This study aimed to determine the measurement error associated with estimates of velocity from a laser-based device during different phases of a maximal athletic sprint. Laser-based displacement data were obtained from 10 sprinters completing a total of 89 sprints and were fitted with a fifth-order polynomial function which was differentiated to obtain instantaneous velocity data. These velocity estimates were compared against criterion high-speed video velocities at either 1, 5, 10, 30 or 50 m using a Bland-Altman analysis to assess bias and random error. Bias was highest at 1 m (+0.41 m/s) and tended to decrease as the measurement distance increased, with values less than +0.10 m/s at 30 and 50 m. Random error was more consistent between distances, and reached a minimum value ( ± 0.11 m/s) at 10 m. Laser devices offer a potentially useful time-efficient tool for assessing between-subject or between-session performance from the midacceleration and maximum velocity phases (i.e., at 10 m and beyond), although only differences exceeding 0.22-0.30 m/s should be considered genuine. However, laser data should not be used during the first 5 m of a sprint, and are likely of limited use for assessing within-subject variation in performance during a single session. [ABSTRACT FROM AUTHOR]