In controlled biological experiments,controls are generally needed to allow for possible effects of uncontrollable non study factors on the outcome measure. In some experiments, where prior knowledge implies that only the study factor could be having an effect, controls may no longer be needed.(eg in dose effect studies).Positive controls may sometimes be needed to test if the experimental set up is sufficiently sensitive.

In epidemiological cohort studies, a reference ( or non exposed ) control group is needed, except possibly in dose effect incidence studies where prior knowledge indicates that basically only one unique factor causes the outcome  eg asbestos and malignant mesothelioma (excluding the possibility of erionite). In a case referent study  non cases (referents) with measurable exposure are needed.

In a few particular types of case referent study namely case only and case crossover, referents can be dispensed with .In case only the exposure distribution of the source population is estimated from prior theoretical considerations and in case crossover the exposure of the cases at different time periods is compared.

see eg Rothman and Greenland ,Modern Epidemiology 2nd ed 1998 p 111-112.