NYC Clean Air Rules Raised Toddler Cognitive Scores

1,109 New York children posted higher age-3 mental scores as prenatal soot fell, Columbia researchers reported on August 15, 2026. The gain showed up in Northern Manhattan and the South Bronx, where mothers had lived beside bus depots and residual-oil boilers. Clean air policies aimed at buses, taxis, and heating oil were never sold as brain policy. The developing fetus was the party those rules actually moved.

Columbia Mailman School’s public note on the paper put the age-3 jump at 26 percent when the first cohort is set beside the latest one. Fine particles at the mothers’ homes had already dropped 51 percent, nitrogen dioxide 48 percent, and backpack-measured combustion chemicals 66 percent.

Northern Manhattan and the South Bronx Carried the Exhaust

The Columbia Center for Children’s Environmental Health built its flagship study in low-income tracts of northern Manhattan and the South Bronx. Asthma rates there rank among the highest in the country, and low birth weight is elevated too. The mothers were healthy nonsmokers who had lived in the neighborhood for at least a year.

Those blocks did not draw a random share of city soot. They sat next to the sources the later laws would target, including diesel bus depots and commercial roadways, plus the heavy residual oil that once heated thousands of large buildings.

WHERE THE COHORTS LIVED WITH THE FUMES

• Bus depots: Diesel fleets laid overnight in the same districts where the mothers received prenatal care through New York-Presbyterian.
• Residual oil: No. 6 boilers packed northern Manhattan and the South Bronx; those tracts later saw the densest conversions and the largest local air-quality lift.
• Roadways: Major truck and bus corridors ran through the enrollment zone, feeding nitrogen dioxide at the doorstep.

Fine particles have long been treated as a neurotoxicant in the same class as lead. Cities almost never get a 20-year stretch in which the dose falls and the same families keep taking the same developmental test. This one did.

1,109 Children Took the Same Test for Two Decades

Three successive birth groups make up the analysis. Together they enrolled 1,446 mother-child pairs; 1,109 of those children sat for at least one Bayley exam at age 1, 2, or 3. The testers kept the second edition of the Bayley Scales so the scoring key would not drift.

THE THREE CCCEH BIRTH GROUPS

The first birth in the Mothers and Newborns group was on March 16, 1998. Women who started prenatal care after week 20, smoked, used illicit drugs, had diabetes or hypertension, were HIV-positive, carried twins, or had not lived in the city for a year were kept out. Fair Start took English- or Spanish-speaking adults over 18. Only the first 126 Fair Start pairs had the backpack PAH protocol before the pandemic shut it down.

PM2.5 and NO2 were modeled at each home on a 25-by-25-meter grid for the first two groups. PAH came from a 48-hour personal backpack the pregnant women wore, a method the center had already checked as a stand-in for longer exposure. Fair Start never got the same PM2.5 and NO2 model because the funding for that work ran out.

What the Age-3 Scores Did as Soot Fell

Mental Development Index scores rose at ages 1, 2, and 3 as the three pollutants fell. The age-3 exam is the one earlier work has found more stable, and more tied to later intellect, than the tests at 1 and 2. Columbia’s newsroom comparison of the first group with the latest group is the 26 percent figure. The journal paper itself reports inverse links to age-3 mental scores for each pollutant after the team adjusted for years of maternal schooling, child sex, gestational age, and a smoker in the home.

PRENATAL POLLUTANTS AND AGE-3 MDI

Among children who had both air data and a Bayley score, PM2.5 fell 0.607 micrograms per cubic meter per year (N=839) and NO2 fell 1.099 ppb per year. PAH fell 0.059 nanograms per cubic meter per year (N=979). The steepest particle and NO2 drop landed between 1998 and 2002, in the early years of the first cohort, then the slope flattened through 2016.

Inside the first cohort, a one-unit drop in PM2.5 lined up with a 0.677-point rise in the age-3 score (N=522). A one-unit drop in log PAH lined up with a 1.874-point rise (N=497). Combined across groups, the same inverse pattern held at ages 1 and 2, with smaller coefficients.

Bus, Taxi, and Boiler Rules Never Named the Fetus

The authors had already tied the exposure drop in these same families to a stack of city rules. None of those bills listed fetal brain development as a purpose. The paper names four: the Clean Fuel Bus Program, Clean Air Taxi legislation, the Clean Heat Program, and the Climate Mobilization Act. They could not pin the gain on a single statute. The cumulative effect is what they measured.

