Four things that help the birds in your yard

The DC study is Narango, Tallamy & Marra 2018, Nonnative plants reduce population growth of an insectivorous bird, in PNAS. Nest boxes in the yards of 159 NestWatch volunteers, monitored 2013 to 2016.

Worth being precise about the mechanism, because it's easy to get backwards. The strongest effect they measured was on settlement: chickadees were markedly less likely to move into a non-native yard in the first place. A second, weaker effect was on the pairs that did settle, who had lower nest success and fledged fewer young. So non-native yards are not luring birds in and killing them. They are mostly being avoided, and they underperform for the few that try.

The thresholds: mean population growth was only positive below about 6% non-native plants. Up to 30% non-native, the confidence intervals still overlapped replacement, so those yards may hold their own. Above 30% non-native, the paper calls them population sinks. That's the same line as the abstract's "more than 70% native plant biomass," from the other direction. 70% native is a higher bar than most people expect.

Honest limits: one region, one species, one study. Carolina Chickadees were picked partly because they're generalists that forage almost anywhere, which makes the result more striking, but it's still one bird.

The 6,000 to 9,000 caterpillar figure is a season-long count for a brood of about five chicks, and it's the ceiling of a range, not a daily rate. It predates the DC study, and the sources quoting it attribute it to "previous studies" without naming one, so treat it as a well-established rule of thumb rather than a single measurement.

Oaks, cherries, elms and maples are named because they came out on top for caterpillar and sawfly larvae counts in this line of work. American plum (Prunus americana) is in the same genus as cherry and is listed as a larval host for coral hairstreak, eastern tiger swallowtail, spring azure, viceroy and red-spotted purple.

One thing to know about American plum before you plant it: NC State's plant database says it "suckers freely and can form large colonies or thickets," and that it will do this with thorn-like side branches unless you stay on top of it. Cut the runners each year and it behaves like a small tree. Ignore it and you have a hedge. If you'd rather not manage that, serviceberry is the better-behaved pick.

The fruit study is Smith, DeSando & Pagano 2013, The value of native and invasive fruit-bearing shrubs for migrating songbirds, in Northeastern Naturalist. Their numbers: native fruits "ranging from 6.57 to 48.72% fat," while "all invasive fruits had ≤0.99% fat." They also found native dogwood fruits were eaten by migrating songbirds at higher rates than the invasive ones. The Center for Conservation Biology has measured sassafras above 60% fat, which is the highest I've seen for a fruit in this region.

Two different jobs, two different rankings, and it's worth not mixing them up. On caterpillars, serviceberry is usually put around 120 species of butterflies and moths and elderberry somewhere in the 30s; NC State doesn't list elderberry as a larval host at all. On fruit, elderberry is excellent, and NC State says plainly that "its fruits are eaten by mammals and many species of songbirds." A plant that's mid-table for one job can be top of the list for another. Plant the elderberry.

On nesting cover I'm being deliberately vague, because I haven't found a clean number for it the way there is for caterpillars and fruit. Audubon's line is that more native plants mean "more choices of food and shelter for native birds," which is true and unquantified, and I'd rather say it that way than dress it up.

The study is Anderson, Code, Malfi, Holladay & Forister 2025, An assessment of risk to pollinators from mosquito control in residential settings, published in Stacks Journal by the Xerces Society and the University of Nevada, Reno.

The method: silicone bands that absorb pesticide were placed at 19 volunteer properties across Georgia, Iowa and Massachusetts, before real spray events and collected 24 to 48 hours after. "Six times the lethal dose" is the average honey-bee risk quotient of 6.6 in yards sprayed by a private company. The neighboring-yard average was much lower, 0.12, and the number I quote in the body is the tail of that: nine of 40 samples, at five of 14 sites, reached medium or high risk.

One number you may see elsewhere: a press release about this study says 75% of neighboring-yard samples showed contamination. That's a detection rate, not a harm rate, and it is not the same as the 23% that reached medium or high risk. I've used the harm figure. Detectable is not the same as dangerous, and it matters that we don't blur them.

The authors' own limits: single spray events rather than a full season of them, a small number of sites, and a method that measures contact exposure, which if anything understates the real risk rather than exaggerating it.

