When studying the properties of various physical objects, experimental data often contain gross errors (outliers) which can lead to significant distortions of the results of statistical processing of such data. For this reason, statistical procedures are being developed that are protected from the presence of outliers in observations. In this paper, two types of robust estimates of the scale parameter, which characterizes the spread (variability) of the random variable under study, are considered. The proposed estimates are asymptotically normally distributed, have bounded influence functions, and therefore, unlike the standard deviation estimates, are protected from the presence of outliers in the sample. The results of comparing the estimates of the scale parameter by the efficiency for different observation models, in particular, for the Gauss model with large-scale contamination, are presented.