Environmental toxins and IQ: what each exposure shows, study by study

0 The short answer

Lead is the exposure with the most pooled analysis and population modeling behind it, while for air pollution, pesticides, mercury, PCBs, manganese, arsenic, flame retardants and fluoride the evidence is real but smaller, newer or more contested, and the estimates use incompatible units. An international pooled analysis of 1,333 children estimated a 6.9 point difference in IQ between blood lead of 2.4 and 30 micrograms per deciliter, and a population model attributed about 824 million lost IQ points to lead among Americans as of 2015. Those are group estimates from observational designs and models; they describe populations and doses, not what happened to any one child.

1 How Do Researchers Measure What a Toxin Does to IQ?

Researchers measure a toxin's effect on IQ by relating a biomarker or environmental measurement taken in pregnancy or childhood to an IQ test given years later, and the unit of the answer changes from study to study. The exposure is measured in blood, urine, hair, umbilical cord plasma, tap water or personal air monitors. The outcome is usually a Wechsler or Stanford-Binet scale given between ages 3 and 13 (the Dunedin lead study tested adults at 38), and results are written in IQ points, conventionally a scale with a mean of 100 and a standard deviation of 15, which the page on the standard deviation of 15 explains. The designs that supply the numbers fall into four families, and they carry different weight.

The first family is the prospective birth cohort. Mothers or infants are enrolled, exposure is measured before the outcome exists, and children are followed to school age. The Dunedin cohort in New Zealand, the Columbia cohort in New York and the Salinas cohort of farmworker families in California are of this kind. The second is the cross-sectional comparison, in which exposure and IQ are measured at about the same time, as in the manganese and arsenic studies below; it cannot establish that exposure came first. The third is the pooled analysis or meta-analysis, which combines studies to estimate a dose-response curve or an average slope. The fourth is the population model, which takes an estimated slope and multiplies it by an estimated distribution of exposure to produce a national or global total. A model of the fourth kind inherits every assumption of the slope it uses.

The same exposure also appears in different units. A lead estimate is a difference across a range of blood concentrations. A pesticide estimate may compare the highest and lowest fifths of a metabolite distribution. A mercury estimate is a slope per part per million in maternal hair. An air pollution estimate is a slope per microgram per cubic meter. A manganese estimate is a change per tenfold increase in tap water concentration. These units cannot be placed on one axis, so this page does not rank the exposures by "worst." It reports each one in its own unit and flags the cases where a conversion would be our arithmetic rather than a finding.

Three difficulties recur across the literature. Exposure is hard to measure for the period that matters, so a single urine or blood sample may misrepresent a pregnancy or a childhood. Confounding is persistent: families with higher exposure often differ in income, parental education, parental IQ and the quality of the home, which is why careful cohorts adjust for maternal intelligence and a home environment score, and why the page on poverty and IQ matters to any reading of this one. And exposures cluster, so a child living near a busy road, an old house or a contaminated well may carry several at once. Where a study tested for interactions or adjusted for a second exposure, this page says so. Where it did not, a single estimate should be read as an association that remains open to explanation.

IQ itself has limits as an outcome. A total score combines several broad abilities, as the page on what IQ measures sets out, and a toxin can be associated with one ability, such as working memory or perceptual reasoning, more than another. Several studies below report index scores for that reason. The page on reliability and validity explains why a one or two point difference between group means is a statement about averages that no single test administration can resolve in one child.

2 The Evidence by Exposure: Design, Sample, Unit and Result

The table below gives the design, sample, exposure measure and result for each study this page relies on, in each study's own unit, and it should be read across a row rather than down a column. The rows are not ranked. A tenfold increase, a quintile contrast and a slope per part per million are different quantities, and several rows report no IQ point figure at all because the source did not.

3 How Much Does Lead Lower a Child's IQ?

The best known estimate comes from a pooled analysis that put the difference at 6.9 IQ points between concurrent blood lead of 2.4 and 30 micrograms per deciliter, with the steepest part of the curve at the low end. The 2005 pooled analysis by Lanphear and colleagues combined seven international population-based longitudinal cohorts, 1,333 children followed from birth or infancy until 5 to 10 years of age, with Full Scale IQ as the primary outcome. The geometric mean blood lead concentration of the children peaked at 17.8 micrograms per deciliter and declined to 9.4 by ages 5 to 7. Of the 1,333 children, 244 (18 percent) never exceeded 10 micrograms per deciliter and 103 (8 percent) never exceeded 7.5.

