BACKGROUND: The relationship between arsenic and birth weight is not well understood. The objective was to evaluate the causal relationship between prenatal arsenic exposure and birth weight considering the potential mediation effects of gestational age and maternal weight gain during pregnancy using structural equation models. METHODS: A prospectively enrolled cohort of pregnant women was recruited in Bangladesh from 2008 to 2011. Arsenic was measured in personal drinking water at the time of enrollment (gestational age <16 weeks, N = 1,140) and in toenails collected ≤1 month postpartum (N = 624) using inductively coupled plasma mass spectrometry. Structural equation models estimated the direct and indirect effects of arsenic on birth weight with gestational age and maternal weight gain considered as mediating variables. RESULTS: Every unit increase in natural log water arsenic was indirectly associated with decreased birth weight (ß = -19.17 g, 95% confidence interval [CI]: -24.64, -13.69) after adjusting for other risk factors. This association was mediated entirely through gestational age (ß = -17.37 g, 95% CI: -22.77, -11.98) and maternal weight gain during pregnancy (ß = -1.80 g, 95% CI: -3.72, 0.13). When exposure was modeled using toenail arsenic concentrations, similar results were observed. Every increase in natural log toenail arsenic was indirectly associated with decreased birth weight (ß = -15.72 g, 95% CI: -24.52, -6.91) which was mediated through gestational age (ß = -13.59 g, 95% CI: -22.10, -5.07) and maternal weight gain during pregnancy (ß = -2.13 g, 95% CI: -5.24, 0.96). CONCLUSION: Arsenic exposure during pregnancy was associated with lower birth weight. The effect of arsenic on birth weight appears to be mediated mainly through decreasing gestational age and to a lesser extent by lower maternal weight gain during pregnancy.