Well-conducted experiments require a “control”—a measurement where you expect no effect at all, and which serves as a kind of idiot-check on what you are measuring. For example, if you wonder what effect beer has on a tulip plant, then also nurture a second tulip plant, identical to the first, but give it water instead. If both plants die—if you killed them both—then you can’t blame the alcohol. That’s the value of a control sample.
