There is rarely the opportunity to perform the classic and most rigorous form of experiment, or The True Experiment, because the first critical requirement is to randomize all selection of test subjects. Second, there must be randomization in all decisions as to which subjects are to be tested and which are not to be tested, or in selecting test and control groups. Third, there is complete control over all variables, and all variables are known. Fourth, detailed definitions and explanations of qualitative and quantitative matters, records, documentation, and identification of the exact formulas and procedures used in the test are sufficient for an objective and uninvolved party to replicate the selection and testing, and to obtain the same results.
That is a tall order.
Non experimental testing is a more realistic goal in the real world of humans and their social systems. Social scientists often seek the “sufficient”, the “substantial” or the “exceptional” rather than the “absolute” in determining whether an individual hypothesis or a system of hypotheses has been considered to be acceptable. Such terms as “supported” replace absolute terms such as “proved”.
In non experimental research, complete randomization is usually not possible, so meticulous documentation of how the test subjects were selected must be made, so that as much replication and correction of errors as possible can be done in future experimentation.
Controls over all variables is not possible in most sociological experimentation. All variables and their effects are not necessarily known, nor can they be known. Meticulous methods for identifying other variables and making sound alternative hypotheses, where inferences about other variables can be simultaneously tested, must be established and documented.
All variables may not be subject to complete isolation and control during the experiment, where one variable is changed or manipulated to see what happens to the hypothetical dependent variable.
As a result, Quasi Experimentation, Survey, interview, historical, statistical, and even the Anthropological methods of research are used by the social scientist. Quasi Experimentation is the closest method to true experimentation, but most commonly lacks the ability to randomize test subjects. Other aspects of the experiment may be well regulated in the documentary or qualitative, control, quantitative and other aspects.
The survey method uses either well controlled examination and summaries of documented or qualitative facts to support a hypothesis, or uses the responses of people to questions. The survey method is highly vulnerable to misinterpretations of the questions and answers, bias in the questions, and false responses by individuals. The survey method is most popular in the political, legal, urban planning, marketing, and other areas where social engineering protocols and programs are developed, and where exacting or immutable results are not possible or necessary.
The historical method can involve a grueling and disturbing form of the Anthropological method called Psycho Historical experimentation to confirm or to develop hypothesis which derive inferences about human behavior and experience during horrific events in history, or to understand such serious deviance as serial killing, generational drug abuse, and other deviant patterns in social and individual behavior. The historical method may involve comparing qualitative data from the past to quantitative date from the present in order to conform hypotheses.
The quantitative methods are legion in the social sciences. Models, algorithms, and statistical processes are complex analytical tools that are used to test hypothesis when the behavior of large populations can be quantified and examined. Other analytical tools involve qualitative information and data, such as the Marxian model and thought, the principles of Emile Durkheim, the Christian religion, the Capitalist economic system, or the Democratic political system.
In summary, the non experimental method is used in most sciences, and not just the social sciences when there is no chance that true randomization or control of variables is possible. There is many a null hypothesis, or hypothesis that leads nowhere in cancer research, where intervening or unknown variables cause a promising new treatment to fail.
In some cases, such as marketing, an experimental method may not needed at all to predict the responses of humans to the release of a new or improved product with enough accuracy to make a profit, then to end production when forecasts indicate that the product will be obsolete.
Chong-Ho Yu, PhD, CNE, MCSE, CCNA