Reply #1: The controlled trial is one model of experiment

[Recursion]

But it's not the only one. Even the way you described wouldn't be the holy grail of experimental science, the double-blind control study (in which neither the subject nor the researcher know whether the subject is in the experimental group of the control group). Of course, with non-human subjects, the subject is always assumed not to "know", so the question is if the researcher would know.

These kinds of experiments are great for certain kinds of questions. But they suck at other kinds of questions -- particularly kinds where we don't really know how many factors we need to sort through (famous case in point: doctors were trying for years to find a substance to help people having heart attacks but every drug was showing the same benefit -- eventually they realized it wasn't actually any of the drugs, it was the glycerine they was being injected along with them).

A better model of experiment for your trees would be Multiple Competing Hypotheses. It gets more use in the life sciences and social sciences.