It is shown that the usual method of microturbulent velocity determination from the "abundance versus equivalent width" plot leads to a systematic overestimate of the microturbulent velocity when the observd equivalent widths are affected by random errors. This overestimate is due to the correlation between errors in observed equivalent widths and in line abundances. This error rises rapidly with decreasing quality of the observational material and may be avoided by using theoretical equivalent widths instead of observed ones as abscissa of the "abundance versus equivalent width" plot.