Research in the area of multivariate outliers is reviewed, emphasizing the problems associated with definition and identification. Treatment of the problem can be traced to 1777 and the work of D. Bernoulli. Most of the many procedures developed for identifying outliers proceed sequentially starting with the most aberrant observation, or proceed without consideration of the influence that an outlier may have on the focus of the analysis. If an outlier does not influence the outcome, there may be no reason to be concerned with it. A major problem in identifying outliers has been the lack of agreement on an operational definition of "outlier," even though most definitions refer to the outlier as being extreme in some manner. Researchers must reach a consensus on the definition in order to proceed in perfecting identification procedures. Once an outlier is identified, its treatment is contingent upon the type of data being studied. Outliers may be of such importance as to be the main focus of some types of study. A list of 64 references is included. (SLD)