The three segments add up to the whole, and the first accounts for about 57 percent of it (3.9 of 6.9, our arithmetic). The authors also reported that, for a given increase in blood lead, the decrement was significantly greater among children whose maximal level stayed below 7.5 micrograms per deciliter than among the others (p = 0.015), and they concluded that exposure in that range is associated with intellectual deficits. A pooled log-linear curve is a statistical summary of seven cohorts that used different tests and sampling; the abstract reports the shape and does not explain it.

An earlier single cohort pointed the same way. In Canfield and colleagues' 2003 study, 172 children had blood lead measured repeatedly from 6 to 60 months and took the Stanford-Binet at 3 and 5 years, with adjustment for maternal IQ and the quality of the home environment. In the linear model each 10 micrograms per deciliter of lifetime average blood lead was associated with a 4.6 point lower IQ (p = 0.004). For the 101 children whose maximal level stayed below 10, the change per unit of lead was greater. In the nonlinear model, IQ declined 7.4 points as lifetime average blood lead rose from 1 to 10 micrograms per deciliter.

Whether the difference persists into adulthood was tested in the Dunedin cohort. Reuben and colleagues followed a 1972 to 1973 birth cohort in New Zealand to age 38. Of 1,007 participants alive at 38, 565 (56 percent) had blood lead measured at age 11, with a mean of 10.99 micrograms per deciliter. After adjusting for maternal IQ, childhood IQ and childhood socioeconomic status, each 5 micrograms per deciliter higher blood lead in childhood was associated with a 1.61 point lower adult IQ on the WAIS-IV (95 percent confidence interval 0.74 to 2.48 points lower). The association was 2.07 points for perceptual reasoning and 1.26 for working memory, and it was not statistically significant for verbal comprehension or processing speed. The authors noted that high blood lead was found among children of all socioeconomic levels in this cohort, which matters because lead and poverty often travel together, as the page on poverty and IQ discusses.

On the public health side, the U.S. Centers for Disease Control and Prevention updated its blood lead reference value to 3.5 micrograms per deciliter on October 28, 2021. The page defines it as the 97.5th percentile of the blood lead distribution in U.S. children ages 1 to 5 in the 2015 to 2018 national survey cycles, and states that it is not a health-based standard or a toxicity threshold. The same page states that no safe level of lead in children's blood has been identified. The reference value identifies children whose level is higher than that of most; it is not a cutoff below which a score is unaffected. The page on what a low IQ is and its documented causes places lead beside other documented factors.

4 What Does the 824 Million IQ Points Figure Mean?

The 824,097,690 figure is a model's estimate of a population total, produced by combining estimated historical exposure with an assumed relation between lead and IQ, and it is neither a measured loss nor an average per person. McFarland, Hauer and Reuben published it in the Proceedings of the National Academy of Sciences in 2022 under the title "Half of US population exposed to adverse lead levels in early childhood." Their summary estimates that over 170 million Americans alive today were exposed to high lead levels in early childhood, several million of them to five or more times the current reference level, and that lead is responsible for the loss of 824,097,690 IQ points as of 2015. The summary does not state the dose-response function behind the total, so a reader cannot judge its uncertainty from the abstract alone; any model of this kind inherits the uncertainty of both its exposure estimates and its slope.

A second model works at world scale. Larsen and Sánchez-Triana used country blood lead estimates from the Global Burden of Diseases study for 2019 and the blood lead to IQ function from the international pooled analysis described above. They estimated that children younger than 5 lost 765 million IQ points globally in 2019 (95 percent confidence interval 443 to 1,098 million), of which 729 million (95.3 percent) occurred in low and middle income countries, and that the loss in those countries was nearly 80 percent higher than a previous estimate. Their cost estimate was 6.0 trillion dollars, or 6.9 percent of global gross domestic product, of which 23 percent was the present value of lost future income from IQ loss and 77 percent was the welfare cost of cardiovascular deaths attributed to lead. The authors note that blood lead measurement data in low and middle income countries still need improvement.

