The data studied in social sciences in general and in sociology in particular, have a hierarchical structure. During the past two decades researches have made serious efforts to develop appropriate statistical methods for the analysis of such hierarchically structured data. As a result an important set of models by the name of hierarchical linear models, along with computer programs, have emerged. This study, considering works already accomplished attempts to demonstrate the significance of multilevel models in social research. Moreover, the characteristics of this method will be pointed out. In addition, and more importantly, the logic of linear models will be discussed. Some of the researches in which multilevel models have been used will be referred to. We will also analyze hypotheses which are being tested, and look into structure of hierarchical data and their multilevel data source. Finally, we will discuss some of the problems with which these models are involved.