In its broadest sense, this paper reviews the general outlier
problem, the means available for addressing the discordancy (or lack thereof)
of an outlier (or outliers), and possible strategies for dealing with them. Two
alternate approaches to the multiple outlier problem, consecutive and block
testing, and their respective inherent weaknesses, masking and swamping, are
discussed. In addition, the relative susceptibility of several tests for
outliers in normal samples to the swamping phenomena is reported.