We have proposed a three step model for the specification of left-right in mammalian embryos. The fundamental assumption is that handedness is imparted by an asymmetrical molecule. Conversion of molecular asymmetry to the cellular level gives a property to one side of the embryo to bias an otherwise random generation of an asymmetrical gradient which can be interpreted by developing organs. Rat embryos, treated at discrete stages, show a window of sensitivity for disruption of handedness, which may reflect the time of conversion/biasing. Heat shock and several chemicals cause left-right inversion in up to 50% of embryos exposed during neural groove formation. Earlier stages are less sensitive; no treatment begun after foregut pocket formation influences asymmetry. Evidence for cellular interactions in left-right specification comes from the apparent rescue of iv/iv mutant embryos in chimeras. We are looking for molecular left-right disparity before morphological asymmetry but detect no differences in two-dimensional protein profiles. Using an indirect measure, we find a right-left gradient of tissue oxygen in embryos at the 20–30 somite stage. This may reflect asymmetrical vasculature, as we have suggested to explain drug-induced asymmetrical limb malformations.