a single person: different years, different cameras, different light, different haircut. Present them as a pair with a caption asking whether these two are the same person.

Viewers hesitate. Some say no. The face is identical by construction, and the pair format has produced disagreement about a fact. Run it the other way, matching crop and grading across two different people, and the same viewers find a resemblance. The format is doing more than either set of faces.

What survives the exercise

A matched pair of two named people built only from photographs does not survive: it is a permanent claim, and an instrument with a known bias builds that claim about people who cannot answer it. A pair the two people themselves acknowledge, like Smith and Ferrell, rests on more than the layout.

The same instrument runs under every scored guessing game, so a quiz score measures the format as much as the player.

Common questions

Is a side-by-side ever appropriate?

For teaching a mechanism, yes, with images of one person or of objects. For asserting that two people resemble each other, it is the wrong instrument for the claim.

Does presenting the two images separately fix it?

It weakens the effect. Sequential presentation forces a memory comparison, which has its own problems, but it removes the manufactured alignment.

Why does a matched pair feel so much more convincing than a description?

Because a picture arrives as perception rather than as an argument, and perception does not present its assumptions for inspection.