The regions established for the National Cotton Variety Testing Program, at its inception, were contiguous geographical areas, delineated by approximate north south divisions. Abou-El-Fittouh, Rawlings and Miller (Crop Science 9:135-140. 1969.) suggested a revision of the regions that assigned northern locations to different groupings than the southern locations. However, at the time the data were not sufficient to assign several locations to specific groups.

The PKM or K-Means clustering algorithm of the BIOMED program was used to determine groupings from the 1966, 1967 and 1968 National Cotton Variety Testing Program. Again the responses in the northern locations appeared to differ from those in the southern locations. Furthermore, the responses in the western area appeared to be similar to those in the southern, rainbelt. The determination of groupings from the 1969, 1970 and 1971 National Cotton Variety Testing Program further confirmed the north-south differentiations, but groupings differed somewhat from those found in the 1966, 1967 and 1968 tests. Additional data are being analyzed in order to determine an optimal grouping of locations.