The two totals cannot be added or compared head to head. One counts Americans of all ages alive in 2015 against their childhood exposure; the other counts children under 5 in a single year across the world. A total is large because many people each carry a small shift, not because any one person lost hundreds of points. David Bellinger's 2012 paper makes this point directly: dismissing chemical effects as clinically insignificant fails to distinguish individual risk from population risk, because the population impact depends on both the size of the effect and how many people are exposed. Estimating total Full Scale IQ points lost among U.S. children from birth to 5, he concluded that methylmercury, organophosphate pesticides and lead make contributions that exceed those of many nonchemical risk factors. The page on the Flynn effect covers population level IQ trends over time, where a shift of a few points across a whole population is the normal unit of discussion.

5 Does Air Pollution Lower Children's IQ?

Air pollution studies point the same direction, but they measure different things: one cohort used personal monitors for polycyclic aromatic hydrocarbons in pregnancy, and one meta-analysis pooled six studies of fine particulate matter. In the Columbia cohort, Perera and colleagues followed children of nonsmoking Black or Dominican-American women in New York City from before birth to age 5, measuring each mother's exposure to airborne polycyclic aromatic hydrocarbons (PAHs) with a personal monitor during pregnancy. At 5 years, 249 children took the Wechsler Preschool and Primary Scale of Intelligence-Revised. After adjustment for maternal intelligence, the quality of the home caretaking environment, environmental tobacco smoke and other factors, children above the median exposure of 2.26 nanograms per cubic meter scored 4.31 points lower on Full Scale IQ (p = 0.007) and 4.67 points lower on Verbal IQ (p = 0.003). With exposure treated as a continuous, log transformed variable, the Full Scale difference was 3.00 points per unit (p = 0.009).

For fine particulate matter (PM2.5), Alter, Whitman, Bellinger and Landrigan searched seven databases for studies relating exposure before or after birth to childhood IQ. Of 1,107 unique publications, six met the inclusion criteria, together representing 4,860 children across North America, Europe and Asia. The mean PM2.5 concentration across the studies was 30.4 micrograms per cubic meter with a standard deviation of 24.4, exposure timing ranged from the prenatal period to mid-childhood, and children were 8.9 years old on average when tested. Each 1 microgram per cubic meter increase was associated with a 0.27 point lower Full Scale IQ (p < 0.001), a 0.39 point lower performance IQ (p = 0.003) and a 0.24 point lower Verbal IQ (p = 0.021). The authors identified the performance score as the most affected.

A slope this small is easier to read when multiplied, but multiplying is our arithmetic and no author reports it: if the slope were exactly linear, a 10 microgram per cubic meter difference would correspond to about 2.7 Full Scale points. That assumption is the weak point. The six studies came from different places and exposure windows, and the large standard deviation shows that concentrations differed widely between them. Readers who meet a different figure for the same review in a summary should check which IQ composite it refers to; the abstract prints 0.27 for Full Scale IQ and 0.39 for performance IQ. Air quality also tracks where a family lives, and where a family lives is tied to income and other exposures, so the adjustment sets in each study carry much of the weight. PAHs and PM2.5 are different measurements, and a PAH finding should not be read as a PM2.5 finding or the reverse.

6 Do Organophosphate Pesticides Lower IQ?

Three cohort studies published in the same journal issue in 2011 linked prenatal organophosphate exposure to lower scores at school age, using different cities, different biomarkers and different ways of stating the effect, so their numbers should not be merged. All three appeared in Environmental Health Perspectives, volume 119, issue 8. Organophosphates are a class of pesticides, and the studies measure exposure either through a specific compound or through urinary breakdown products.

In New York, Rauh and colleagues measured the pesticide chlorpyrifos in umbilical cord blood plasma for 265 children in the Columbia Center for Children's Environmental Health cohort, then tested them with the WISC-IV at 7. For each standard deviation increase in exposure (4.61 picograms per gram), Full Scale IQ declined by 1.4 percent and working memory by 2.8 percent. Covariates were maternal education, maternal IQ and the quality of the home environment, and the authors found no significant interactions with prenatal tobacco smoke or PAHs. A percentage change in a score is not a point change; for a score near 100, 1.4 percent is roughly 1.4 points (our arithmetic).

In California, Bouchard and colleagues studied predominantly Latino farmworker families in an agricultural community. Dialkyl phosphate (DAP) metabolites were measured in urine during pregnancy and from the children at 6 months and at 1, 2, 3.5 and 5 years, and 329 children took the WISC-IV at 7. Analyses were adjusted for maternal education and intelligence, a home environment score and language of testing. Averaged maternal DAP concentrations during pregnancy were associated with poorer working memory (Gwm), processing speed (Gs), verbal comprehension, perceptual reasoning and Full Scale IQ, and children in the highest fifth of maternal concentrations had an average deficit of 7.0 IQ points compared with the lowest fifth. The children's own urinary concentrations were not consistently associated with scores, and the authors noted that maternal levels were higher than in other studies but within the range measured in the general U.S. population. The page on the CHC model explains the abilities behind these index names.

