Core subsets of germplasm collections facilitate their evaluation and
use. Field evaluations were conducted to develop a core subset of Cari
bbean maize (Zea mays L.) accessions from the CIMMYT maize germplasm b
ank. Most accessions were from the West Indies and others were from Ce
ntral and South America. The accessions were divided into two trials o
f 249 accessions and seven common checks that were evaluated in a 16 b
y 16 alpha-lattice design with two replications. The trials were plant
ed in two seasons at two sites in Mexico during 1992 through 1994. Dat
a were taken on agronomic and morphological traits and the combined an
alysis was performed by a mixed linear model. The adjusted means of pl
ant height, ear height, senescence, ear diameter, ear length, days to
silk, days to anthesis, moisture (%), rating on ease of shelling, and
number of kernel rows were used to cluster accessions into homogeneous
groups. Twelve non-overlapping dusters were formed and patterns of ph
enotypic diversity among and within clusters were determined using can
onical discriminant analysis. A selection index based on yield (Mg ha(
-1)), ear rot (%), erect plants (%), and moisture (%) was calculated f
or each accession. The upper 20% of the accessions (100 in total) whic
h represent the phenotypic diversity of the clusters and have superior
selection indexes were used to form a core subset of the Caribbean ma
ize collection.