Journal of Tropical Ecology (2006) 22:719–722. Copyright © 2006 Cambridge University Press doi:10.1017/S026646740600349X Printed in the United Kingdom SHORT COMMUNICATION Positive effect of seed size on seedling survival in fire-prone savannas of Australia, Brazil and West Africa Ga¨ elle Lahoreau ∗1 ,S´ ebastien Barot † , Jacques Gignoux ∗ , William A. Hoffmann ‡ , Samantha A. Setterfield § and Paul R. Williams # ∗ Laboratoire de Biog ´ eochimie et d’Ecologie des Syst ` emes Continentaux, UMR 7618, Ecole Normale Sup ´ erieure, 46 rue d’Ulm, 75230 Paris Cedex 05, France † Laboratoire d’Ecologie des Sols Tropicaux, UMR 137, IRD, 32 avenue H. Varagnat, 93143 Bondy, France ‡ Department of Botany, Campus Box 7612, North Carolina State University, Raleigh, NC 27695-7612, United States § Faculty of Education, Health and Science, Charles Darwin University, Darwin, NR 0909, Australia # Queensland Parks and Wildlife Service and School of Tropical Biology, James Cook University, PO Box 5597 Townsville, Q 4810, Australia (Accepted 18 May 2006) Key Words: buds, resource allocation, resprouting, root reserves, seed mass, seedling size All plant species face a fundamental reproductive trade- off: for a given investment in seed mass, they can produce either many small seeds or few large seeds. Whereas small seeds favour the germination of numerous seedlings, large seeds favour the survival of seedlings in the face of common stresses such as herbivory, drought or shade (Leishman et al. 2000). One mechanism explaining the better survival of large-seeded species is the seedling size effect (SSE) (Westoby et al. 1996): because seeds with large reserves result in bigger seedlings, seedlings from large-seeded species would have better access to light and/or to reliable water supply than seedlings from small- seeded species. However, the relationship between seedling size and seed mass usually weakens through time, as seedling relative growth rate tends to be negatively correlated with seed mass (Leishman et al. 2000). Therefore, the advantage of large-seeded species is expected to be greatest during early seedling establishment and to diminish until seed reserves have been fully exhausted (Leishman et al. 2000, Moles & Westoby 2004). In accordance with this expectation, Dalling & Hubbell (2002) and Walters & Reich (2000) found no relationship between the survival probability of well-established seedlings and seed size in a species comparison experiment. In contrast, Baraloto et al. (2005), Hoffmann (2000) and Moles & Westoby (2004) found a positive relationship. With the exception of Hoffmann (2000), all these studies were on forest tree species. 1 Corresponding author. Email: gaelle.lahoreau@gmail.com As in forests, seedling survival is a critical process in savannas that influences both the abundance and species diversity of trees. But in contrast to forests, the factors affecting seedling survival, and potentially selecting for seed size, are expected to be different in savannas. In contrast to forests, seedlings in savannas are strongly competing with grass for access to light, water and nutri- ents (Kanegae et al. 2000, Nardoto et al. 1998). Conse- quences of competition are more severe for small seedlings than for tall seedlings due to the higher grass biomass encountered. The smaller seedling size of small-seeded species is thus likely to persist, as they are continually unfavoured. We therefore expected the SSE to last in savannas even after exhaustion of the seed reserves. Fire is also a major constraint on seedling survival in moist savannas, which commonly burn at intervals of 1–3 y. Grass fire almost invariably destroys the entire shoot of seedlings so that seedlings must be able to resprout to survive (Bond & Midgley 2001). Resprouting requires sufficient root reserves of total non-structural carbohydrates (TNC) and nutrients to regenerate shoots and leaves, which depend both on the root biomass and on the resource allocation strategy of the species (Bond & Midgley 2001, Canadell & Lopez-Soria 1998). Tree species with high root:shoot ratio are indeed likely to accumulate more TNC in roots than tree species with low root:shoot ratio (Chapin et al. 1990). Although differences in resource allocation among tree species may weaken the relationship, we expected an overall positive effect of seed size on seedling survival due to the SSE in savannas. To test this hypothesis, we re-analysed data on savanna tree species available in