In the Mount Sinai cohort in New York, Engel and colleagues enrolled 404 women between 1998 and 2002, analyzed third-trimester urine samples for 360, and tested children at 12 months (200 children), 24 months (276) and 6 to 9 years (169). They also measured the mother's genotype for paraoxonase 1 (PON1), an enzyme involved in metabolizing organophosphates. In later childhood, higher prenatal dialkylphosphate levels were associated with decrements in perceptual reasoning among children of mothers with the QQ genotype, with a monotonic trend; the abstract gives no IQ point figure. Only 169 of 404 enrolled women (about 42 percent, our arithmetic) had a child tested at school age, so loss to follow-up is a question for that estimate, and a result limited to one genotype group is a subgroup finding.

The 7.0 points in the California study contrasts the top and bottom fifths of a biomarker distribution, while the New York study states a percentage change per standard deviation. They are different quantities and neither is "the effect of pesticides." What they share is the design: prenatal exposure measured before the outcome, school-age testing and adjustment for maternal IQ and the home environment, which addresses the confounders that most often explain apparent environmental effects.

7 What Do the Mercury and PCB Studies Show?

Mercury and PCB studies are the clearest cases where cohorts disagree or where the exposure arrives bundled with something else, so the honest summary is a range and a caveat rather than a single number. For methylmercury, the question is whether prenatal exposure from fish lowers IQ at the levels people actually encounter. Axelrad and colleagues combined dose-response coefficients from three cohorts, in the Faroe Islands, New Zealand and the Seychelles, using a Bayesian hierarchical model. Their central estimate was 0.18 IQ points lower for each part per million increase in maternal hair mercury (95 percent confidence interval 0.009 to 0.378 points lower). The estimate was similar to those for the Faroe Islands and Seychelles studies and lower in magnitude than that for the New Zealand study, and sensitivity analyses gave central estimates between 0.13 and 0.25. The authors added that IQ may not fully represent deficits in attention and motor skills.

The Seychelles cohort is the one that most directly addresses fish eating. In Myers and colleagues' report, 779 mother and infant pairs were followed, the mothers reporting about 12 fish meals per week. Mean prenatal exposure, measured in maternal hair growing during pregnancy, was 6.9 parts per million (standard deviation 4.5). When the children were 9, only two of the study's endpoints were associated with exposure, one a decreased grooved pegboard score with the non-dominant hand in males and the other an improved score on a teacher-rated hyperactivity index. The authors concluded that the data do not support a neurodevelopmental risk from prenatal methylmercury exposure resulting solely from ocean fish consumption. This page gives no advice about what to eat; it reports that the integrated analysis gave a small slope while the Seychelles report, taken alone, gave a null finding.

For polychlorinated biphenyls (PCBs), Jacobson and Jacobson tested 212 children at 11 who had been recruited as newborns to overrepresent infants of women who had eaten Lake Michigan fish contaminated with PCBs. Prenatal exposure was a composite of PCB concentrations in umbilical cord serum, maternal serum and milk, and the maternal serum and milk concentrations at delivery were slightly higher than in the general population. After control for confounders such as socioeconomic status, prenatal exposure was associated with lower Full Scale and Verbal IQ (p = 0.02), with the strongest effects on memory and attention. The most highly exposed children were three times as likely to have low average IQ scores (p < 0.001) and twice as likely to be at least two years behind in reading comprehension (p = 0.03). Deficits were associated only with exposure before birth and not with exposure through breast milk. The abstract gives no IQ point difference, so this study has no row in a comparison of points.

8 What About Manganese, Arsenic and Flame Retardants?

Manganese and arsenic estimates come from cross-sectional studies, and the flame retardant estimate from four pooled studies with a wide interval, so each supports an association but leaves the order of events and the dose unresolved. Bouchard and colleagues studied 362 children aged 6 to 13 living in communities supplied by groundwater. Manganese was measured in home tap water (median 34 micrograms per liter, range 1 to 2,700) and in the children's hair, and IQ was measured with the Wechsler Abbreviated Scale of Intelligence. After adjustment for maternal intelligence, family income and other confounders, a tenfold increase in tap water manganese was associated with 2.4 fewer IQ points (95 percent confidence interval 0.9 to 3.9 fewer), and children in the highest quintile scored 6.2 points below those in the lowest. The association was stronger for performance IQ than for verbal IQ, and hair manganese tracked intake from water but not from the diet. Because exposure and IQ were measured at about the same time, the study cannot show that exposure came first. The page on average IQ in Canada cites this study; this page gives its design.

