Previous toxicity testing with fathead minnows (Pimephales promelas) indicated that some unsaturated acetylenic and allylic alcohols can be metabolically activated, via alcohol dehydrogenase, to highly toxic a, @-unsaturated aldehydes and ketones or allene derivatives. Although several in vivo and in vitro toxicological and biochemical endpoints can differentiate these alcohols by toxic mechanism, the use of stereoelectronic molecular descriptors to discriminate these toxicants, and subsequently to predict potency, had not been previously attempted. Exploration of several descriptors indicated that soft electrophilic characteristics of acetylenic or allylic moieties in the suspected metabolites unambiguously discriminated reactive and narcotic toxicants. The acute toxicity of alcohols acting as narcotics was accurately predicted using an existing quantitative structure-activity relationship for nonpolar narcosis. The toxicity of alcohols mediated by metabolic activation was estimated quantitatively using acceptor superdelocalizability of specific carbon atoms of the acetylenic or allylic moiety in the putative electrophilic intermediates. and tetany in fathead minnows exposed to those alcohols suspected of being activated [2], and the finding that these alcohols can be metabolized in vitro using both horse and rainbow trout liver preparations of alcohol dehydrogenase to reactive intermediates [4]. ' While previous work provides a mechanistic rationale for the observed toxic responses as well as the empirical toxicological and biochemical data to differentiate narcotic from nonnarcotic alcohols, identifying specific parameters to discriminate toxic mechanisms and establishing QSARs based on those mechanisms remains incomplete. The lack of a quantitative analysis can be attributed, in part, to the absence of a uniform approach to model in vivo electrophile reactivity. Hermans [5, 61 used the reactivity rate constants with model nucleophiles to predict the toxicity of electrophiyes. These empirical rate constants are useful; however, reactive species must first be classified as hard and soft reactants [7], and care is needed not to use a hard model nucleophile to simulate a soft electrophile.The unsaturated aldehyde and ketone metabolites in question, with alkene and alkyne bonds activated by an adjacent carbonyl, are soft electrophiles and have small rate constants (if measurable at all) with a hard nucleophile. They can also have immeasurably fast rate constants with soft nucleophiles such as thiols and amines [6, 8, 91.