Experiment A was conducted in the 1996/1997 dry season. The soil was slightly acidic, infertile, sandy loam with low organic matter and total N (Table 1). Methods used for determination of soil properties are; pH meter for pH (1:1 water), Walkley and Black for organic matter, Kjeldahl for total N, Bray 2 extraction and spectrophotometry for available P, 1 N ammonium acetate extraction–atomic absorption spectrophotometer for extractable K, and ammonium acetate saturation distillation for cation exchange capacity (CEC). Experiment B was conducted in the 1997/1998 dry season. Because only mild drought stress was induced in experiment A, the differences in drought resistance or susceptibility among populations and RILs might not have been fully expressed. Therefore, experiment B was conducted at a higher position in the toposequence so as to increase the chance of inducing severe drought stress. The soil in experiment B was similar to that in experiment A (Table 1).