Wasserman and colleagues studied 201 children aged 10 in Araihazar, Bangladesh, drawn from a prospective cohort of about 12,000 residents examining arsenic exposure. Arsenic and manganese were measured in the tube well at each child's home. Intellectual function was assessed with tests drawn from the WISC-III, and the scores are weighted raw sums for Verbal, Performance and Full Scale, not IQ points on the standard scale. After adjustment for sociodemographic covariates and water manganese, children whose water arsenic exceeded 50 micrograms per liter scored significantly lower on Performance and Full Scale than children below 5.5 micrograms per liter, in a dose-response pattern. The association was generally stronger for well water arsenic than for urinary arsenic.

For flame retardants, Lam and colleagues conducted a systematic review of developmental exposure to polybrominated diphenyl ethers (PBDEs). Fifteen studies met the inclusion criteria, 10 of them on intelligence. The authors rated the overall body of evidence as moderate in quality, with sufficient evidence for an association between PBDEs and IQ, and a meta-analysis of four studies estimated a 3.70 point decrement for each tenfold increase in exposure (95 percent confidence interval 0.83 to 6.56). The lower bound of that interval is under one point, so the data fit a small effect as well as a moderate one. Exposures also cluster: the Bangladesh study measured manganese in the same wells and adjusted for it. The page on stress and IQ covers a different set of exposures, those of the social environment, that can accompany the physical ones.

9 What Did the 2024 NTP Monograph and the 2025 Meta-Analysis Conclude About Fluoride?

The National Toxicology Program reported moderate confidence that fluoride exposure above 1.5 milligrams per liter in drinking water is associated with lower IQ in children, and it said the data were insufficient to decide whether the 0.7 milligram per liter level used for U.S. community water supplies has an effect. This section reports the evidence and the official statements with their dates. It takes no position on whether any community should fluoridate, and the page on fluoride and IQ covers the topic at greater length.

The NTP monograph was released in final form on August 21, 2024, after work that began in 2016. According to the program's fluoride page, the draft was peer reviewed by the National Academies of Sciences, Engineering, and Medicine, by other external experts and by experts in federal health agencies before a final review by the NTP Board of Scientific Counselors. The review evaluated studies published through October 2023. The program states that its review was designed to evaluate total fluoride exposure from all sources and not the effects of fluoridated drinking water alone; that the determination rested primarily on epidemiology studies in countries such as Canada, China, India, Iran, Pakistan and Mexico where some participants had total exposures above 1.5 milligrams of fluoride per liter of drinking water; that it found no evidence that fluoride exposure had adverse effects on adult cognition; that an association "does not prove a cause and effect"; and that the monograph does not assess fluoride's benefits. The program uses four confidence levels, high, moderate, low and very low, and assigned moderate to this finding.

The companion meta-analysis was published on January 6, 2025. Taylor and colleagues searched eight databases through October 2023 and included 74 studies of children's IQ: 64 cross-sectional and 10 cohort studies, 45 of them conducted in China. By the review's risk-of-bias tool, 52 studies were high risk and 22 low risk. Sixty-four studies reported inverse associations between fluoride exposure measures and children's IQ.

Two features of the results matter for how far they travel. First, in the drinking water analysis the association was null below 1.5 milligrams per liter when all studies were included, but it remained inverse at that exposure in the analysis restricted to low risk-of-bias studies; the authors concluded that data were limited and the dose-response uncertain for drinking water alone below 1.5 milligrams per liter. Second, urinary fluoride and drinking water fluoride are different measurements: 1 milligram per liter in urine is not 1 milligram per liter in tap water. If a standardized difference of 0.15 were expressed on a scale with a standard deviation of 15, it would be about 2.3 points (our arithmetic; the included studies used various tests).

