between the two: smaller, nonshowy, maroon flowers with a fetid odor; and larger, showy (white, yellowish, or pink) flowers with a fragrant aroma.

Fruit varies in size from 4 gm to 400 gm. The flavor varies from sweet, aromatic and delicious to bitter, resinous and insipid. The skin ranges from thin and tender to rough, leathery, and nearly hard. The proportion of fruit which is seed varies from 4 percent to 45 percent.

Seeds vary from 1 to 2.5 cm in length, and from .2 to 2.0 gm in weight. The shape varies from oblong-flat to round; color varies from tan to chestnut to ebony; the surface varies from dull to lustrous.

The dwarf pawpaws exhibit considerable variation within each species. Fruit quality is not consistently poor. Although most are terrible, some are merely poor, and occasional clones have good flavor; for instance, the author tasted an A. reticulata fruit in 1980 whose flavor was typical of a good A. triloba. The seed:fruit ratio varies considerably, from a low of 11 percent to a high of 45 percent and a mean of 28 percent. The skins are usually tough and thick, even hard. Some clones of A. obovata have tender skin like A. triloba. Asimina obovata is also notable for having a variety of fruit skin tones; the green ground color, similar to A. triloba, is often overlaid with shades of lavender and coated with a dense waxy bloom. Other species have skins that are deep brown or black (A. incana and A. longifolia), or that are coated with a fine reddish pubescence (A. incana and A. parviflora).

Few reports have discussed variation in *A. triloba* in the context of breeding (Little, 1905; Amer. Genet. Assn., 1917; Ourecky and Slate, 1975; McKay, 1975) and even fewer have presented quantitative data7 (Peterson et al, 1982; Peterson, 1989). Since 1986, the author has taken measurements of tree growth, blossom count, cluster count and fruit count, individual fruit weight, seed weight, etc. from over 130 individual six-year old seedling trees in the collection at Wye, Maryland, U.S.A. Table 4 presents statistics from these data. Obviously, these data are not representative; the trees were in their first year of bearing; and furthermore, the Wye collection is a decidedly nonrandom sample of the *A. triloba* population. These data, however, being the first such available, tentatively provide a baseline for later studies assessing variation in horticulturally relevant traits, and begin to establish a context for judging the breeding potential of individual clones.

Of the 13 variables listed in Table 4, five appear normal, that is, distributed fairly symmetrically about their mean with relatively small coefficients of variation. These normal variables are height, growth rate, seed size, the number of seeds per fruit, and the seed:fruit ratio. The remaining 8 variables are highly skewed, i.e., distributed nonsymmetrically as shown by a great disparity between the maximum and mini-

587