The quotient method (QM), a laboratory-based risk assessment methodology us
ed by the U.S. Environmental Protection Agency to evaluate pesticides for r
egistration and use, has not been thoroughly field-tested and its performan
ce has not always been reliable. Our objective was to compare predictions o
f risk using the QM with results from small mammal field studies to identif
y factors that cause variation in the performance of the QM. We regressed e
stimated mortality of herbivorous gray-tailed voles (Microtus canicaudus) a
nd omnivorous deer mice (Peromyscus maniculatus) against predictions of mor
tality based on LC50 probit analyses for pesticide concentrations in spray-
tank mixtures and residues on the top and bottom 15 cm of alfalfa plants. P
redictions of risk generally were correlated with estimated-mortality in th
e field, but slopes often differed significantly from the predicted 1:1 rel
ationship and were much greater than 1.0 for several relationships during t
he 1993 experiment. Risk predictions may be influenced by a number of facto
rs, including estimation of estimated environmental concentration (EEC), ve
getation structure, weather patterns, species of concern, and demographic c
haracteristics of the exposed populations. More accurate risk assessments m
ay be accomplished by improving vegetation nomogram estimates of EEC and by
incorporating vegetation structure and differences in natural history amon
g species into QM predictions. Weather following application and the demogr
aphic characteristics of the exposed populations are likely to remain unkno
wn and will cause imprecision in risk assessments.