10 How Has the Fluoride Meta-Analysis Been Criticized, and What Are the Official Positions?

The criticisms concern study design, risk of bias, exposure level and how exposure was measured, and the official positions on fluoridation have shifted during 2024 and 2025, so any statement about them needs a date. In the same issue of JAMA Pediatrics, two commentaries took different views: Levy's is titled "Caution Needed in Interpreting the Evidence Base on Fluoride and IQ," and Lanphear, Den Besten and Till's is titled "Time to Reassess Systemic Fluoride Exposure, Again." Their titles show the split; the abstracts are not available, and this page does not characterize their arguments beyond that.

Published letters raised specific questions, and the authors answered in a reply dated May 12, 2025. One letter asked whether alternative dose-response models and a threshold had been adequately considered, and another raised the inclusion of cross-sectional studies and other methodological points. The authors replied that they evaluated linear, quadratic and cubic spline models and that linear fit best, that no statistically significant association was found below 1.5 milligrams per liter in drinking water, and that excluding studies by their design label risks bias. The authors of the meta-analysis later answered a longer list of critiques in Annals of Global Health, on December 12, 2025, arguing that analyses restricted to high-quality studies still show the inverse association. Those authors work at the National Institute of Environmental Health Sciences and ICF, so that paper is a defense by the people who did the work, and it is reported as such. In the meta-analysis itself, 52 of 74 studies were rated high risk of bias and 64 were cross-sectional, which is the substance of the main objections to the pooled figures.

The official positions, as documented in a Congressional Research Service report dated May 16, 2025, are these. The U.S. Public Health Service recommends 0.7 milligrams per liter for community water fluoridation. The Environmental Protection Agency set an enforceable maximum contaminant level of 4.0 milligrams per liter in 1986 and a nonenforceable secondary level of 2.0 milligrams per liter against dental fluorosis, and the NTP page notes that the World Health Organization guideline is 1.5 milligrams per liter. On September 24, 2024, a federal district court in Food & Water Watch v. EPA found that fluoridation at 0.7 milligrams per liter presented an unreasonable risk of injury to health under the Toxic Substances Control Act and ordered the agency to initiate rulemaking; judgment was entered on November 20, 2024, and EPA appealed on January 17, 2025. A court finding is a legal determination under a statute, not a systematic review. In April 2025 the Secretary of Health and Human Services directed the end of the department's water fluoridation recommendation and the EPA Administrator announced a review of the science, and the report notes that no official agency statement on the HHS actions had been released as of its date. Utah's prohibition took effect on May 7, 2025, and Florida's governor signed legislation that month. This page does not track what has happened since; agency pages carry the current status.

11 Why Can a Few IQ Points in a Group Not Be Read as a Loss for One Child?

An effect of two to seven points in a group average tells you how much a population shifts, not how much any single child's score differs from what it would otherwise have been. Bellinger's 2012 paper frames the same distinction: the population impact of a risk factor depends on its effect size and on how many are exposed. A small shift per person can add up to a large total, and the same small shift is invisible inside the ordinary spread of scores in any one family.

The arithmetic shows the scale. On a scale with a mean of 100 and a standard deviation of 15, a 3 point shift moves a score from 100 to 97, or from the 50th percentile to about the 42nd, and the 6.9 point difference in the lead curve moves it to about the 32nd (our arithmetic, assuming a normal distribution; the page on IQ scores and percentiles explains the conversion). Scores vary between children for many reasons unrelated to any exposure, and the page on heritability and IQ explains why a group statistic about genes or environment is not a statement about one person. Individual scores also carry measurement error, which the page on reliability and validity describes and which is why a score is reported with an interval.

Three consequences follow. A score cannot reveal a child's exposure history: a lower than expected score has many explanations, and no IQ test can attribute a result to lead, air, a pesticide or any other single cause. An exposure does not predict a score: the studies report adjusted averages, and an average says nothing about where one child falls around it. And the studies do not say that a deficit is fixed. The Dunedin follow-up measured associations at 38, and the page on whether IQ can be improved sets out what is known about change within a person. Questions about a particular child's exposure belong with a pediatrician or the local health department, and this page gives no testing or treatment advice.

ACIS is an online, unsupervised assessment and not a clinical or diagnostic instrument. It reports a Full Scale IQ and six index scores with percentiles and a 95 percent confidence interval, and it cannot say what produced any pattern in them. The page on what IQ scores mean describes how to read a score without over reading it.

12 What the Evidence Supports, Stated Narrowly

The evidence supports these statements and no stronger ones. Each is limited to the design that stands behind it.

