There is indirect evidence in the literature that, when calculated by the classical method of growth analysis involving unrestricted random pairing of replicate plants between two harvests, the within-sample variability of mean relative growth rate is greater than it should be. An experiment is described in which two parallel batches of plants, each of four species, were grown in solution culture. The f. wts of all the replicate plants in one batch were measured non-destructively at each of four sampling times, while the other batch was sub-sampled destructively at the same times for both f. wts and d. wts.In general, within any one species, there were no differences between the mean RGRs over any one time interval, whether calculated from the f. wts of the continuously-monitored plants or from f. wts or d. wts of the sub-sampled plants. However, the within-sample variability of the continuously monitored plants was always lower, and usually much lower, than those of the sub-sampled plants. It was concluded that this result provided further indirect evidence that the true within-sample variability of RGR is lower than that indicated by classical growth analysis involving unrestricted random pairing. A correction factor, which may be applied to classical growth analysis experiments, is suggested. © 1991 Annals of Botany Company.
