Research

Open Field Protocols

8 replicable test procedures, each with materials, controls, a scoring scale, sample size guidance and a datasheet. Run one in your own yard and send us the numbers.

How to read this section

  • These are protocols, not results. Every one is open for replication and none carries a findings table yet. When runs come in, results get their own pages with the sample size attached and the raw datasheets downloadable.
  • They are written to be run by one person with ordinary equipment: a kitchen scale, a thermometer, a phone camera and patience. None needs a lab.
  • Every protocol is constrained by the welfare rules on the methods page. No withheld food, no trapping, nothing on a pole or in food that could hurt a bird.
  • Casual observations are welcome too and do not need a protocol. That is what the Feeder Log is for.

Why publish protocols before results#

Because a protocol published in advance is a commitment. It fixes what will be measured, how, and what would count as the test failing, all before anyone knows how it turns out. That removes the easiest way to produce a confident-sounding result: run something loosely, see what happened, then describe the method as if it had been designed to find that.

It also means the work is not bottlenecked on us. Backyard feeding questions are local by nature, and a hundred people running the same procedure in a hundred yards is a genuinely better instrument than one person running it in one.

Running one#

  1. Pick a protocol and read all of it before you start. The controls sections are where the value is, and most of them exist because the obvious way to run the test does not work.
  2. Download that protocol's datasheet, which is a CSV with the exact columns to record.
  3. Run it. Record everything, including the runs that go wrong, because a run abandoned when a hawk moved in is information.
  4. Send the filled datasheet through the contact page, and log the summary in the Feeder Log.

Null results are wanted. "I ran this for six weeks and nothing changed" is a real result, and it is the one that never gets submitted anywhere, which is exactly why published advice drifts toward whatever produced a visible effect once.

What we will do with your data#

Aggregate it, publish it with its sample size, and credit you if you want crediting. Raw datasheets are published alongside any results page so the analysis can be checked. The thresholds and claim-language rules on the methods page apply to protocol results exactly as they apply to everything else here.