• Higher childhood blood lead is associated with lower IQ in seven pooled cohorts of 1,333 children, in a single cohort of 172 children, and in a Dunedin follow-up to age 38 of 565 participants. The pooled curve is steepest at low concentrations, and the estimated difference from 2.4 to 30 micrograms per deciliter is 6.9 points.
• The 824,097,690 and 765 million figures are modeled totals for populations, not measured losses and not averages per person.
• Prenatal air pollutants, organophosphate pesticide biomarkers and PCBs are associated with lower scores in prospective cohorts, with effects stated in different units that cannot be ranked against each other.
• Mercury results diverge across cohorts, and the integrated estimate is a small slope with a wide interval.
• Manganese and arsenic findings come from cross-sectional studies, and the arsenic study reports raw scores, not IQ points.
• For fluoride, the NTP reported moderate confidence in an association at exposures above 1.5 milligrams per liter in drinking water and found the data insufficient for 0.7 milligrams per liter; the 2025 meta-analysis rests mostly on cross-sectional studies, most rated high risk of bias, and its critics and authors disagree about what follows.
• None of this identifies the cause of any one person's score, and none of it supports advice about testing or treatment.

The Standards for Educational and Psychological Testing (AERA, APA and NCME, 2014) are explicit that an interpretation of a score must be supported by evidence for the specific use, and a group association between an exposure and a mean score does not supply that evidence for attributing one person's result to an exposure. The APA Ethical Principles of Psychologists and Code of Conduct, Standard 9.06, asks psychologists to interpret results in light of the purpose of the assessment and the characteristics of the person assessed. Both point to the same limit: an exposure study describes a population, and an IQ score describes a person on one occasion, with an interval around it. The page on how IQ scores are normed against a reference group explains what a score is compared with.

13 Sources Behind This Page

These sources are grouped by design: pooled and cohort analyses, population models, systematic reviews, official statements and the published debate, with each figure given as its source prints it. Where a figure is our arithmetic the body says so. Journal records were checked through Crossref and PubMed, and the web pages were opened on October 7, 2026.

