ABSTRACT
In a way, the whole of our penal system is one big experiment because we often call an experiment any kind of social action the ultimate effects of which are uncertain. In this sense, the legal provisions concerning the establishment of detention centres, the setting up of other new types of correctional institutions, the replacement of a prison sentence by a probation order or the earlier release of an offender on licence may be called experiments just as, in Popper's words, a grocer who opens a new shop is conducting a social experiment. 2 This, however, is not the meaning attached to the term in the present chapter, where it is throughout used in the narrower sense of a strictly controlled experiment which it has in the natural sciences and in experimental psychology. The criteria of such an experiment are as follows: It requires, first, two or more contrasting situations which are usually created by the experimenter (artificial or created experiment); in exceptional cases, however, they are already in existence without any interference on his part (‘natural’ experiment). 3 Greenwood 4 uses as an illustration the experiment which tries to discover the cause of goitre in certain communities by arranging the supply of water as a suspected cause in such a way that community A receives its supply from source X and community B from source Y (see Chapter 6, II, above). If this arrangement had been in existence by chance already before the experiment started no interference with the sources of supply would have been required. Secondly, two groups of people who drink the water have to be formed and carefully matched with regard to any variables which might possibly also be responsible for the causation of the disease such as age, sex, health, food, occupation, etc. The experimenter has now to make sure that group A uses only water from source X and group B only water from source Y, which is usually called the introduction of the stimulus. Thirdly, the incidence of goitre in the two groups has to be examined after a certain time has elapsed. In this example, which is an illustration of the experiment in its narrowest form, the water would be called the independent and the effect, the disease, the dependent variable or the criterion. 5 If this first experiment should be inconclusive, similar arrangements would have to be made with regard to other variables such as certain items of food, but ‘the fundamental rule of the experimental method is to vary only one condition at a time’ (Chapin). A statistical correlation has been discovered between the prevalence of dental caries in children and the fluoride content of their water supply. A controlled alteration in the fluoride content has produced a reduction in caries, and this experiment, while revealing nothing about the causal mechanism behind it, makes control and prevention possible. 6 As stressed by Greenwood and others, controlled experiments in this strictest sense of the word are rare in the social sciences, and many sociologists believe them even to be almost impossible. Moreover, where no serious attempt is made to control the situation and to keep the influence of variables out which are not the subject of the experiment it would be misleading to use this term at all. As will be pointed out below (IV) attempts have been made to fill the resulting gap by inventing certain ‘quasi-experimental’ techniques.
