529, Report oF THE HortTicuLTURIST OF THE 
3-pound to 5-pound sizes being preferred. ‘The smaller sizes may 
cramp the larger clusters and the larger sizes offer too much resist- 
ance to the wind. ‘The shoots on which the clusters are borne 
being still green and tender at blooming time are more lable to 
be broken off in wind storms when the large bags are attached to 
them than they are when the smaller sizes are used. 
In preparing a bag for covering a cluster it is slit downward 
from the mouth for a short distance on opposite sides so that the 
open end may include and project beyond the shoot on which the 
cluster is borne as shown in Figure 1. The mouth is then closed 
and fastened with a wired label as shown by Figure 2. In this way 
a portion of the shoot is included which holds the bag more firmly 
than when it is fastened to the stem of the cluster only and the 
cluster is less apt to be broken off. 
NUMBER OF VINES AND CLUSTERS TESTED. 
Number of clusters.— The number of clusters tested among the 
cultivated grapes varies from four to seventy-one and averages 
about fifteen to the variety. This does not include the varieties 
Little Blue, Norfolk and Illinois City. No account is taken of 
clusters which have been lost or destroyed by accident. 
Number of vines.— No exact record has been kept of the num- 
ber of vines tested to the variety. It varies from one to perhaps 
ten or more. “Probably in the majority of cases at least two vines 
have been tested to the variety. 
Tests of but few clusters.— Are tests of two or three clusters 
on a single vine sufficient to show the characteristic degree of self- 
fertility of a variety? The records of these tests give some evi- 
dence on this question. Nineteen varieties were tested by cover- 
ing only two clusters of each in one season. The same varieties 
were afterwards tested by covering a larger number of clusters. 
The latter tests gave results practically similar to those of the first 
tests except in the case of three Rogers hybrids which in the first 
tests gave only perfect clusters and in the later tests gave some 
