Acrylamide (ACR) is used in the manufacture of polyacrylamides and has recently been shown to form when foods, typically containing certain nutrients, are cooked at normal cooking temperatures (e.g., frying, grilling or baking). The toxicity of ACR has been extensively investigated. The major findings of these studies indicate that ACR is neurotoxic in animals and humans, and it has been shown to be a reproductive toxicant in animal models and a rodent carcinogen. Several reviews of ACR toxicity have been conducted and ACR has been categorized as to its potential to be a human carcinogen in these reviews. Allowable levels based on the toxicity data concurrently available had been developed by the U.S. EPA. New data have been published since the U.S. EPA review in 1991. The purpose of this investigation was to review the toxicity data, identify any new relevant data, and select those data to be used in dose-response modeling. Proposed revised cancer and noncancer toxicity values were estimated using the newest U.S. EPA guidelines for cancer risk assessment and noncancer hazard assessment. Assessment of noncancer endpoints using benchmark models resulted in a reference dose (RfD) of 0.83 microg/kg/day based on reproductive effects, and 1.2 microg/kg/day based on neurotoxicity. Thyroid tumors in male and female rats were the only endpoint relevant to human health and were selected to estimate the point of departure (POD) using the multistage model. Because the mode of action of acrylamide in thyroid tumor formation is not known with certainty, both linear and nonlinear low-dose extrapolations were conducted under the assumption that glycidamide or ACR, respectively, were the active agent. Under the U.S. EPA guidelines (2005), when a chemical produces rodent tumors by a nonlinear or threshold mode of action, an RfD is calculated using the most relevant POD and application of uncertainty factors. The RfD was estimated to be 1.5 microg/kg/day based on the use of the area under the curve (AUC) for ACR hemoglobin adducts under the assumption that the parent, ACR, is the proximate carcinogen in rodents by a nonlinear mode of action. When the mode of action in assumed to be linear in the low-dose region, a risk-specific dose corresponding to a specified level of risk (e.g., 1 x 10-5) is estimated, and, in the case of ACR, was 9.5 x 10-2 microg ACR/kg/day based on the use of the AUC for glycidamide adduct data. However, it should be noted that although this review was intended to be comprehensive, it is not exhaustive, as new data are being published continuously.