In evidence-based medicine (EBM) hierarchy, randomized controlled trials (RCTs) are ranked higher than cohort studies. However, cohort intervention studies are frequently, and RCTs rarely, used to investigate long-term psychotherapy (LTP). The authors compare the two methods and provide critical discussion of their acceptability, feasibility, and decisive power in LTP. The only essential and unchangeable difference between RCTs and cohort studies is that the former always include randomized control groups and the latter never do, giving RCTs a head start on internal validity that cohort studies cannot match. However, randomization nearly always has dramatic consequences for LTP research: The control conditions that are most informative (no treatment, wait list, placebo) are so unacceptable for the patients that decisive RCTs are, in most cases, unfeasible, but more feasible RCTs are less decisive. In contrast, the decisive power of cohort studies is determined by their methodological quality and knowledge of the natural course of the investigated disorders. Cohort studies are as capable as RCTs of meeting all quality criteria for intervention research, except for randomization. The knowledge of the natural course of the disorders suitable for LTP treatment is limited but not nonexistent. In most cases of LTP research, decisive RCTs present insurmountable, method-inherent feasibility problems and represent not the highest but rather an irrelevant level of evidence. The authors conclude that cohort studies provide the best available evidence.