• Lanphear B, Hornung R, Khoury J, Yolton K, Baghurst P, Bellinger D, Canfield R, Dietrich K and colleagues. Low-Level Environmental Lead Exposure and Children's Intellectual Function: An International Pooled Analysis. Environmental Health Perspectives, 2005, volume 113, issue 7, pages 894 to 899.
• Canfield R, Henderson C, Cory-Slechta D, Cox C, Jusko T and Lanphear B. Intellectual Impairment in Children with Blood Lead Concentrations below 10 micrograms per Deciliter. New England Journal of Medicine, 2003, volume 348, issue 16, pages 1517 to 1526.
• Reuben A, Caspi A, Belsky D, Broadbent J, Harrington H, Sugden K, Houts R, Ramrakha S and colleagues. Association of Childhood Blood Lead Levels With Cognitive Function and Socioeconomic Status at Age 38 Years and With IQ Change and Socioeconomic Mobility Between Childhood and Adulthood. JAMA, 2017, volume 317, issue 12, pages 1244 to 1251.
• Centers for Disease Control and Prevention. Update to the blood lead reference value (page published May 15, 2024; last reviewed August 25, 2026 as printed), opened October 7, 2026.
• McFarland M, Hauer M and Reuben A. Half of US population exposed to adverse lead levels in early childhood. Proceedings of the National Academy of Sciences, 2022, volume 119, issue 11, article e2118631119.
• Larsen B and Sánchez-Triana E. Global health burden and cost of lead exposure in children and adults: a health impact and economic modelling analysis. The Lancet Planetary Health, 2023, volume 7, issue 10, pages e831 to e840.
• Bellinger D. A Strategy for Comparing the Contributions of Environmental Chemicals and Other Risk Factors to Neurodevelopment of Children. Environmental Health Perspectives, 2012, volume 120, issue 4, pages 501 to 507.
• Perera F, Li Z, Whyatt R, Hoepner L, Wang S, Camann D and Rauh V. Prenatal Airborne Polycyclic Aromatic Hydrocarbon Exposure and Child IQ at Age 5 Years. Pediatrics, 2009, volume 124, issue 2, pages e195 to e202; and Alter N, Whitman E, Bellinger D and Landrigan P. Quantifying the association between PM2.5 air pollution and IQ loss in children: a systematic review and meta-analysis. Environmental Health, 2024, volume 23, issue 1, article 101.
• Rauh V, Arunajadai S, Horton M, Perera F, Hoepner L, Barr D and Whyatt R. Seven-Year Neurodevelopmental Scores and Prenatal Exposure to Chlorpyrifos, a Common Agricultural Pesticide. Environmental Health Perspectives, 2011, volume 119, issue 8, pages 1196 to 1201; Bouchard M, Chevrier J, Harley K, Kogut K, Vedar M, Calderon N, Trujillo C, Johnson C and colleagues. Prenatal Exposure to Organophosphate Pesticides and IQ in 7-Year-Old Children. Environmental Health Perspectives, 2011, volume 119, issue 8, pages 1189 to 1195; and Engel S, Wetmur J, Chen J, Zhu C, Barr D, Canfield R and Wolff M. Prenatal Exposure to Organophosphates, Paraoxonase 1, and Cognitive Development in Childhood. Environmental Health Perspectives, 2011, volume 119, issue 8, pages 1182 to 1188.
• Axelrad D, Bellinger D, Ryan L and Woodruff T. Dose-Response Relationship of Prenatal Mercury Exposure and IQ: An Integrative Analysis of Epidemiologic Data. Environmental Health Perspectives, 2007, volume 115, issue 4, pages 609 to 615; and Myers G, Davidson P, Cox C, Shamlaye C, Palumbo D, Cernichiari E, Sloane-Reeves J, Wilding G and colleagues. Prenatal methylmercury exposure from ocean fish consumption in the Seychelles child development study. The Lancet, 2003, volume 361, issue 9370, pages 1686 to 1692.
• Jacobson J and Jacobson S. Intellectual Impairment in Children Exposed to Polychlorinated Biphenyls in Utero. New England Journal of Medicine, 1996, volume 335, issue 11, pages 783 to 789.
• Bouchard M, Sauvé S, Barbeau B, Legrand M, Brodeur M, Bouffard T, Limoges E, Bellinger D and Mergler D. Intellectual Impairment in School-Age Children Exposed to Manganese from Drinking Water. Environmental Health Perspectives, 2011, volume 119, issue 1, pages 138 to 143; and Wasserman G, Liu X, Parvez F, Ahsan H, Factor-Litvak P, van Geen A, Slavkovich V, LoIacono N and colleagues. Water Arsenic Exposure and Children's Intellectual Function in Araihazar, Bangladesh. Environmental Health Perspectives, 2004, volume 112, issue 13, pages 1329 to 1333.
• Lam J, Lanphear B, Bellinger D, Axelrad D, McPartland J, Sutton P, Davidson L, Daniels N, Sen S and Woodruff T. Developmental PBDE Exposure and IQ/ADHD in Childhood: A Systematic Review and Meta-analysis. Environmental Health Perspectives, 2017, volume 125, issue 8, article 086001.
• National Toxicology Program. NTP Monograph on the State of the Science Concerning Fluoride Exposure and Neurodevelopment and Cognition: A Systematic Review. NTP Monograph 08, August 21, 2024; and the program's fluoride assessment page, opened October 7, 2026.
• Taylor K, Eftim S, Sibrizzi C, Blain R, Magnuson K, Hartman P, Rooney A and Bucher J. Fluoride Exposure and Children's IQ Scores: A Systematic Review and Meta-Analysis. JAMA Pediatrics, 2025, volume 179, issue 3, pages 282 to 292; with Levy S, Caution Needed in Interpreting the Evidence Base on Fluoride and IQ, pages 231 to 234; Lanphear B, Den Besten P and Till C, Time to Reassess Systemic Fluoride Exposure, Again, pages 234 to 236; the authors' reply to letters, JAMA Pediatrics, May 12, 2025, volume 179, issue 7; and Taylor K, Eftim S, Sibrizzi C, Blain R, Magnuson K, Hartman P, Bucher J and Rooney A, Addressing Critiques of the Evidence Linking Fluoride and Children's IQ, Annals of Global Health, 2025, volume 91, issue 1, article 83.
• Congressional Research Service. The Development of Federal Recommendations and Regulations for Fluoride in Drinking Water, report R48539, May 16, 2025, opened October 7, 2026.

The interpretive framework is the Standards for Educational and Psychological Testing (AERA, APA and NCME, 2014) and the APA Ethical Principles of Psychologists and Code of Conduct, effective January 1, 2017, Standard 9.06.

14 Frequently Asked Questions