Open Field Protocol

Paired-Choice Seed Preference Trial

Two seeds, two identical feeders, side by side, positions swapped on a schedule so the yard's own geometry cannot masquerade as a preference. You measure grams eaten and visits made, not impressions.

The question this answers#

When a bird can have either of two seeds for equal effort, which does it take, and by how much? That is a cafeteria design, also called a paired-choice test, and it is the one feeder experiment a backyard can run that yields a comparable number.

It will not tell you which seed is nutritionally better or what a species prefers everywhere. It tells you what happened in one yard, in one season, between two products, with the side of the pole controlled for. Stack enough of those and a pattern emerges; one alone is an anecdote with a scale.

Why the existing evidence is thin#

Seed preference rankings circulate widely and trace back to little: a few decades-old cafeteria studies, manufacturer copy, and one person watching one feeder. The recurring failures are structural and each is fixable.

• Sequential testing. Seed A one week, seed B the next, with temperature, day length and flock arrivals all changing in between.
• No position swap. Two feeders sat where they were for a month. In most yards the one nearer cover wins regardless of contents, and that gets reported as preference.
• Eyeballed levels. A narrow tube drops visibly faster than a wide one holding the same mass.
• Uncounted losses. Seed kicked to the ground, taken by squirrels or blown out in wind is scored as eaten.
• Unstated duration. Unequal blocks compared as equals.

What you need#

• Two feeders of the same model, port count, capacity and color, ideally bought together. Not "both tube feeders": the same feeder twice.
• A digital scale reading to 1 g (0.04 oz) with a capacity of at least 5 kg (11 lb). A hanging luggage scale is not precise enough.
• Two spill trays or tarps at least 3 ft (0.9 m) square, one per feeder, kept apart so catches cannot mix.
• Two labeled sealed containers, each filled from a single bag used all trial.
• A timer, an outdoor thermometer, a printed datasheet, and baffles or squirrel-resistant feeders, so non-target mammals are excluded rather than noted.
• Optional: a trail camera per feeder, and a third identical feeder screened from birds as a moisture control.

How to set it up#

Pick two seeds differing in one interesting way: hulled against unhulled sunflower, black oil against safflower, a premium mix against a supermarket mix. Buy one bag of each and use only those bags, because lots vary.

Mount both on the same crossarm or rail, within 1 in (2.5 cm) of each other vertically and 24 to 36 in (60 to 90 cm) apart. Closer and birds treat the pair as one feeder. Farther and the positions stop being comparable, because cover and wind exposure begin to differ.

Label the two positions, left and right, never the feeders by seed. Every entry is recorded against a position first, then the seed in it. That is what makes a check for side bias possible.

The procedure#

1. Weigh each empty feeder, clean and dry, and write the tare mass on tape inside the lid. Hang both full and let birds use them for 3 to 5 days, recording nothing.
2. Run the same-seed baseline: identical seed in both positions for two matched blocks, weighed as below.
3. Fix your block window and keep it. A set daylight window such as 07:00 to 17:00 beats a rolling 24 hours, because activity is not spread evenly through the day. Blocks must be equal in length.
4. At block start, fill both to the same fill line, weigh each loaded feeder to 1 g (0.04 oz), and record mass, clock time, seed by position, temperature and weather.
5. Leave them alone. Do not top up mid-block. If one drops below a quarter full early, end the block, note it, and shorten the window from then on.
6. At block end, weigh each loaded feeder again, then sweep each spill tray separately, dry wet contents and weigh each catch.
7. Compute consumption per feeder as start mass minus end mass minus spill mass, in grams.
8. Swap the feeders between positions, moving each whole feeder with its seed still in it. Refill both and weigh: those are the next block's start masses.
9. Run two 20 minute observation sessions per block at fixed times, counting visits and species per position. A visit is one bird landing and taking at least one seed; a bird returning after 30 seconds away counts again.
10. Clean both feeders at the same time, by the same method, and record the date. Never clean one and not the other. Repeat until you have at least 8 scored blocks, 4 with each arrangement.

What you record#

One row per block, with these fields.

Controls, and the ways this goes wrong#

Position is the big one. Birds approach from cover, and the feeder nearer cover gets more traffic no matter what is in it. Swapping at every weigh-in, with equal blocks in each arrangement, turns that bias into something measurable. If the result reverses on swapping, you measured the yard, not the seed.

Eaten faster is not the same as preferred. A seed carrying more calories per gram meets a bird's energy need in fewer grams, so equal calorie intake shows up as the denser seed losing less mass, which on a grams-only scoreboard reads as a loss. Report grams per hour and kilocalories per hour. Handling time cuts the same way: an unhulled seed takes longer to open, so mass leaves slowly even if the bird would rather have it. That is why visit counts sit alongside mass instead of replacing it.

Moisture moves the number. Seed gains mass in rain and loses it drying out, so a feeder can finish a wet block heavier than it started. The screened control feeder holds seed no bird can reach, so its mass change is moisture alone: subtract it from both trial feeders.

Non-target consumers. Exclude them rather than estimating them. Baffle the pole and use feeders a squirrel cannot open. If a squirrel gets access mid-block, void the block. Ground feeders working the spill trays are a separate problem: keep trays raised and shallow, and note any juncos or doves cleaning them out.

How many runs before the number means anything#

Eight scored blocks is the floor: four with seed A on the left, four with it on the right. Below that, one windy afternoon carries the result.

For each block compute the difference in grams per hour between the two seeds, keeping the sign, then check two things. Is the sign the same in most blocks? A direction holding in 7 or 8 of 8 is worth reporting; a 5 to 3 split is noise. Is the average difference larger than the scatter between blocks? If not, run more blocks rather than reaching a bolder conclusion.

Then average within each arrangement separately. If seed A wins on the left and loses on the right, position beat seed, and the write-up says so. A result in one yard is about one yard: three independent sites is where a finding starts to travel, and a second season tells you whether it holds when energy needs change.

Submitting your results#

Fill in one row per block on the datasheet for this protocol, including voided blocks with the reason. Voided blocks are data about how the protocol behaves in the field, and deleting them quietly is how a clean dataset turns misleading.

Send it with the site details, the exact seed products and lot codes, your feeder model, and photographs showing both positions. Results, including runs that find no difference, appear in the Feeder Log with the contributor credited. A null result from a careful trial beats a strong result from a trial that never swapped positions.