Abstract
The values of the power of Tiku's (1975) T statistic for testing outliers in normal samples are evaluated. The statistic T is shown to be more powerful than other comparable statistics under Tiku's outlier model, although slightly less powerful under Dixon's (1950) contamination model.