We investigated the differences in performance between 100-m sprints performed on a sprint treadmill recently validated versus on a standard track. To date, studies comparing overground and treadmill running have mainly focused on constant and not maximal "free" running speed, and compared running kinetics and kinematics over a limited number of steps, but not overall sprint performance. Eleven male physical education students including two sprinters performed one 100-m on the treadmill and one on a standard athletics track in a randomized order, separated by 30 min. Performance data were derived in both cases from speed-time relationships measured with a radar and with the instrumented sprint treadmill, which allowed subjects to run and produce speed "freely", i.e. with no predetermined belt speed imposed. Field and treadmill typical speed-distance curves and data of maximal and mean speed, 100-m time and acceleration/deceleration time constants were compared using t tests and field-treadmill correlations were tested. All the performance parameters but time to reach top speed and deceleration time constant differed significantly, by about 20% on average, between field and treadmill (e.g. top speed of 8.84 ± 0.51 vs. 6.90 ± 0.39 m s(-1)). However, significant correlations were found (r > 0.63; P < 0.05) for all the performance parameters except time to reach top speed. Treadmill and field 100-m sprint performances are different, despite the fact that subjects could freely accelerate the belt. However, the significant correlations found make it possible to investigate and interpret inter-individual differences in field performance from treadmill measurements.