You don't really talk about confounding variable, singular. Confounding variables are two or more variables whose effects on a subject are impossible to separate, and thus "confound" one another.
For example, if I begin a diet, and at the same time start exercising, and a month later I have lost 8 pounds, I can't tell how much to attribute to the diet and how much to the exercise.
In most experiments, it is paramount that all variables be as controlled for as possible, particularly variables which may have an impact on the results you are looking for. To continue the example, if you are conducting a study on the effectiveness of exercise on weight loss, you would want to control your subjects' diet by making them all eat the same amounts and types of foods as much as possible.
Often times, confounding variables are also lurking variables; they are not readily identified before or during the study, and are therefore not controlled in any manner, often leading to misinterpreted or incorrect results.