Numerical Analysis of Seed Morphology in Cucurbita pepo

Quantitative data were collected for 24 commercial cultivars of Cucurbita pepo, four C. pepo collections from Mexico, six populations of C. texana from Texas, and three spontaneous populations of texana-like plants from Alabama, Arkansas, and Illinois. Canonical variate and prin- cipal component analyses revealed variation in the size and shape of C. pepo seeds which was similar to recently documented patterns of allozyme variation. Additionally, variation in seed characters appears to reflect effects of human selection. Stepwise selection was used to select four characters useful in discriminating between C. texana and C. pepo var. ovifera seeds. Most significant were characters describing the sinus area near the seed scar. A discriminant function based on these characters was capable of correctly identifying seeds with 86% accuracy, indicating inherent differences between these taxa. On the basis of this discriminant function, the spontaneous popu- lations from Alabama, Arkansas, and Illinois were classified. Although 80% of the seeds from these populations were classified as C. pepo var. ovifera, the general nature of these populations was one of intermediacy between the two taxa. This intermediacy is an important consideration in inter- preting the relationship between C. texana and C. pepo. Although generally assumed to be feral populations of relatively recent origin, these populations could also represent remnants of the wild species in an area northeast of its current distribution in Texas. Cucurbita pepo L., represented by cultivated