The first question you have to ask is what your experimental population is going to look like. Try to formulate as clearly as possible the subject characteristics that you are looking for. As you saw in Part II, a daunting number of external factors can impact on the process of language attrition, and these may often interact in unexpected or unpredictable ways. Such characteristics of the participants in an experiment, which may influence the outcome variables but which the researcher cannot or does not attempt to manipulate (for example, we cannot make a subject older or younger than she is, nor can we change how often she has spoken her first language or how long ago she emigrated), are called independent variables or predictors.

Be thorough

The fact that attrition is conditioned by so many predictors confronts the researcher with the dilemma of which factors to include in the research design, and which to exclude through controlling them. It is impossible to investigate everything. How many factors you can investigate depends largely on the size of your sample.