The introduction of species sensitivity distribution (SSD) approaches to ecological risk assessment offers the potential for a more transparent scientific basis for the derivation of predicted no-effect concentrations. However, conventional SSD methodologies have relied on standard distributions (e.g., log logistic, log normal) that are not necessarily based on sound ecological or statistical grounds. More recently, bootstrap resampling techniques that do not rely on distributional assumptions have been applied to the problem. Here we describe how a more advanced bootstrap methodology may be applied to derive better point estimates and confidence intervals for SSD estimates of safe environmental concentrations. Motivated by the fact that the true SSD may not fit any standard model category, we go on to consider a hybrid bootstrap regression approach. This can yield a substantially different estimate for the SSD when compared with both the basic bootstrap and the more frequently used parametric curve approaches. With increasing use of SSDs in ecological risk assessment, it is now imperative that the scientific community develops agreement over appropriate methods for their derivation.