The long-term focus of the McKim Conferences is the paradigm shift which is occurring in chemical risk assessments whereby the test-all-chemicals-for-all-hazards paradigm needs to evolve to a more efficient hypothesis-driven testing paradigm. This paradigm shift has already been called for by regulators in many countries, and the 2006 McKim Conference heard first-hand accounts of the scientific challenges involved from both the U.S. EPA and the European Commission. In brief, the Conference heard that the scientific foundation of the current testing paradigm must be extended to include predicting the behavior of untested chemicals with enough reliability to determine if testing is needed. The McKim Conference was convened to begin reshaping the foundation of risk assessment to include a new generation of models that extrapolate existing experimental data for tested chemicals to related untested chemicals, related species, or other hazard endpoints. QSAR, interspecies extrapolation models and biological systems models, respectively, serve as the primary elements new paradigm in risk assessment. The scientific barriers which stand between the special case models in the literature and the comprehensive models needed to support a new regulatory paradigm was primary topic of discussion at the 2006 McKim Conference. For the sake of simplicity, we have outlined four of these major barriers which will serve as the focus for future McKim Conferences. First, one major obstacle to maximizing the use of existing test data is that the data are scattered throughout the literature and private databases in inconsistent, non-digital formats. Second, the QSAR models to predict the intrinsic behavior of chemicals cannot adequately model reactive toxicity or hydrogen bonding at receptors. Third, interspecies correlation models do not uniformly model the metabolic differences of species or predict vulnerabilities of species. Fourth, systems biology and virtual animal models are in their infancy and do not adequately link a chemical perturbation of a system to the most important biological effect. Following two days of presentations and discussions, the participants of the 2006 McKim Conference provided a number of recommendations for next steps. The overwhelming reaction of the participants was that the topic being addressed is very broad and that future conferences should focus and prioritize topics of discussion. This recommendation is, itself, a major scientific challenge of scale because narrowing the focus creates models only useful in special cases while addressing the global perspective may require simplifications which do not apply to specific cases. To assist in planning, the participants suggested three general guidelines for setting priorities: identify the adverse outcomes which are the "big worries" to regulators, identify the adverse outcomes which have the best current understanding of the mechanisms involved, and identify the QSAR and other models which are currently most useful and build on them. The second major recommendation of the participants is that the McKim Conference should assist in setting a timeline for near-term and long-term research needs so that regulators and scientists recognize "what is needed when". Even if the science needed is extremely broad, we do not have to solve all aspects of the problem at once. In the near-term, the scientific community should be focused on building acceptance among regulators as well as focusing on chemicals where little or no data exists but regulatory decisions must be made. The third recommendation of the participants is that the McKim Conference should invite people with experience in the respective areas who might not be familiar with the more global perspective of the problem. The participants repeated that restricting the conference to the present-day experts is likely to lead to many of the same opinions that have perpetuated the test-everything-for-everything paradigm.