1. looking for unpredicted, post hoc effects in a body of data.

2. examining data before an experiment has been completed, which can sometimes result in erroneous or misleading conclusions. See also data dredging.