The window has not changed. This is the part that takes a moment to accept, standing in a lamp-lit room at night watching your own bookshelf hover over the garden. Ordinary window glass reflects a small, fixed fraction of the light that strikes it—about four percent at each surface, or roughly eight percent round trip for a plain pane—and it does this at noon, at midnight, and at every hour between. At midday that eight percent is nothing. The view through the glass, carried by the other ninety-odd percent, is built from a sunlit street and swamps the reflection completely. At night the street goes dark. Your room does not. The reflection of your lamp-lit ceiling becomes the brightest signal the pane sends toward your eyes, and the window behaves, as far as your retina is concerned, like a mediocre mirror with a dim picture trying to fight through it. Nothing about the glass turned on. The balance of the light did. Glass, in fact, always has opinions about light—they accumulate visibly in the green tint deep inside a hall of mirrors—but on an ordinary pane the opinion is a whisper that only becomes the whole story when the competition leaves.

Reflection was happening all along

The numbers come from the Fresnel equations, the nineteenth-century rules that predict how much light bounces at a boundary between two transparent materials. The mismatch between air and glass is modest, so at a perpendicular angle only about four percent reflects; at grazing angles—light arriving almost parallel to the glass—the reflected share climbs steeply, which is why distant windows viewed at a shallow angle blaze like silver even by day. Watch a row of houses from across a valley at sunset and every window is a mirror. You already live in a world where glass reflects constantly; the daytime transparency of your living-room window is the special case, maintained by the brightness being on the far side.

Two reflections are stacked in every pane, one from each surface, and their interference in thin coatings is a whole industry—low-emissivity window films are engineered to tweak exactly this reflective behavior. But for plain glass the story is arithmetic: transmit the majority, reflect a sliver, absorb almost nothing. The sliver never goes away.

The contest that flips at dusk

What you see in a pane at any moment is a sum. Reflected interior light travels one path; transmitted exterior light travels another; both land on the same patch of retina and your visual system reports the stronger one, with the weaker riding on top as a ghost. Dusk is the moment the two terms cross. As the sky dims, the transmitted term sinks while your lamps keep the reflected term steady, and the reflection that was invisible at lunch becomes unmistakable after dinner. Your eyes adapt to the darker scene too, which raises the reflection's apparent brightness further—the same shift in sensitivity that gives a gray car a blue cast at dusk as daylight vision hands off to dim-light vision.

The flip has practical moods. Interior designers place lamps to make glass walls glow; privacy-conscious people learn that the mirror works both ways in reverse: at night, with your lights blazing, you see yourself and cannot see out, while anyone outside your dark yard sees in perfectly. The one-way effect belongs to whoever holds the darkness.

Interrogation-room mirrors are just lighting design

The classic two-way observation mirror uses the same physics with the dial turned up. A thin metallic coating raises the reflected share and lowers the transmitted share, so the glass is half mirror, half window. But the coating alone does not make it one-way. The observing room is kept dark and the observed room bright, which rig precisely the contest described above; reverse the lighting and the 'one-way' mirror obligingly reverses, letting the suspect study the police. Glass manufacturers' descriptions of one-way glazing stress exactly this: performance depends on the luminance ratio between the two sides, not on any optical one-way valve, because no such valve exists in a plain pane. It is the same lesson as the mirror tunnel’s accumulating green: glass has physical opinions about light, and given enough trips or enough imbalance, those opinions become visible.

Living with a two-way window

Once you see the mechanism, the nightly mirror becomes readable. If you want the view back after dark, you have three levers: dim your own side, light the outside—porch lights and path lights raise the transmitted term—or block the reflection mechanically with curtains, which is why every night-time room with a view ends up with a lighting plan. Photographers fight the same ghost at every window, angling shots to keep lamps out of the reflection or polarizing against it, a small cousin of the interference tricks in what polarized lenses reveal in car windows. And on a walk at night you can spot which neighbors keep bright rooms—streetward, their windows are aquariums; inward-facing, they are mirrors—while the moon, reflectance no better than worn asphalt, still sends enough light through a dark window to cast a shadow indoors.

The window is always doing both jobs, transmitting and reflecting, without preference. Which one you witness is not a property of the glass. It is a property of where the light happens to be.

Two signals, one paneIllustrative balance, not measured values
Day
View (transmitted): dominant
Room reflection: present but swamped
Dusk
View: fading fast
Room reflection: unchanged
Night
View: nearly dark
Room reflection: now the winner

The reflection bar never grows. The other bar shrinks beneath it.