A management treatment may be successful or it may be a failure. If the first, the
manager needs to know whether the success flowed from the treatment itself or whether
it would have happened anyway. Otherwise an expensive and unnecessary manage-
ment scheme might run indefinitely. Alternatively, the management may not achieve
its stated aim, in which case the manager must first establish that fact without doubt
and then find out why. Was the failure caused by some extraneous factor that formed
no part of the treatment? Was the entire management treatment inappropriate or only
a part of it? Would a higher intensity of the treatment have been successful whereas
the same treatment at a lower intensity was not?
To find out, the management must be run as an experiment. There are rules to
designing an experiment that are there for one very important reason: if they are
broken the questions the experiment is designed to answer cannot be answered
unambiguously.
Suppose a manager wished to increase the density of quail in an area by supple-
menting their supply of food with wheat. How that is done determines whether any-
thing will be learned from the exercise. There is a graded series of approaches, ranging
from useless in that they yield no verification of the worth of the treatment, through
suggestive in that their results allow a cautious choice between alternative interpre-
tations, to definitive where the results can be interpreted without error:
1 Grain is scattered once a month but density is not monitored. The manager assumes
that since the treatment shouldincrease the density of quail that it willincrease their
density.
This is no test because the outcome of the treatment is assumed rather than observed.
2 The manager measures density on two occasions separated by 1 year, the first before
supplementary feeding is instituted. If density were higher on the second occasion
the manager might assume that the rise resulted from the feeding.
This is the classic fallacy of “before” being taken as a control on “after.” Interpreta-
tion of the result rests on an assumption that the density would have remained stable
without supplementary feeding, and there is no guarantee of that. It may, for example,
have been increasing steadily for several years in response to a progressive and general
increase of cover.
3 The manager designates two areas, one on which the birds are fed (the treatment)
and the other on which they are not (the experimental control). The density of quail
is measured before and after supplementary feeding is instituted. If the proportion-
ate increase in density on the treatment area is greater than that on the control area,
the difference is ascribed to the effect of feeding.
This design is a radical improvement but still yields ambiguous results: the dif-
ference in rate of increase may reflect a difference between the two sites rather than
between the two treatments. We say that the effect of site and the effect of treatment
are confounded. Perhaps the soil of one site was heavy and that of the second light,
the vegetation on the two sites thereby reacting differently to heavy winter rainfall
and the quail reacting to that difference in plant growth.
Flawed as it is, this design is often the only one available, particularly if the treated
area is a national park. In such cases the control must be chosen with great care to
ensure that it is in all important respects similar to the treated area, and the response
variable should be monitored on each area for some time before the treatment is insti-
tuted to establish that it behaves similarly in the two areas. Another way around this
problem is to reverse the two treatments and see if the same result is obtained.272 Chapter 1616.4 The nature of the evidence
16.4.1Experimental
evidenceWECC16 18/08/2005 14:47 Page 272