No-effect concentrations (NECs) for toxicants are of interest from a biological and a legislation point of view. Using artificial, but typical, examples of the results of a bioassay on survival and two different models for the concentration-effect relationship, we show that the likelihood based confidence set of the NEC as parameter of the hazard model has quite acceptable statistical properties. Contrary to the hazard model, the NEC of the standard log-logistic model did not differ significantly from zero.