Methylmercury is a neurotoxin at high exposures, and the developing fetus is particularly susceptible. Because exposure to methylmercury is primarily through fish, concern has been expressed that the consumption of fish by pregnant women could adversely affect their fetuses. The reference dose for methylmercury established by the U.S. Environmental Protection Agency was based on a benchmark analysis of data from a poisoning episode in Iraq in which mothers consumed seed grain treated with methylmercury during pregnancy. However, exposures in this study were short term and at much higher levels than those that result from fish consumption. In contrast, the Agency for Toxic Substances and Disease Registry (ATSDR) based its proposed minimal risk level on a no-observed-adverse-effect level (NOAEL) derived from neurologic testing of children in the Seychelles Islands, where fish is an important dietary staple. Because no adverse effects from mercury were seen in the Seychelles study, the ATSDR considered the mean exposure in the study to be a NOAEL. However, a mean exposure may not be a good indicator of a no-effect exposure level. To provide an alternative basis for deriving an appropriate human exposure level from the Seychelles study, we conducted a benchmark analysis on these data. Our analysis included responses from batteries of neurologic tests applied to children at 6, 19, 29, and 66 months of age. We also analyzed developmental milestones (age first walked and first talked). We explored a number of dose-response models, sets of covariates to include in the models, and definitions of background response. Our analysis also involved modeling responses expressed as both continuous and quantal data. The most reliable analyses were considered to be represented by 144 calculated lower statistical bounds on the benchmark dose (BMDLs; the lower statistical bound on maternal mercury hair level corresponding to an increase of 0.1 in the probability of an adverse response) derived from the modeling of continuous responses. The average value of the BMDL in these 144 analyses was 25 ppm mercury in maternal hair, with a range of 19 to 30 ppm.