Background In recent studies in Bangladesh and elsewhere, exposure to arsenic (As) via drinking water is negatively associated with performance-related aspects of child intelligence (e.g., Perceptual Reasoning, Working Memory) after adjustment for social factors. Because findings are not easily generalizable to the US, we examine this relation in a US population. Methods In 272 children in grades 3–5 from three Maine school districts, we examine associations between drinking water As (WAs) and intelligence (WISC-IV). Results On average, children had resided in their current home for 7.3 years (approximately 75% of their lives). In unadjusted analyses, household well WAs is associated with decreased scores on most WISC-IV Indices. With adjustment for maternal IQ and education, HOME environment, school district and number of siblings, WAs remains significantly negatively associated with Full Scale IQ and Perceptual Reasoning, Working Memory and Verbal Comprehension scores. Compared to those with WAs < 5 μg/L, exposure to WAs ≥ 5 μg/L was associated with reductions of approximately 5–6 points in both Full Scale IQ (p < 0.01) and most Index scores (Perceptual Reasoning, Working Memory, Verbal Comprehension, all p’s < 0.05). Both maternal IQ and education were associated with lower levels of WAs, possibly reflecting behaviors (e.g., water filters, residential choice) limiting exposure. Both WAs and maternal measures were associated with school district. Conclusions The magnitude of the association between WAs and child IQ raises the possibility that levels of WAs ≥ 5 μg/L, levels that are not uncommon in the United States, pose a threat to child development.