Complex data produce better characters.

Abstract

Two studies were conducted to explore the use of complex data in character description and hybrid identification. In order to determine if complex data allow the production of better characters, eight groups of plant systematists were given two classes of drawings of plant parts, and asked to divide them into character states (clusters) in two separate experiments. The first class of drawings consisted only of cotyledons. The second class consisted of triplets of drawings: a cotyledon, seedling leaf, and inflorescence bract. The triplets were used to simulate complex data such as might be garnered by looking at a plant. Each experiment resulted in four characters (groups of clusters), one for each group of systematists. Visual and statistical analysis of the data showed that the systematists were able to produce smaller, more precisely defined character states using the more complex drawings. The character states created with the complex drawings also were more consistent across systematists, and agreed more closely with an independent assessment of phylogeny. To investigate the utility of complex data in an applied task, four observers rated 250 hybrids of Dubautia ciliolata X arborea based on the overall form (Gestalt) of the plants, and took measurements of a number of features of the same plants. A composite score of the measurements was created using principal components analysis. The correlation between the scores on the first principal component and the Gestalt ratings was computed. The Gestalt ratings and PC scores were significantly correlated, demonstrating that assessments of overall similarity can be as useful as more conventional approaches in determining the hybrid status of plants.