The measurement uncertainty in the absolute calibration of accelerometers using laser interferometry was reduced by a time capture method. Outputs of a photodetector and accelerometer were digitally sampled at a rate of 256k samples per second. Displacement amplitudes and accelerometer sensitivities were found from the sampled data. During data processing, a technique was used that decomposes a multivariable nonlinear optimization into a series of single-variable optimizations. This decomposition technique not only reduces the computational complexity but also avoids the convergence to local minima. Results of calibrations showed that the measurement uncertainty at the high-frequency end (800 Hz) for a small acceleration of 10 ms[sup -2] was less than 0.9% at 95% confidence level. This time capture method can be applied to any time-variant (nonsinusoidal, nonperiodic) displacement measurement. With the above method, measurement requirements such as frequency stability and acceleration amplitude stability, can be less stringent. The effects of airborne noise and vibration can also be reduced.