From: "carr0023@xxxxxxxxxxxxxxx" <carr0023@xxxxxxxxxxxxxxx>
Date: Sun, 25 Dec 1994 16:59:40 CST
Excerpt: Advice to a Young Scientist P.B. Medawar, c. 1979 h&r pub.
Chapter 9: ~Experiment~and~Discovery
"No experiment should be undertaken without clear pre-
conception of the forms its results ~might~ take; for unless a
hypothesis restricts the total number of possible happenings or
conjunctions of events in the universe, the experiment will yield
no information whatsoever. If a hypothesis is totally permissive
-- if it is such that ~anything~ goes -- then we are none the wiser.
A ~totally~ permissive hypothesis says nothing.
The "~result~" of an experiment is never the ~totality~
~of~observables~; the result of an experiment is almost always the
~difference~ between at least two sets of observables. In a simple,
one-factor experiment, the two sets of observables are called the
"experiment" and the "control." In the former, the factor under
investigation is allowed to be present or to exercise its effects,
and in the latter it is not.
The "result" of the experiment is then
the difference between the readings or counts in the experiment and
control. An experiment executed without a control is not Galilean
in style but still might qualify as an experiment in the Baconian
style-- that is, as a little contrived performance of nature, though
not a very informative one. In the performance of what is intended
to be a critical experiment, clarity of design and fastidiousness of
execution are the qualities to be aimed at."