OrientationObservationIn-depth interviewsDocument analysis and semiologyConversation and discourse analysisSecondary Data
SurveysExperimentsEthicsResearch outcomes
Conclusion
9.4 Summary and conclusion Experimentation is a means of testing a hypothesis. It is not a data collecting method like social surveying, or observation. The researcher is not involved in collecting data from the 'real world'. Instead, a situation is set up, and the researcher notes what happens when an experimental variable is manipulated.
This contrived situation might be providing subjects with information to see if it changes attitudes. Or seeing if people behave differently in groups from the way they do when acting individually. Or seeing how people cope in a crisis, or when things they take for granted no longer apply.
The researcher records the outcome of the experiment, much as a physical scientist might make a note of readings from a set of gauges.
Positivist 'classic' experiments are designed to test causal assumptions. The experiment is thus the paradigm method for positivists.
Although little used by sociologists, the experiment is the model that the widely used multivariate-analysis approach attempts to recreate. To be credible, positivist experiments must ensure that other variables are controlled.
Ethnomethodological 'breaching' experiments, on the other hand, are designed to see what people take for granted. They are thus concerned with meanings, not causes. For them, the issue of control is not important.
In all forms of experimentation, though, the issue of external validity is important. The sample must be representative if the results are to be generalised (population validity). Similarly, it is necessary to ask whether the contrived setting of the experiment can be generalised to 'real life' situations (ecological validity).
All forms of experimentation raise ethical problems (see Section 10.2 and Section 10.5.2). It is not usually possible to tell the subjects what the point of the experiment is until it is over because prior knowledge may lead people to change their behaviour and render the experiment useless. You must then consider carefully how much you can ethically manipulate people for the sake of an experiment (see Section 10).