The errors involved in the use of laser-two-focus velocimetry for measurements in subsonic turbulent flows have been discussed. In particular, measurements show that the finite dimensions of the laser beams cannot be neglected in the signal processing of the data obtained in turbulent shear flows. A simple algorithm which takes into account the finite dimensions of the laser beams has been presented. Reasonable results were obtained. A stochastic model for analysing statistical counting noise has been presented and the results indicate the signal-to-statistical noise ratio decreases with both increasing turbulence intensity and sampling rate. Results obtained by this stochastic model provide a guidance to the choice of an optimum sampling frequency.