When an experiment is carried out, it's important that only one variable is tested so that any changes observed can't be attributed to another factor. A control set-up "controls" for other factors. Ideally it's the exact same set-up as the experiment except for one difference - the variable being tested.
Let's say that you've designed an experiment to test the effect of a nutrient on the rate of growth of juvenile rats. You will be feeding several rats their regular feed with the nutrient added, and recording each rat's length and mass at regular intervals.
The rats are most likely going to grow with or without the nutrient. For your data to be meaningful you need to show that the nutrient made a difference in the rate of growth. To do this you need to have a second set-up with everything the same except for the nutrient. You would have a similar-sized cage with the same number of rats, of the same age, that are genetically related (preferably litter mates), and of the same gender. The rats in both set-ups should have access to the same amounts and types of feed with the nutrient added only to the food of the experimental group, the same amount of water and the same opportunity for exercise (e.g.exercise wheel). Temperature and amount of light should also be the same. Each of these factors would then be controlled, meaning they would not be able to contribute to variation between the two groups.