A rapid (7 day) solution-based screening test was developed using 15 annual Medicago cultivars and one M. sativa. Based on a relative root regrowth after exposures to aluminium (Al), Zodiac (M. murex), Orion (M. sphaerocarpos) and the M. polymorha cultivars Santiago, Cavalier and Serena had the greatest Al tolerance. Herald (M. littoralis) and Rivoli (M. tornata) were most sensitive. Ranking for Al tolerance from the solution culture correlated well (r = 0.80) with ranking for tolerance of the 16 genotypes grown in an acidic soil (unlimed pHCa 4.1). We screened 17 Australian populations of lucerne (M. sativa) using a 24 h 'pulse' of 75 Âµmol/L Al, and a three day 'recovery' of 10 Âµmol/L Al. We identified and recovered plants with a root regrowth of '5 mm in all 17 populations with selection intensities of 2 to 4%. Four of these selected populations (Aurora, UQL-1, A513 and TO2-011) were polycrossed within each population to produce four populations of seed from the cycle 1 selections. The length of root regrowth under Al stress was improved for all four populations of cycle 1 selection (P ' 0.001; from 2.6 mm for the original populations to 6.3 mm for the cycle 1 selections). In a subsequent experiment the cycle 2 selections from Aurora, UQL-1 and TO2-011 had significantly greater root regrowth than both the cycle 1 selections (P ' 0.001; 8.3 cf. 6.6 mm) and the unselected populations (3.0 mm). The selections from TO2-011 appeared to have greater improvement in the average length of root regrowth after 2 cycles of selection. Selected germplasm was more tolerant than GAAT in our evaluation. Based on estimation of realised heritability, it seemed likely that higher selection intensities would give more rapid improvements in tolerance. Our studies have not investigated the physiological basis of any tolerance of Al which we observed.