Open Field Protocol

Window Marker Spacing Field Trial

A structured way to measure whether a 2 inch (5 cm) marker grid changes the strike rate at one specific window, using a real baseline instead of an impression. It takes a year, and most of that year is spent counting evidence of collisions nobody saw.

The question this answers#

The site's guidance is settled: markers on the outside surface, no more than 2 inches (5 cm) apart across the whole pane, and feeders either under 3 feet (1 m) or over 30 feet (9 m) from glass. That rests on tunnel testing and building-scale monitoring, which describe populations of windows rather than yours.

What you cannot look up is the size of the change at a particular pane: strikes before, strikes after, and how much of the difference survives once season and yard traffic are accounted for. This protocol produces that number in a poolable form, and separates four things people change at once: spacing, contrast, surface and feeder distance.

Why the existing evidence is thin#

Almost every home-scale report is an after-only account: someone had strikes, applied a treatment, stopped seeing strikes, and reported success. Three problems sit inside that story.

The first is that nobody counted before. People treat a window after a bad week, and a bad week is usually followed by a quieter one, so you cannot tell a working treatment from a return to normal.

The second is detection. A large share of collisions leave no body: birds fly on and die out of sight, or a scavenger clears the evidence within hours. If you only count what you find, and you look harder after treating the glass than before, you will manufacture the result you expect.

The third is timing. Collisions cluster in migration, so a July baseline against a September treatment period compares two different worlds. Few reports record season, feeder distance, or how much bird traffic the yard had.

What you need#

• One study window, picked by standing outside at bird height, 15 to 30 feet (4.5 to 9 m) back, and confirming the pane reflects sky or vegetation.
• A tape measure and a cardboard or paper spacing template so the grid is uniform.
• Marker material: tempera paint and a sponge stencil, exterior dot film, or tape squares. Any material works if the spacing is right.
• A flashlight for raking light across the glass, which reveals smudges and feather prints that are invisible head-on.
• A phone for the fixed reference photograph, and if you have one, a doorbell or trail camera aimed along the pane.
• Flags marking a ground search zone 10 feet (3 m) out from the base of the wall.
• A ventilated cardboard box, paper towel, gloves and the number of a licensed wildlife rehabilitator, assembled before the trial starts, not after the first strike.

How to set it up#

Describe the pane before you touch it: glazed area in square feet and square meters, orientation, what it reflects, and the distance to the nearest feeder in feet and meters. Note whether that distance falls in the dangerous 3 to 30 foot (1 to 9 m) band, because a feeder inside it is a variable you must hold still all year.

Take the reference photograph from outside at bird height, from a marked spot, in mid-morning light, and repeat that exact shot monthly to score coverage as the markers wear.

Set an exposure index: twice a week, at a fixed time, count birds at the nearest feeder for 10 minutes. Strike counts mean nothing if yard traffic fell from thirty birds a day to five.

The procedure#

1. Phase 0, baseline. Run the daily checks on the untreated window for at least 8 weeks, ideally a full season. Skip this phase entirely if the window is already killing birds.
2. The daily check, twice. Once within an hour of first light, before scavengers work the ground, and once in late afternoon. Sweep the flashlight across the glass at a low angle for smudges, dust prints and the wing-and-body outline a feathered bird leaves, then walk the ground zone on the same route.
3. Log every piece of evidence with a certainty class: witnessed strike, body, feather print, smudge only, or clean check. Photograph each smudge, then wipe it, so tomorrow starts from a clean pane.
4. Record the outcome of any bird found alive: stunned and recovered, stunned and transferred to a rehabilitator, or dead. Care comes first, notes afterward.
5. Phase 1, treatment. Apply the grid to the outside surface across the entire pane, edge to edge, at 2 inch (5 cm) spacing in both directions. Log the date, material, contrast against the glass, and measured spacing.
6. Hold everything else still: same feeder position, seed, curtains and lighting. Run phase 1 for at least as long as phase 0, matched to the same calendar window if you can.
7. Test spacing spatially, not sequentially. To compare 2 inch (5 cm) against 4 inch (10 cm) spacing, treat two similar hazardous windows at once with different grids. Never widen a grid on a pane that has stopped killing birds.
8. Test the surface question photographically. On an untreated pane, apply markers inside and shoot the reference photo from bird height in bright light, then score how much of the pattern survives the reflection. That costs no birds to show.
9. Re-score coverage monthly. Tempera thins in rain and film lifts at the corners. A stretch where coverage fell below about 80 percent is a different treatment and gets logged as one.
10. Continue for 12 months so both migration peaks appear in the record.

What you record#

One row per daily check, whether or not anything happened. The clean checks are your denominator.

A stunned bird sits upright, blinks and usually recovers within two hours in the box. That is a different outcome from a fatality and both belong in the record. Do not log a bird that flew off as unharmed, because delayed mortality is real and you cannot observe it.

Controls, and the ways this goes wrong#

Effort drift. The commonest bias here is looking harder in one phase than the other. Fix the route, times and duration, and log a check as completed only if you did all of it.

Scavenger removal. Crows, cats and chipmunks clear casualties fast. That is what the first-light check is for, and why body counts alone underestimate strikes.

Two changes at once. Applying markers the same weekend you move the feeder makes the run uninterpretable. One variable per phase.

Regression to the mean. Start the day after a bad week and the next period will look quieter whatever you did. Long phases dilute this.

Season mismatch. Eight baseline weeks in midsummer against eight treated weeks in late September compares the calmest period of the year with the noisiest.

Marker wear. Score coverage monthly, or you will report a treatment that stopped existing in March.

How many runs before the number means anything#

Strikes at one house are rare events, and rare events are statistically expensive. As a rule of thumb you cannot detect a halving of a rate unless the baseline period is expected to contain roughly ten or more events, and a window producing three or four strikes a year will not deliver that in a season.

So the protocol is pooled. One contributor supplies one window-year of standardized rows, and the answer comes from a few hundred, analyzed as counts per window-year with season and exposure as covariates. Your own run is not the finding.

Minimum accepted submission: 8 weeks of baseline and 8 weeks of treatment in matched calendar windows. Strong submission: 12 continuous months. Zero-strike windows are wanted, and so are treatment-only runs from windows too dangerous to leave untreated.

Submitting your results#

Download the datasheet for this protocol, keep one row per check, and send it through the Feeder Log. Use the same window_id all year so phases link, and give your location no more precisely than state or province.

Include the monthly photographs if you can, since coverage scoring varies most between contributors. Submit incomplete runs rather than discarding them: a run that stopped in month four with a note is usable, one reconstructed from memory is not.