Sheffe's test is used to test whether an unplanned contrast is zero. In other words, it can be used on an arbitrary contrast, after the experimental data is collected, to test a research hypothesis which may not have been anticipated in advance.

See the attachments for an example.