This study demonstrated that, in rice seedlings, genotypic difference in tolerance to anoxia only occurred when anoxia was imposed at imbibition, but not at 3 d after imbibition. When seeds were imbibed and grown in anoxia, IR22 (anoxia-'intolerant') grew much slower and had lower soluble sugar concentrations in coleoptiles and seeds than Amaroo (anoxia-'tolerant'), while Calrose was intermediate. After 3 d in anoxia, the sugar concentrations in embryos and endosperms of anoxic seedlings were nearly 4-fold lower in IR22 than in Amaroo. Sugar deficit in the embryo of IR22 is presumably due to the limitation of sugar mobilization rather than the capacity of transport as shown by similar sugar accumulation ratios of 1.8 between embryo and endosperm in IR22 and Amaroo at 3 d in anoxia. With 20 mol m-3 exogenous glucose, coleoptile extension and fresh weight increments in anoxic seedlings of IR22 were much closer to those in the two other genotypes, nevertheless protein concentration remained lowest on a fresh weight basis in the coleoptiles of IR22; indicating that protein synthesis has a lower priority for energy apportionment during anoxia than processes crucial to coleoptile extension. In contrast to these responses to anoxia imposed at imbibition, IR22 had nearly the same high tolerance to anoxia as Calrose and Amaroo, when anoxia was imposed on seedlings subsequent to 48 h aeration followed by 16 h hypoxic pretreatment. In fact, coleoptiles of anoxic IR22 had higher sugar concentrations and grew faster than Calrose, and exogenous glucose had no effect on the coleoptile extension of IR22. Excised coleoptile tips of IR22 and Amaroo with exogenous glucose had similar rates of ethanol production and were equally tolerant to anoxia. In conclusion, much of the anoxia 'intolerance' of IR22 when germinated in anoxia could be attributed to limited substrate availability to the embryo and coleoptile, presumably due to slow starch hydrolysis in the endosperm.