Background: If the amount of resources allocated to reproduction (K) is fixed, then an increase in seed mass (S) can only be achieved by a decrease in seed number (n = K/S). Thus, log(n) = log(K)-log(S) producing a slope of -1 when seed mass and number are plotted on log-log axes. However, in comparative studies, empirical support for a slope of -1 is limited and contentious, leading some to question the utility of this concept.
Methodology/principal findings: First, we show that the expected slope depends on whether genotypes and species producing seeds of different mass are expected to reach the same adult size and that this in turn depends partly on the nature of growth. Second, we present experimental results using a population of recombinant inbred lines (RILs) of Arabidopsis thaliana. When these RILs are grown in large pots with plentiful nutrients, they exhibit a trade-off between seed size and number with a slope of -1.68 (+/-0.18) on log-log axes. This occurs because of genetic correlations between seed mass and adult size so that, under the right growth conditions, lines producing lighter seeds have the genetic potential to produce larger rosettes and hence a greater total mass of seeds. We re-grew lines in small pots (10 and 40 mm diameter) in a nutrient-poor substrate so that final adult size was heavily restricted by pot size.
Conclusions/significance: Under our growth conditions, small-seeded lines were unable to produce a greater total mass of seeds. Hence a trade-off emerged between seed mass and seed number with a slope of -1.166+/-0.319 on log-log axes in 40-mm diameter pots (close to the expected value of -1), although the slope was 0.132+/-0.263 in 10-mm diameter pots, demonstrating that the nature of the trade-off is sensitive to the growth conditions.