An experimental control, often referred to as a "control group", is a population or set of objects that is statistically similar to the set being tested, on which no changes are implemented.
This is in order to make sure that any observed results are statistically related to the tests being performed, and not simply random occurrences that happen anyway.
When the possibility of experimenter or subject bias is an issue, the control group is managed through double-blind testing.
In psychology and biology, the control group is very important, since results are often statistical rather than concrete. For instance, in testing a drug for a malady, some percentage of the test subjects will heal with no intervention (or heal at some rate in a third group, people getting an existing remedy). The control group yields this number, and the group getting the treatment under test can be compared to this to determine efficacy.
While it is less of an issue in the physical sciences, part of the description of how to set up an appropriate experiment should always address what controls are used to limit the independent variables to the one of concern.