In 1995, the Dutch Minister of Health proposed that a randomized clinical trial (RCT) with heroin-maintenance for severe abusers be conducted. It took nearly four years of lengthy debates before the Dutch Parliament consented to the plan. Apart from the idea of prescribing heroin, the minister and her scientific advisers had to defend the quite high material and non-material costs that would arise from employing the randomized controlled design. They argued that the RCT represented the truly scientific approach and was (...) the royal way to unambiguous results. In the present article, I question this common dual justification of RCTs. First, I situate the historical origins and the basic assumptions of the ideal experiment in 20th-century economic liberalism. Secondly, using the Dutch heroin experiment as an example, I discuss human-science experimentation as an attempt to create reality rather than merely record it. Finally, I discuss some surprising responses by heroin users. These responses display the assumptions of RCTs discussed in the historical section, and underline the importance of the culture of heroin use. In the epilogue, I suggest that cultural aspects of heroin consumption can best be studied by thorough ethnographic research. (shrink)