A flock of birds migrating over a city at dawn

The science

How birds
see glass.

Vision, reflection and testing, in three chapters, drawn from the papers behind our commitments.

01 · Eyes

Birds don't see glass the way we do.

People learn what glass is, from cues like frames and right angles. Birds do not grasp glass as a transparent or reflective barrier, so a row of decals that tells us there is a wall looks to a bird like a row of obstacles to fly around or between.4

Close-up of a songbird's eye, set on the side of its head
Eyes set to the side, tuned for open air.
A glass facade as a person sees it

Illustrative rendering. It is not a measurement of what any bird sees.

Colour

In most birds, colour vision runs on four types of cone photoreceptor, one more than humans have. One type is tuned to very short wavelengths: ultraviolet or violet.8

Field of view

Most birds have eyes on the sides of the head, with little overlap between visual fields and modest 3D vision. They look to the side for cues about flight speed.4

Contrast

Humans are better than birds at telling objects apart at a distance, so a faint pattern we spot easily may not stand out to a bird.4

Ultraviolet

UV is not a magic fix. Many bird groups cannot see it, and there is little UV at dawn or on low-UV days. In tunnel trials, UV materials gave varied results, and higher contrast scored better.4

02 · Failure

Why glass fails.

Reflections are the most serious threat glass poses to birds: they cannot tell reflected habitat from the real thing.4

Sky and trees reflected in a window pane
Sky and trees, reflected in a pane.

Danger per building isn't the whole story.

Homes · 1–3 stories
2.1
Low-rises · 4–11 stories
21.7
High-rises · 12+ stories
24.3

Birds killed per building per year. Per building, high-rises are the deadliest, but there are so few that they account for under 1% of the total. Homes kill about two birds each, but there are so many that they account for 44%.14

03 · Testing

How we know what works.

Bird-safe is not a feeling. The standard method is a flight-tunnel choice test: no injuries, one flight per bird, and a plain clear control to beat.4

A bird-banding and testing setup
Testing, not guessing.

Same coverage, different result.

2 cm stripes, vertical · 16% coverage
94
2 cm stripes, horizontal · 16% coverage
78
Frit lines, vertical · 20% coverage
90
Frit lines, horizontal · 20% coverage
94
Frit dots, 3.2 mm · 20% coverage
59

Tunnel score. The left marker is 50 (no effect), the right marker is 70 (recommended minimum).

Pattern coverage did not predict the score. Two stripe patterns covering the same 16% of the glass scored very differently depending on orientation, and a pattern of tiny, closely spaced dots scored no better than random. The study's author speculates that such dots may look to a bird like empty sky or fog.45 It is why dots are not a free pass, and why every Double Take product is tunnel-tested.

Principle 1

Outside surface

Put the pattern on the outside of the glass, where reflections cannot hide it.4

Principle 2

Mind the spacing

Small songbirds are frequent victims, so patterns need relatively narrow spacing tied to body size.4

Principle 3

Test, don't guess

Coverage alone does not predict effectiveness. Orientation and spacing matter, so patterns have to be tested.45

Questions

Common questions.

Do birds learn where a window is?

They can learn about particular pieces of glass, but they do not grasp the concept of glass as a barrier.4

Are small birds the main victims?

Small songbirds are frequent victims.4 In one long-running study, window strikes were equally lethal for small and large species.7

Does hitting a window always kill a bird?

Not always, but often. At the two single homes Klem monitored, at least one in two birds that struck the windows was killed.7

Is ultraviolet the answer?

Not on its own. Many bird groups cannot see UV, there is little UV at dawn or on low-UV days, and results with UV materials varied in tunnel tests.4

Where should a pattern go on the glass?

On the outside surface, so reflections cannot hide it. If a pattern has to go on an inside surface, the outside surface needs low reflectivity.4

Sources on this page: 1 4 5 7 8. Full list


See our commitments

Read the evidence