Improving your next eye scan accuracy

I’m an ophthalmology patient and get my eyes tested once a year. Because my case is rare, I’ve tried almost every kind of eye exam. One of them was color perimetry.

What the test does

Color perimetry measures how wide the color-sensing area in the fovea is, for red, green, and blue separately.

You sit in a dark room and focus on a dot in the center of the device. A curved line in front of you rotates to different angles. A small light appears from the far end of the line and moves toward the center. You say what color you see. Each point where you can name the true color is marked on a spiderweb-like graph. Then the device changes angle and the process repeats.

Two test runs, two protocols

• First time: the nurse kept switching colors. Each change made a click, so I knew something had changed.

• Second time: a different nurse tested one color across the whole visual field, then moved on to the next color. When I asked to randomize the colors, I was told, “Just tell me what you see.”

The problem

What we know affects what we see.

In the second test, I knew green was being tested and “saw” it in my peripheral vision. That should be impossible for the eyes to truly detect. My report wasn’t just vision. It was expectation.

Why this happens

Perception is how we identify what we sense. To do that, the brain pulls from memory and makes predictions about what we’re experiencing and what will come next. As we keep experiencing things, those predictions get updated.

Tests that rely on patient responses are shaped by these predictions. If we can guess what’s coming, our answers shift.

The fix

Randomize the stimuli.

• For color perimetry: randomize the colors.

• For common vision tests: show each eye different, mixed slides of letters or numbers so what you saw with one eye won’t affect the other’s vision through memory.

When my second color perimetry wasn’t randomized, it looked like my fovea had “grown.” It hadn’t. My expectations had.

Takeaway for products and services

This is a small but useful example of improving outcomes with cognitive psychology. Sometimes the change isn’t in the tool itself, but in how it’s used. Randomizing inputs can remove bias from memory and expectations, leading to more accurate results in tests like these.