This study compared the energy cost (EC) (J x kg(-1) x m(-1)) of running on grass and soft dry beach sand. Seven male and 5 female recreational runners performed steady state running trials on grass in shoes at 8, 11 and 14 km x h(-1). Steady state sand runs, both barefoot and in shoes, were also attempted at 8 km x h(-1) and approximately 11 km x h(-1). One additional female attempted the grass and sand runs at 8 km x h(-1) only. Net total EC was determined from net aerobic EC (steady state VO2, VCO2 and RER) and net anaerobic EC (net lactate accumulation). When comparing the surface effects (grass, sand bare foot and sand in shoes) of running at 8 km x h(-1) (133.3 m x min(-1)) in 9 subjects who most accurately maintained that speed (133.3 +/- 2.2 m x min(-1)), no differences (P>0.05) existed between the net aerobic, anaerobic and total EC of sand running barefoot or in shoes, but these measures were all significantly greater (P<0.05) than the corresponding values when running on grass. Similarly, when all running speed trials (n = 87) performed by all subjects (n = 13) for each surface condition were combined for analysis, the sand bare foot and sand in shoes values for net aerobic EC, net anaerobic EC and net total EC were significantly greater (P<0.001) than the grass running measures, but not significantly different (P>0.05) from each other. Expressed as ratios of sand to grass running EC coefficients, the sand running barefoot and sand in shoes running trials at 8 km x h(-1) revealed values of 1.6 and 1.5 for net aerobic EC, 3.7 and 2.7 for net anaerobic EC and 1.6 and 1.5 for net total EC respectively. For all running speeds combined, these coefficients were 1.5 and 1.4 for net aerobic EC, 2.5 and 2.3 for net anaerobic EC and 1.5 and 1.5 for net total EC for sand running barefoot and in shoes respectively. Sand running may provide a low impact, but high EC training stimulus.