(1) In the classical theory of sociological experiment it was customarily believed thet the experimenter must hold all (independent) variables constant except one-either directly (by means of the Method of Removal, the Constant Method the Screening Method, the Counter-balanced Method, or Systematic Randomization) or indirectly (by means of the Matched Group Method). The 'rule of the single variable' was formulated before moreef ficient statistical methods and experimental designs had been developed. Theclassical experimental logic is only applicable when the complete identification of all factors is possible, when there is complete arithmetical correspondence between cause and effect, and when no interaction between the factors exists ; but in the complex social reality these conditions are never in fact satisfied. Therefore, modern experiments, with the statistical tools now available, handle several independent variables within the same design and also include as many dependent variables as seem necessary. One experimental design of this sort is the 'factorial design'. In this connection, the 'pure case' method advocated by K. Lewin and F. S. Chapin is logically defective.
(2) Sociological experiments can be classified according to various criteria. For example E. Greenwood used three criteria ; namely, artificiality of the situation, simultaneity of the comparison and direction of relationship, and produced the following classification :
(a) The Projective successional experiment.
(b) The Projective simultaneous experiment.
(c) The ex-post-facto cause to effect experiment.
(d) The ex-post-facto effect to cause experiment.
However, the most significant dichotomy of experiments would seem to be the classification into 'field experiment' and 'laboratory experiment'. This because of the importance of the 'strategy of social research' and the emphasis which must be placed on the supplemental relationship between field survey, field experiment, laboratory experiment and the clinical analysis of cases. Experimental methods must be viewed in the context of various other social research techniques. Otherwise they are doomed to futility.
(3) However, traditional sociological experiments have at the same time many other weak points. In the first place, they have mostly handled uninstitutionalized small groups and have lost couch with macroscopic studies of societies. In experiments with small groups it is necessary to examine the cultural setting of the experimental situation which places a limitation on the general validity of the experimental conclusion, and at the same time to elaborate methods of transposing the essential structure of social reality into the experimental situation. Secondly, since experimentation with social phenomena is itself a historical event, we must examine the nature of experiments from the viewpoint of the sociology of knowledge. Thirdly, the spontaneity of the subjects is a necessary conditions for succesful sociological experimentation. Only experiments which call for the spontaneous response of the subject, experiments which are conducted with a view to the subject's welfare, experiments which gain the cooperation of the subject, can hope to succeed. Fourthly, in order to carry out large-scale social experiments, the experimenter must have wide control over social phenomena in order to make possible the setting up of experimental conditions. Except in the classless planned society these conditions are unlikely to be satisfied.
