Parallax
Seo-yun had been interpreting stereo pairs for the geological survey for six years. The work was simple to describe — look at two aerial photographs taken from slightly different positions, and the flat images would fuse into three dimensions. Terrain would rise. Valleys would open. The landscape that each photograph compressed into flatness would, in the space between two viewpoints, become deep.
Simple to describe. The actual skill was in the looking.
New analysts always asked the same question: "How do you make it work?" As if stereoscopic fusion were a technique you could apply. Seo-yun had learned to answer honestly: "You don't make it. You wait. And then you stop seeing two pictures."
The depth arrived between moments. The left eye saw the left photograph, the right eye saw the right, and somewhere between the two — not in either image, not in either eye — the ground rose. The photographs remained flat. The depth existed only in the gap.
She'd worked with Dr. Haines during her first year. He was exacting about methodology but imprecise about his own expertise. When she'd asked him to describe what he saw during fusion, he'd said: "I don't see anything during fusion. I see terrain after it."
"But in the moment —"
"There's no moment. There's two flat images and then there's depth. The transition isn't visible."
That had frustrated her then. Now she understood. The parallax worked by erasing awareness of itself. When stereoscopic fusion succeeded, you stopped seeing two photographs. You saw the world. The instrument's disappearance was the sign that it was working.
The problem cases were the instructive ones. Sometimes a pair wouldn't fuse — the overlap too narrow, the angular difference too great. Each photograph remained stubbornly itself: flat, bounded, a rectangle of compressed information. She could see both images but not through them. She could compare them but not combine them.
Other times the overlap was too complete. Both photographs taken from nearly the same position, the angular difference so slight that there was nothing to triangulate. The images fused instantly but the terrain remained shallow. Everything aligned. Nothing rose.
Seo-yun began to think of separation as its own kind of information. The gap between viewpoints wasn't an obstacle the parallax had to overcome — it was the raw material the parallax was made of. Too little separation: everything flat. Too much: everything incoherent. The depth required a specific distance between positions, and that distance was the one thing neither position could measure.
Her supervisor, Park, noticed she was spending more time on the problem pairs.
"Are you stuck?"
"Not stuck. Thinking."
"About?"
"About what the pairs that don't work can tell me. The failures show the structure. When a pair fuses, I forget there were two photographs. When it doesn't, I can see the mechanism."
Park considered this. "That's the description of every functional tool. You stop seeing the hammer when the nail goes in."
"But the hammer is still there. With stereo vision, the two images genuinely disappear. You're not looking through them anymore. You're looking at the world."
"You're looking at a model of the world."
"Is there a difference?"
Park didn't answer. Seo-yun noticed he rarely answered questions that interested him.
She found herself returning to one particular failed pair. Photographs of a river delta, taken on two separate flights three months apart. The angular geometry was correct — both images covered the same terrain from appropriately different positions. They should have fused. But the river had shifted between flights. The left image showed one channel; the right showed another. When she tried to combine them, the delta flickered — the brain insisting that the same point couldn't be in two places, the parallax engine producing not depth but conflict.
Each photograph was accurate. The terrain was the same terrain. But the ground had moved between viewpoints, and the fusion couldn't reconcile a surface with itself.
She showed it to Haines, who'd retired but still came by the lab on Thursdays.
"What do you see?" she asked.
He looked for a long time. "I see two rivers."
"There's one river."
"There's one river that moved. The parallax is trying to give it depth, but the depth doesn't exist — what exists is time. The separation between the photographs isn't spatial. It's temporal."
"Can you get depth from temporal separation?"
"Not in photogrammetry. In life, constantly."
She sat with that. Two versions of the same thing, taken from different moments, overlaid. The parallax produced not terrain but change — the shape of what had moved between one looking and the next.
She couldn't use the pair for the survey. It failed every formal criterion: no consistent depth model, no reliable elevation data, no measurable surface. But she kept it pinned above her desk. Two flat images that couldn't become one world, because the world they were both faithful to had become two worlds while they weren't watching.
The functional pairs didn't teach her anything new. They fused, terrain rose, she mapped it. The process worked by disappearing. The failed pair stayed visible — two images that remained two images, each accurate, each insufficient, the gap between them holding something neither could contain alone.
She hadn't been measuring terrain. She'd been measuring the space between positions. The depth was always in the gap. It was just that successful parallax made you forget the gap was there.