In constructing a single variable, it is recommended that the data first be carefully examined. A sample set of data is examined from a variety of perspectives to determine if a variable can be derived from them. Some techniques for inspecting data are presented. Then some models for measuring are described. These are the tools used to build variables and make measures. Five different measurement models, each of which has been developed for a particular type of data, are portrayed. The models are members of a family of models which share the same basic structure. How to use the models to get results is indicated as four different estimation procedures are described. The last phase in the construction of a variable is its quality control. The 'separation' index described can be used to assess the extent to which items and persons have been separated along a common line, which will determine if a variable has been successfully defined. Some procedures are also described for identifying and summarizing misfit. The study analyzes the response patterns of some persons with unusual or inconsistent responses and discusses some frequently encountered problems. The concluding chapters outline and illustrate techniques for comparing and monitoring the performances of items over time and from group to group. The use of the techniques is illustrated with data collected to measure attitudes towards drug use, fear of crime, knowledge of elementary physics, and the development of prekindergarten children. Fifty-nine references and a subject index are provided.