New EU legislation is providing an impetus for research aimed at replacing acute fish toxicity testing with in vitro alternatives. In line with such research, the objective of this study was to determine what factors influence the correlation between in vitro and fish toxicity data. Basal cytotoxicity (IC₅₀) and acute toxicity data from fathead minnow (LC₅₀) of 82 industrial organic chemicals were obtained from the Halle Registry of Cytotoxicity and the US EPA Fathead Minnow Database. A good correlation between IC₅₀ with LC₅₀ data was found (r 0.84). Yet, IC₅₀ data were less sensitive than LC₅₀ data by an order of magnitude. Using multiple regression analysis, the octanol–water partition coefficient (K_OW) and the Henry’s Law Constant (H) were found to significantly explain the low absolute sensitivity. The mode of action (MOA) of the chemical was found to significantly explain the general variation in the log IC₅₀/log LC₅₀ regression line. These results support the notion that (a) the bioavailability of hydrophobic (high K_OW) and volatile (high H) chemicals is significantly lower in in vitro assays than in the fish bioassay and (b) multiple cell types and endpoints should be included to mimic the modes of action possible in the whole organism.