Eliminating nuisance parameters from the likelihood function so that inference can be made about the parameters of interest is an important problem in statistics. Conditioning on an ancillary statistic is ol1e of the methods of parametric reduction. Andersen (1970)and Basawa (1981) suggested different conceptsof ancillarity. Liang (1983) showed that they are equivalent for exponential families of distributions. This paper investigates relationship between them in general situation. The relation to F -information is also discussed.