When performing screening level and baseline risk assessments, assessors usually compare estimated exposures of wildlife receptor species to toxicity reference values (TRVs). We modeled the exposure of American robins (Turdus migratorius) to ten elements (As, Cd, Cr, Cu, Hg, Mn, Pb, Se, Zn and V) in spring and early summer, a time when earthworms are preferred prey. We calculated soil benchmarks associated with possible toxic effects to these robins from six sets of published TRVs. Several of the resulting soil screening level benchmarks were inconsistent with each other and below soil background concentrations. Accordingly, we examined the derivations of the TRVs as a possible source of error. In the case of V, a particularly toxic chemical compound (ammonium vanadate) containing V, not normally present in soil, had been used to estimate a TRV. In the cases of Zn and Cu, use of uncertainty values of ten in estimating TRVs led to implausibly low soil screening values. In the case of Pb, a TRV was calculated from studies demonstrating reductions in egg production in Japanese quail (Coturnix coturnix japonica) exposed to Pb concentrations well below those causing toxic effects in other species of birds. The results on quail, which were replicated in additional trials, are probably not applicable to other, unrelated species, although we acknowledge that only a small fraction of all species of birds have been tested. These examples underscore the importance of understanding the derivation and relevance of TRVs before selecting them for use in screening or in ecological risk assessment. This article is protected by copyright. All rights reserved.