A document from the Institute of Education Studies similarly notes that "well-designed and implemented randomized control trials are considered the `gold standard` for evaluating an intervention`s effectiveness, in fields such as medicine, welfare and employment policy and psychology" (U.S. Department of Education, 2003, p. 1), and therefore, also applicable to educational research. In another governmentally supported research paper, a similar analogy to medicine is made, with "treatment" the operative intervention relevant whether of the doctor to the patient or the teacher to the student. On this premise, socio-cultural interpretations of education are eliminated, as is the realm of values, which are viewed as outside the purview of legitimate science. The focus becomes, rather, on the technology of "what works" (Stanovitch & Stanovitch, 2003) in which the litmus test of verifiability stems from such scientific principles as "control, manipulation, and randomization" (p. 11) based on the ideal standard of statistically-valid "meta-analysis" (p. 18).

In certain lab-like environments where independent variables can be tightly controlled, experimental design can be a valuable, and, depending on the nature of the problem under study, an essential instrumentality. The methodology, even according to the more qualifying precepts of quasi-experimental design, is more problematic where variables interact in less than precisely discernable ways. This is particularly the case over regions of research like motivation and the murkier, yet, for the social sciences, the critical arena of consciousness, if one is seeking to understand, rather than simply to report on, the behavior of agents. In these regions evidence pointing to causal attribution may be susceptible to multiple explanations, which may require the "thick" description of case study analysis. In postpositivist design (see below) such areas of subjectivity do not simply collapse into the relativism of constructivism. They remain subject to the rigors of "experimental inquiry" (Peirce, 1955, p. 47) and to the ideal of "versimilitude" (approximation to the truth) (Popper, 1963) even if in a manner that would be difficult, if not impossible or irrelevant to be precisely broken down into statistically discernable categorizations or certain-like truth statements, however provisionally they may be held.

The underlying problem in the positivist quest for certainty is that of reductionism based on foundational sources of scientific analysis that stem from inductionist principles of verification through objective observation of given empirical data, or a priori rational principles of logic. Critics have noted that perception is theory-laden from inception and that both the selection and even the definition of what counts as data is a construct that cannot be accepted simply as given. Additional concerns include the underdetermination of theory by evidence that undercuts high levels of generality allegedly discerned through positivistic methodologies, as well as challenges to claims that analysis can be simply broken down into component parts given the ubiquity of situational contexts in which data is embedded. Critics also point to the centrality of the social dimension of social science research in which "variables" complexly interact in a manner that may be susceptible to multiple causations and interpretations (Phillips and Burbules, 2000, pp. 14-25).