The site was clearly very variable with the rows climbing steeply from row 1 to row 5 and with poor
growth at the right hand side of the area which curled some way round the hill. Sensible groupings would
appear to be
 (A) row 5,
 (B) rows I and 2
with either (Cl)row 3 or (C2) plots (7,8,9,13,14,15)
and (DI)row 4 and (D2) plots (10,11,12,16,17,18)

The actual design had two blocks of twelve plots (for the twelve treatments) and there was plainly a great
deal of variation within the blocks (to be fair some of that variation would have remained even within the
better blocks.

One more example from the Poza Rica experiments to show how too swift assumptions about natural
patterns may lead to faulty conclusions. The 14 plots (for seven treatments) were as shown:


14 13 12 11 10 9 8
1 2 3 4 5 6 7


There was a small road running along the side of plots 1 to 7. Each row of seven plots covered a long area
and it seemed unlikely that plots (1 to 7) and (8 to 14) would be sensible blocks. However when walking
between the two blocks there was a very strong suggestion that those plots on the left (plots 1 to 7) for each
treatment showed better crop growth than those on the right. This impression can obviously be verified at
harvest but if true demonstrates how first impressions can be misleading. Further examination of the site
showed that the second block (8 to 14) was clearly, if only slightly, higher than the first, providing further
justification for the blocking pattern used.

In each on-farm experiment it is important, at the planning stage, to identify how the plots should be
grouped so as to both minimise variation within groups and maximise that between groups. This may, and
perhaps often should, produce groups of different sizes and later in the design process these ideal groupings
may be modified. Nevertheless it is vital that the ideal grouping be first identified.

4.2. Recording and Use of Ancillary Information

Whenever experimenters, or even statisticians, observe field plots during an experiment there are many
differences and patterns of growth which are apparent. Two which we observed during observation of the
Poza Rica experiments concerned patchiness of Johnsongrass and a possible trend along each block away
from the edge of the field. Field notebooks should, and often do, include anecdotal information collected in
a systematic manner. If that anecdotal information were recorded in a crude quantitative form it could be
used, later, through covariance analysis to improve the precision of treatment comparisons. The minimum
form of record would be presence(l) or absence(0) of a characteristic. More usefully a three- or two- point
scale to record nil(0) mild(l) substantial(2) or total(3) level of the defect characteristic should allow
adjustment for the effect of the characteristic.

The possible distance adjustment occurred for a tillage and cover crop experiment, for which the plots were
as shown.


Orange 15 16 17 18 19 20 21
 14 13 12 11 10 9 8
Orchard 1 2 3 4 5 6 7