THE RULES THAT CUT THE PRENATAL DOSE

1. June 2000: The MTA sets up its Clean Fuel Bus program, buying cleaner diesel, rebuilding old engines, and moving the fleet onto ultra-low-sulfur fuel with particle filters. By 2003 every bus carries a filter.
2. July 2010: Local Law 43 requires 2 percent biofuel in heating oil and cuts the sulfur cap on No. 4 oil from 3,000 ppm to 1,500 ppm.
3. May 2011: A Department of Environmental Protection rule phases out No. 6 oil by 2015 and No. 4 oil by 2030, and forces a switch to No. 2 or gas when a boiler is replaced.
4. December 31, 2015: All 5,300 buildings off No. 6 oil, the city’s count on February 9, 2016. Models put the annual particle cut at 1,200 tons, with 210 premature deaths and 540 hospital stays avoided, plus 800,000 metric tons of carbon-dioxide equivalent. Northern Manhattan, northern Queens, and the South Bronx had the densest boiler conversions and the largest local air-quality gain.
5. 2019: The Climate Mobilization Act, led by Local Law 97, caps greenhouse gases from mid-size and large buildings and, with them, the co-emitted soot those boilers still throw.

Taxi law sits in the same stack. The city’s administrative code sets clean-air rules for medallion taxis, scoring vehicles on EPA pollution ratings and estimated carbon and stretching retirement dates for the cleaner cabs. The paper describes that package as a push toward hybrids, compressed natural gas, and other low-emission yellow cabs.

Enforcement on the boilers was not a pamphlet campaign. DEP staff ran nearly 4,000 inspections, wrote more than 2,000 notices of violation, and issued 162 cease-and-desist orders. The Clean Heat outreach desk helped more than 4,800 buildings leave No. 6 and nearly 1,000 leave No. 4. State rules also cut sulfur in No. 2 oil to 15 ppm, a 99 percent reduction.

Manhattan Borough President Gale A. Brewer, speaking when the No. 6 conversions closed, put the stake in one line: the city had a moral obligation to make sure kids grew up breathing cleaner air. The Columbia scores are the first 20-year test of whether that obligation reached the womb.

The First 727 Mothers Lived Through the Steepest Drop

The signal is not evenly spread across the 20 years. Significant inverse ties between prenatal PM2.5, NO2, and PAH and the age-3 score appear in the large opening cohort, more than 700 women enrolled from 1998 to 2006, the group that took the highest dose and then the largest decline. In the smaller Sibling group, which breathed cleaner air from the start, the PM2.5 links stayed inverse but lost statistical significance. NO2 did not show the same inverse pattern there.

That split is the piece a citywide “kids got smarter” headline flattens. If rising schooling or a gentler neighborhood were the whole story, the later groups, whose mothers had still more education, should have shown the cleaner, stronger slope. They showed a weaker one. The authors read that as support for the pollution drop itself. They also warn, in the same breath, that a non-significant result at today’s lower dose is not a safety finding.

They left “cohort” out of the main exposure-outcome models on purpose. Each later group is a proxy for both calendar time and cleaner air, so putting cohort in the equation over-adjusts and shrinks the pollution coefficient. When they added it anyway, the inverse shape survived and the numbers got smaller, as they expected.

Age 3, not ages 1 or 2, is where the first cohort’s PM2.5, NO2, and PAH links reach significance. The team had seen that age pattern before with PAH in these same families. Subtle prenatal injury can sit quiet until the test is stable enough to catch it.

A Coal Plant Shutdown in Tongliang Showed the Same Pattern

The New York series is a slow fade. The same laboratory already ran the abrupt version. In Tongliang, in Chongqing, a coal-fired power plant ran until May 2004. Nonsmoking women and their newborns were enrolled in 2002, while it still burned, and again in 2005, after it went dark. PAH-DNA adducts in cord blood fell. Head circumference rose. The Gesell developmental links that had tied adducts to worse motor and average scores in the 2002 group were gone in the 2005 group.

A follow-up found higher levels of mature BDNF, a protein the brain uses to wire itself, in the later Tongliang newborns, and better developmental quotients. Children with higher prenatal PAH had smaller heads at birth and slower growth; the longer a pregnant woman lived near the running plant, the sharper the hit to birth weight and height. Follow-up through age 5 tied prenatal PAH, with secondhand smoke, to IQ. In short, toddlers born after the plant closed did better on developmental tests than the children who had been in utero while it ran.

