This paper intends to clarify a conceptual confusion in the classification of educational research methods. Most of current studies on the subject evince the problem of indiscriminate mixture of various levels of methodology. Against this trend I divide research methods in three levels: normative, epistemological, and technical. In the normative level, research methods can be classified in two realms, critical and scientific, according to the value researchers attach to their studies. Critical research method strives for the debunking of oppressive nature of current social structure and the negative contribution of education to its maintenance, while scientific method takes social reality for granted and tries to find factual order as is reflected in data sentences. In the epistemological level, research method can be divided into positivistic and interpretative camps, depending upon how researchers attempt to investigate educational phenomena. Positivistic orientation incorporates the natural scientific methods in the study of social phenomena to establish the universal law, while anti-positivistic orientation refutes the homogeneity of social and natural objects and argues for the usefulness of interpretative methods for the study of peculiarity of human behavior. And in the technical level, research methods can be categorized into qualitative and quantitative methods, by the way researchers collect and analyze empirical data. Despite theoretical and practical relevance among these three levels, it is absurd to assume the existence of an inherent and necessary connection penetrating them. Rather, we recognize the disjunction between them, as well as within the technical level itself. For instance, though quantitative methods in the technical level are embedded in the positivistic tradition in the epistemological level, we can find some qualitative methods, like ethnographic and ethnoscientifc studies, which purport to attain the goal of generality as in the former. An ignorance of this multi-dimensionality of research methods constitute one important source of the confusion in their classification. The commonality in the classification of various levels of methods is the division between qualitative and quantitative research methods, as they are applied in the actual processes of educational research. Each method has its advantages and limits. Through quantitative methods we not only exactly discern the relative importance of each independent variable over a dependent variable, but also control the compounding effects of independent variables. However, since these methods detach the variable out of social context it is actually situated, we may draw distorted conclusion when one variable possesses different, often quite contradictory, effect on the dependent variable in different social situation. Furthermore, when they handle the variables with skewed distribution, they often reveal what is known as the universalitic fallacy, discounting its internl variation. On the contrary, qualitative methods generally rely upon the holistic approach and thus are good for the comprehensive analysis of some limited number of cases, which at the same time make up its disadvantage of the particularistic fallacy. From this discussion of the characteristics of each method we can find the possibility and desirability of their integration into a more encompassing methodological framework. This strategy of integrative methodology starts from the application of qualitative methods. By employing them we can derive out some strategically important independent variable accounting for the dependent variable, without abstracting them from social context. After establishing the potentially general law with them, we need to extend the research to include more extensive number of cases, which will be analyzed by the statistical methods. The repetition of this strategy will ensure both the scientificity of educational research and the establishment of generality of research results. This is the way how we can transcend the limits of both technical methods.