“Subjects-and-Controls” Experiments

Situations wherein several causes acting together result in one or more noticeable effects are not rare. Our coffee lover’s experiment was but one example. If the intention of the experiment is to study the effect of a new cause in addition to the existing ones, then a comparison between two cases, one with the additional cause and another without it, all other conditions being similar, becomes necessary. The study may pertain to objects, animate or inanimate, or to nonobjective phenomena; each item in the study is referred to as an individual. The normal procedure in such a study is to

1. Select a fairly large, reasonable number of individ­uals by a random method, the individuals repre­senting the population.
2. Divide these by a random method into two groups of equal number.
3. Apply to the first group all the usual causes.
4. Apply to the second group all the above causes, plus the additional cause, to find the effect of which forms the experiment.
5. Observe the result, the effect(s), in the first group as a statistical average.
6. Observe the effects in the second group as a simi­lar, statistical average.
7. Compare the two results obtained above.

The difference in effect(s) in the groups is attributed to the additional cause. In such experiments, individuals in the first group are known as controls, and those in the second group are known as subjects.

Example 7.3

Suppose it is the intention of an experimenter to try the effect of a new plant food on the yield of a certain variety of annual flow­ering plant. Then, the experimenter randomly selects, let us say, one hundred seedlings from a larger lot. He plants these in one hundred pots, one plant in each, identical in all respects: size, material, soil, moss, manure, fertilizer, moisture, and so forth. He marks by a random method fifty of these as subjects. The other fifty plants become controls. He cares for all one hundred pots with identical gardening procedures, such as space, sunshine, warmth, watering, pesticide, and so on. He does not distinguish one in any way from the others in his routine care all through their growth until the flowering time, with the only difference that he adds the dose or doses of experimental plant food uni­formly to all subjects in predetermined quantities and on a prede­termined schedule. In time, the plants begin flowering. He records the yield from each subject and arrives at a statistical aver­age for all subjects. Whether the “yield”—the criterion—is greater number of flowers, bigger flowers, brighter flowers, or flowering early in the season or any other single or combination of criteria is either predetermined or left open by the experi­menter to determine later. Similarly, he obtains a statistical aver­age for the controls. The effect of the experimental plant food may be beneficial, deleterious, or neutral. This is decided by comparing the two statistical averages. If the subjects yielded more flowers, for example, the effect of the new plant food is beneficial to that extent.

1. Varieties within Subjects and Controls: Paired Comparison Design

Example 7.4

Let us imagine that an investigator in the gardening experiment was very pleased with the efficacy of the plant food as he has a stake in its promotion. He wants to extend his investigation to know if many other kinds of flowering plants benefit as well from the plant food. Let us call the particular variety of annuals he has recorded his success with as A and the other varieties as B, C, D, E, F, G, H, and J. Now he plans his experiment differently. He selects two plants of A at random from a larger lot. After considering the features and properties that he knows are signifi­cant in deciding plant fitness or healthiness, leading to yield, and after analyzing such features and properties, he is convinced that the two plants of A are of equal merit. Following the same procedure, he selects two plants each of B, C, D, E, F, G, H, and J.Between the two plants of A, he decides, by tossing a coin, which one of the two will be the subject and which the control. Let us now call the subject A¹ and the control A. Along the same lines, he decides between B¹ and B, C¹ and C, and so forth, and, accordingly, tags the pots containing the plants. This procedure of selecting two individuals that are as close as one can find in regard to the discernible features and properties that make the two individuals of equal merit or potentiality is known as pair­ing. As described above, the purpose is to use one as the subject and the other as the control. Our experimenter has now on hand eighteen pots of plants: A, B, C, D, E, F, G, H, and J to be used as controls and A¹, B¹, C¹, D¹, E¹, F¹, G¹, H¹, and J¹ to be used as subjects.

At this stage, we will assume that the routine treatments— pot size, the kind and quantity of soil and moss, the amount and frequency of watering, weather control, including need for pesti­cides, and so forth—required for A and A¹ are different from those required for B and B¹ and C and C¹, and so on. All these treatments—and we assume our experimenter knows these pro­cedures well—are met as required. All the routine treatments given to A, B, C, . . . J, are also given correspondingly to A¹, B¹, C¹, . . . J^1. The one and only difference is that plant food, not given to A ,B, C, . . . J, is given to A¹, B¹, C¹, . . . J¹ as an “extra treatment.” The “dosage” appropriate for A, B, C, . . . J is decided on a somewhat arbitrary basis but subject to the exper­tise of the experimenter. Such routine cares as are considered appropriate are given through the flowering periods. Using the criterion of “improvement,” the yield of A¹ is compared with that of A, the yield of B¹ with that of B, the yield of C¹ with that of C, and so forth. At this time, we assume that each subject’s yield was better than that of the corresponding control, and to that extent our experimenter has reason to be convinced that the experimental plant food benefits “all” annual flowering plants. If the question arises in his mind about the degree of benefit, then he needs to quantify his criteria and follow the appropriate numbers in terms of statistical analysis, which we will deal with in Chapter 19.

2. Experiments with Humans

Investigations like that above, known as paired comparison design, are fairly common in medical research, when experimenting with drugs and procedures. The problem of designing the experiment in such studies becomes more complex for the following reasons:

1. When dealing with humans, it is difficult to find identical patients. Age, race, sex, heredity, physical features, educational level, social background, philosophical attitudes, habits, and temperament are some of the factors that distinguish individu­als, and these factors have significant influence on reactions and responses to treatments. It is no exaggeration to say that there are no two exactly identical persons. Identical twins may be closest to the ideal, but to find sufficient numbers of such twins, simultaneously available as “patients,” is extremely rare. Otherwise, forming several pairs, one half of which is to serve as control and the other half as subject, is fairly common; the choice of who the subject will be is determined by the toss of a coin. With several such pairs, a set of con­trols and another set of subjects, can be obtained; all this means that the population from which the individuals can be selected for the experiment is very large and the search correspondingly difficult.
2. Humans, unlike plants, have rights. It is often nec­essary to obtain informed consent from the indi­viduals in an experiment. If some individuals do not consent, the required population needs to become even larger.
3. Even after the consent is obtained, the decision, though made purely by chance, that one of the pair will receive the treatment and the other will not has ethical implications. If the treatment is known to be beneficial, the subject individuals stand to gain. If the treatment turns out to be harmful, the subjects are victimized, with the questionable justification that there is a greater good beyond the individuals involved. If Imman­uel Kant could breathe, he would shout from his grave that this justification is improper. Is not each individual an end to himself, not just a means? Industrial products, at the other extreme, fortu­nately for the experimenter involved, do not pose such disturbing questions.

Source: Srinagesh K (2005), The Principles of Experimental Research, Butterworth-Heinemann; 1st edition.