Statistical methods insensitive to the effects of outliers (which may be mistakes or contaminated data). The methods rely on medians rather than means, and use more information from the central than from the outlying observations. The ideas are associated with exploratory data analysis.