Del Siegel of the University of Connecticut notes the selection of the subjects that will participate in the study is the first design decision. Based on the research question, you need to determine the best subjects for the study. Many social science research studies are done using college students because they are available, but they may not be the best sample to generalize from.
Once the type of subjects is determined, it is necessary to develop a method for sampling them. Random sampling is best, but is not always possible. Ultimately the goal is to conduct the research on a group of subjects that is representative of the larger population you would like to generalize your findings from.
Based on the research question, the subjects to be studied and the context of the study, the research needs to determine what kind of data to collect. In a study of museum visitors, the subjects can be observed as they go through an exhibit; they can be given a brief survey questionnaire about their experience in the exhibit or the researcher can observe and collect survey data. For some studies, photographs or videos of subjects might provide the most appropriate data. In all cases, the instrument must be chosen that brings in data that best addresses the research question.
Specifying procedures to be followed in data collection is an important part of the design. To minimize bias, a researcher collects all data in a similar fashion. If the study is using a survey questionnaire administered face-to-face by a research assistant, he can be approach the potential subjects and give each of them identical directions about how to complete the questionnaire. With good procedures, research assistants are also instructed about what to say if a subject has questions during the process of completing the survey. Usually the instructions remind him to say as little as possible and encourage the subject to interpret the survey questions as best as she can.