The sample size must be at least 1,000.

The variables must be measured on an interval scale.

The expected frequencies in each cell should generally be at least 5.

The population must be normally distributed.