New York never switched a single stack overnight. It swapped bus fuel, taxi engines, and boiler oil over 15 years, then scored the children who were in utero while that swap ran. The direction matches Tongliang. The dose change is slower, and the test is a Bayley mental index rather than a Gesell quotient, but the stakeholder is the same: a fetus who does not vote and does not ride the bus.

This study in over a thousand children enrolled over a twenty-year period provides new evidence that policies to reduce fossil fuel emissions benefit developing brains and strongly supports continued efforts to reduce these emissions.

Frederica Perera, Professor Emerita of Environmental Health Sciences, Columbia Mailman School

Perera founded the center and directed translational research on this paper. Coauthors are Yuqi Miao, Shuang Wang, Molly Algermissen, and Julie Herbstman, the center’s director. The John and Wendy Neu Foundation, the New York Community Trust, and National Institutes of Health grant OD023290 paid for the work. The authors declared no competing interests.

The Authors Cannot Call Later Air Safe

The paper’s own limits sit in plain view. Sibling and Fair Start numbers with full covariates and air data are modest, so several later analyses simply could not be run. Fair Start lost backpack PAH sampling in March 2020 when COVID-19 stopped visits, and later lost Bayley testing when the protocol shifted and the money for the old scale ran out. Age-3 MDI in Fair Start was never modeled.

WHAT THE MODELS HOLD, AND WHAT THEY DO NOT

• Adjusted for: Years of maternal schooling, child sex, gestational age, a smoker in the household, and, in the trend models, maternal ethnicity. All three groups stayed inside the New York-Presbyterian system.
• Still open: Unmeasured neighborhood change and other secular trends that can lift test scores. The authors say that residual confounding is possible, as it is in any 20-year observational series.
• Not a safety line: A weaker, non-significant inverse slope at today’s lower dose does not mean current air is safe for the fetal brain.
• Who is in the sample: Direct export to other cities is hard. The authors note that many large U.S. cities have substantial Latina and African American populations and the same exposure gaps, which is the case they make for wider relevance.

Possible pathways listed in the paper are neuroinflammation, oxidative damage, and changes in synaptic plasticity or structure. Earlier center work had already tied prenatal PM2.5 to altered brain structure and function on MRI in these families. This analysis does not re-run the scanners. It asks whether the policy-era drop in dose shows up in the toddler scores, and it does, hardest in the children whose mothers inhaled the worst of it.

Frequently Asked Questions

What Does the Bayley Mental Development Index Measure?

The Mental Development Index is the cognitive scale of the Bayley Scales of Infant and Toddler Development, second edition. Scores are set to a mean of 100 and a standard deviation of 15 in the original norming sample, so 85 to 115 is the broad average band. Columbia kept BSID-II from 1998 onward because later Bayley editions change the scoring, and a mixed scale would have broken the 20-year comparison.

Which New York City Laws Does the Study Tie to the Air Drop?

The paper points to four packages rather than one bill: the MTA Clean Fuel Bus program, the city’s clean-air taxi statutes, the Clean Heat residual-oil phaseout, and the 2019 Climate Mobilization Act. A prior analysis by the same group, using all exposure data without requiring a Bayley score, is the source of the 51, 48, and 66 percent drops. The cognition paper does not claim any single law produced the mental-score gain.

Who Could Enroll in the Columbia CCCEH Cohorts?

Mothers and Newborns took self-identified African American and Hispanic women ages 18 to 35 who got care in the New York-Presbyterian system. Fair Start required age 18 or older and English or Spanish, and 90 percent of that group identified as Hispanic. The Sibling group is later children of Mothers and Newborns mothers, which is why the statistics treat family as a cluster.

Why Were the Later Cohorts Less Conclusive?

They were smaller, cleaner, and, in Fair Start’s case, missing whole chunks of air and test data after March 2020. The Sibling PM2.5 associations stayed inverse and lost significance. Fair Start had too few age-2 and age-3 Bayley exams for the PAH models. The authors treat that quieting as consistent with a lower dose, not as proof the remaining dose is harmless.

Fair Start’s Bayley line ended in 2020. The mothers who remain in the center’s later work will be scored on other instruments. The first 727 pairs, the ones who rode the steepest drop from depot diesel and No. 6 soot, already left their exam on the table.