Neighborhood competition in several violet populations.

Research into local (neighborhood) competition has heretofore focused on either natural populations of long-lived woody plants or artifical populations of herbaceous annuals. To explore the usefulness of these techniques for herbaceous perennials, 1 applied them to 11 natural populations of 4 species of violet (Viola blanda, V. pallens, and V. incognita, alone and with V. adunca). Ramet density ranged from 132 to 508 per square meter. The analysis tried to account for the size of each ramet (expressed as the number of leaves) using information on the number, size, and angular aggregation of neighboring plants in three concentric annuli (0-3, 3-6, and 6-9 cm). Simple rank correlations between leaf number and these measures of local competition were usually of the expected sign, but were consistently significant in only two of the populations (1 pallens and 1 incognita). No significant correlations were of the "wrong" sign. When the spatial independent variables were combined into multiple regression equations they accounted for between 5 and 59% of the total variance in leaf number. The equations were best in predicting individual size in the Newfoundland populations of V. incognita which had the greatest range in individual size. Total plant density appeared unrelated to predictive success.The neighborhood competition approach can be applied to natural populations of herbaceous perennials, but low density, inaccurate measures of plant performance and microvariation in site quality limit its usefulness. In particular, it seems impossible to determine the mechanism by which neighbors exert their competitive effects with this approach.