As New York's air got cleaner, children's cognitive scores rose. Proving it is harder

Air quality regulation is usually defended with mortality and asthma statistics. A study published on 15 August in Environmental Health asks a different question: whether the same policies show up in how children's brains develop.

Researchers used successive longitudinal birth cohorts run by the Columbia Center for Children's Environmental Health in New York City between 1998 and 2020 [s1]. Prenatal exposure to fine particulate matter (PM2.5), nitrogen dioxide and polycyclic aromatic hydrocarbons was measured — PM2.5 and NO2 in the first two cohorts covering 1998 to 2016, PAH across three cohorts covering 1998 to 2020 [s1]. Children were assessed at ages 1, 2 and 3 using the Bayley Scales of Infant and Toddler Development, yielding a Mental Development Index score [s1].

What moved

Across the cohorts, prenatal exposure to all three pollutants declined significantly, and Mental Development Index scores showed a significant upward trend [s1]. The analytic sample was 1,109 children [s1].

Pooling the cohorts, all three pollutants were significantly and inversely associated with Mental Development Index at ages 1, 2 and 3 — higher prenatal exposure, lower score [s1].

The cohort-specific results are where the structure of the finding shows. Significant inverse associations between prenatal PM2.5, NO2 and PAH and 3-year scores appeared in the largest and earliest cohort, of more than 700 children enrolled between 1998 and 2006 — the group with the highest exposures and the steepest subsequent decline [s1]. In the later cohort enrolled between 2008 and 2016, which experienced lower exposure throughout, the association between PM2.5 and cognitive development was weaker, though still inverse [s1].

Why that pattern is both encouraging and difficult

A weakening association at lower exposures is what you would expect if the relationship is real and the exposure range has moved down toward a less harmful part of it. It is also what you would expect if the association was never as strong as the early cohort suggested and the earlier result was partly driven by other differences between the 1998 cohort and the city around it.

The study design cannot fully separate those. Comparing cohorts recruited two decades apart means comparing eras — in schooling, in maternal health, in poverty rates, in the assessment context itself. The authors adjusted for demographic and other factors that contribute to neurodevelopment and that changed over time [s1], which is the correct response and not a complete one.

The paper's own claim is appropriately scoped: during the period of the largest reduction in air pollution levels previously associated with air pollution and climate policies enacted in New York City, children's cognitive test scores improved, providing new evidence that such policies are beneficial to children's neurodevelopment [s1]. That is a statement about co-occurrence with a plausible mechanism, in one city, in one set of related cohorts. It is not an experiment, and no observational study of this kind can be.

The pathway problem

A systematic review published a week later in Ageing Research Reviews addresses the other end of the lifespan and lands on the same weakness — the mechanism.

The review identified 16 studies examining 72 unique exposure-mediator-outcome associations across six pollutants: PM2.5, PM10, NO2, NOX, black carbon and PM1 components [s2]. PM2.5 dominated, accounting for 65% of analyses [s2]. Cognitive outcomes included memory, processing speed, and clinically diagnosed conditions such as dementia incidence [s2].

The candidate mediators span most of internal medicine: mental health outcomes including depression, anxiety and stress; lung function measured as FEV1, FVC and PEF; vascular disease including stroke and hypertension; inflammation measured by CRP; type 2 diabetes; sleep; and neuroanatomical changes [s2].

Cardiovascular and cerebrovascular disease emerged as a potential key mediator [s2]. Mental health outcomes, sleep and lung function also showed mediating potential, but the authors state that further research is needed to confirm those findings [s2].

Their assessment of the underlying literature is candid: the quality of mediation analysis reporting was generally good, but most studies carried some risk of bias, particularly in outcome assessment and confounding adjustment [s2]. The review calls for rigorous causal mediation analyses and standardised reporting [s2].

What the two papers say together

The association between ambient air pollution and cognitive function is observed repeatedly, at both ends of life, across pollutants and study designs. What remains thin is the causal chain — the specific biological route from an inhaled particle to a lower score or an earlier dementia diagnosis.

That gap has a practical consequence. Without a validated mediator, there is no way to check whether a given pollution reduction should have produced a given cognitive benefit, and no way to identify which populations would gain most from targeted intervention. It also leaves the field dependent on natural experiments like New York's, which are informative and are not designed.

What to watch

The most useful next studies are the ones the review specifies: properly conducted causal mediation analyses with standardised reporting, and outcome assessment less vulnerable to bias [s2]. On the child development side, the test is whether the pattern seen in the New York cohorts replicates in other cities whose air improved over the same decades, with different confounding structures. One city's cohorts, however carefully assembled, are one city's cohorts.