I believe the answer is: one cannot separate their effects on a response variable
The existence of confounding variables would ruin the whole experiment and make it produce useless results. To prevent confounding variables from happening, researches tend to separate the subjects into two groups. One group would be given the real treatment while the others would be given placebo.
