In conventional static or semi-static embryo toxicity assays with fish, the nominal concentrations of hydrophobic chemicals are often used to establish the toxic thresholds, which often far exceed the solubility limits of test compounds. Saturators and continuous-flow diluters have been used to provide stable concentrations below solubility but are complex, use large amounts of test substance, and produce large volumes of waste. We present a partition-controlled delivery (PCD) method that maintains the concentrations of chemicals in test solutions at or below solubility limits for extended exposure times. Concentrations are maintained by equilibrium partitioning of test chemicals from a series of poly(dimethylsiloxane) films loaded with a range of concentrations of each chemical. The efficacy of the PCD assay was tested by comparisons with static (no renewal) and semi-static (24-h renewal) embryo-larval toxicity tests. The test species was Japanese medaka (Oryzias latipes) exposed to retene (7-isopropyl-1-methylphenanthrene), a compound causing blue sac disease (BSD) in fish embryos. In the PCD assay, the median effective concentration (EC50) for BSD was 10 microg/L, below retene's solubility of 17 microg/L. In contrast, the nominal EC50 values for the semi-static 24-h and static assays were about 10 (150 microg/L) and 150 times (2500 microg/L) greater than solubility, respectively. The PCD method is a more sensitive and realistic method for assessing toxicity of nonpolar compounds than (semi)-static assays.