One more thing. "Natural" and "botanical" barrier sprays are usually pyrethrins, which are the chemical parent of the synthetic pyrethroids in the conventional stuff. Same broad-spectrum effect on whatever lands on the leaf. It isn't a different thing, it's a different word.

Worth being fair about: barrier spraying does give real, temporary relief in the yard it's applied to. The argument here isn't that it does nothing. It's what it costs, and that the cheap permanent fix is the standing water.

On timing. The eight-to-ten-day cycle figure is Xerces'. The February cutoff comes from the same DC chickadee study linked above, which monitored nests March through June, and Cornell's NestWatch shows the same March start for most of the range. Chickadees also start building roughly three weeks before the first egg is laid, which is what puts early March inside the season rather than safely before it. October is the least damaging month, not a harmless one: fall migration runs through mid-October and migrating birds refuel on insects, monarchs are still moving through the mid-Atlantic into about the third week, and the overwintering stages of a lot of insects settle into leaf litter and bark in October and November. "Late October, once the leaves are down" is the most accurate version of this advice.

I want to be straight that this timing argument is reasoning from nesting calendars and how long the residue lasts. It is not a study measuring an October spray against winter bird food. As far as I can tell, nobody has run that study.

The estimate is Loss, Will, Loss & Marra 2014, Bird–building collisions in the United States, in The Condor, built from 23 studies. Their exact wording: 365 to 988 million birds, median 599 million, "roughly 56% of mortality at low-rises, 44% at residences, and <1% at high-rises." The remaining 56% is low-rise commercial buildings, four to eleven stories, which is where office parks and strip malls live.

The authors flag their own bias, and I'll pass it along: most collision data was collected during migration and in the eastern US, so the list of most-affected species leans that way.

On spacing, the numbers in circulation don't all match, so here they are side by side. American Bird Conservancy's headline is the "two-inch-by-two-inch rule," and it warns that small birds and hummingbirds "can easily fly through four-inch-wide gaps." But ABC's own decal guidance on the same page allows "two to four inches apart," and BirdSavers specifies 3.5 to 4.25 inches for hanging cords. ABC also lists BirdSavers by name as a recommended product, so nobody here is calling anybody wrong.

My read, and I'll flag it as a read rather than a citation: two inches is the conservative target, two to four covers most birds, and a vertical cord is a different visual cue to a bird in flight than a flat sticker. If you get hummingbirds, go tighter. I couldn't find a source that explains exactly why cords hold up at four inches when the general rule is two, so I'd rather point at the gap than paper over it.

The source is Loss, Will & Marra 2013, in Nature Communications. The number people usually quote from it is 2.4 billion birds a year, and I've deliberately not led with that, because roughly 69% of it is un-owned cats: strays, feral colonies, barn cats. That's a policy problem, not something a cat owner can act on, and using it to talk to a cat owner isn't fair.

The owned-cat figure is the paper's own separate estimate: a median of 684 million birds a year, with a range of 221 million to 1.68 billion. That's the "up to 1.7 billion" in the body. It is still, by a wide margin, the largest number on this page that a single household can change on its own.

Why the range is so wide: the authors ran a sensitivity analysis and the biggest single source of uncertainty is that nobody knows how many un-owned cats there are in the country. That one unknown accounts for about 42% of the variation in the total. The paper also produces three different estimates depending on which countries' data are included; I'm using the US-and-Europe version, which is the one the authors chose to focus on because it's the lowest of the three.

The fair criticism, and it's a real one: a mortality total is not the same as a population-level impact. Knowing that X hundred million birds die doesn't automatically tell you which populations are declining because of it. That objection is worth knowing about before you repeat the number at a dinner party.

Also from the paper, and often left out: the majority of birds cats catch are native species, not house sparrows and starlings. Only about a third of identified prey were non-native.

On hunting and hunger: the 2021 review is Cecchetti, Crowley & McDonald in Mammal Review, and the exact line is "Most cats retain some propensity to express hunting behaviours, even if hunting is not required for nutrition." The cats-leaving-their-food experiment is Adamec 1976, in Behavioral Biology. Note the review hedges, and so should we: it says most cats, not all, and "propensity," not compulsion. The same review also reports that better nutrition and more play can reduce how much a cat hunts, which is worth knowing, but reducing is not the same as stopping, and it isn't a substitute for the door.