farmer may harvest the whole field before data are
collected. An advantage of the sampling procedure is that
the farmer has no means of previously identifying the
check area, so it cannot be treated differently from the
rest of the field.
 If check measurements are to be taken by sampling, it
is important that the locations be chosen completely at
random. One convenient method is to choose a row and a
distance into the field (in that row) from a table of
random numbers, and to harvest a set distance in that row
(probably the net row length in the trial plot). The
procedure is repeated until sufficient replications have
been made to make the area harvested by sampling equal to
the net size of the treatment plot. Another method of
measuring the results of the farmer's practice is to
locate a check plot on paper, positioned randomly in the
field, and mark its location at the time of harvest.
 Plots should be large enough to allow realistic
estimates of labor requirements and other economic
factors. They should also facilitate evaluation by
farmers. The unit or units used should be related to
those commonly used by farmers in the area (hectare,
manzana, cuerda, tarea, etc.). It is important that the
area be large enough on each farm that it can be of
significance to the farmer. If possible, 10% to 20% of
the area allocated for that crop on each farm should be
included, so that the farmer must give it appropriate
attention.

Number of Sites

 The number of sites needed in an FMT to test an
alternative in a homogeneous recommendation domain will
depend on several factors, including:
1) the nature of the alternative to be tested;
2) the budget available to researchers;
3) demands on their time for other kinds of trials; and
4) the number of farmers in the recommendation domain.

Usually, the fourth factor is not limiting. The first
factor is important because a more complex technology or
alternative requires more learning time from each farmer,
and the researcher will have to visit each farm more often
for supervision.