LETTER How specialised must natural enemies be to facilitate coexistence among plants? Brian E. Sedio* and Annette M. Ostling Department of Ecology and Evolutionary Biology, University of Michigan, 830 N. University, Ann Arbor, MI, 48109, USA *Correspondence: E-mail: bsedio@umich.edu Abstract The Janzen-Connell hypothesis proposes that plant interactions with host-specific antagonists can impair the fitness of locally abundant species and thereby facilitate coexistence. However, insects and pathogens that associate with multiple hosts may mediate exclusion rather than coexistence. We employ a simulation model to examine the effect of enemy host breadth on plant species richness and defence community structure, and to assess expected diversity maintenance in example systems. Only models in which plant enemy similarity declines rapidly with defence similarity support greater species richness than models of neutral drift. In contrast, a wide range of enemy host breadths result in spatial dispersion of defence traits, at both landscape and local scales, indicating that enemy-mediated competition may increase defence-trait diversity without enhancing species richness. Nevertheless, insect and pathogen host associations in Panama and Papua New Guinea demonstrate a potential to enhance plant species richness and defence-trait diver- sity comparable to strictly specialised enemies. Keywords Coexistence, community structure, frequency dependence, host specialisation, Janzen-Connell, plant–enemy interactions, species richness, tropical forest. Ecology Letters (2013) 16: 995–1003 INTRODUCTION How large numbers of plant species manage to co-exist in the face of intense competition for light, water and other shared resources remains a fundamental challenge to community ecology (Wright 2002; Silvertown 2004). A now-classical hypothesis put forth inde- pendently by Janzen (1970) and Connell (1971) proposes that plants fail to recruit in the neighbourhood of conspecific adults as a result of attack by specialised natural enemies, such as insects or patho- gens, that respond either to the presence of adult plants or to the density of offspring of their host species. This mechanism is thought to limit the abundance of any given species, and thereby facilitate coexistence among plants. A considerable amount of evidence has accrued in the last 40 years in support of the predictions of Janzen and Connell, namely that offspring fail to survive at high densities or in the vicinity of conspecific adults in tropical forests (e.g. Wills et al. 1997; Webb & Peart 1999; Harms et al. 2000; HilleRisLambers et al. 2002; Comita et al. 2010; Terborgh 2012). Yet, recent large- scale community surveys suggest that within at least some commu- nities of both insect herbivores (Basset 1992; Odegaard et al. 2000, 2005; Novotny et al. 2002, 2010) and fungal pathogens (Gilbert 2005; Gilbert & Webb 2007; Liu et al. 2012), enemies seldom spe- cialise on a single host, but rather associate with a range of often closely related species. In addition, recruitment patterns consistent with the expectations of Janzen and Connell are not restricted to the tropics (Johnson et al. 2012), despite some suggestions that host specificity declines with latitude (Dyer et al. 2007). More broadly, knowledge of how specialised enemies are, and how large of an effect specialised enemies have on host fitness as compared with their more generalist counterparts, is rather limited. Shared natural enemies may result in coexistence or competitive exclusion depending on the degree to which competitors partition niche space defined by antagonists (Bever 2003; Chesson & Kuang 2008). Hence, it remains unclear whether plants differ in their interactions with enemies sufficiently to facilitate the maintenance of diversity in tropical forests (Freckleton & Lewis 2006) or else- where. Are plant antagonists specialised enough to facilitate coexis- tence? How specialised must they be? Even if natural enemies are not specialised enough to foster coex- istence of competing plant species, they may still influence the phy- logenetic or defence trait composition in plant communities. The degree to which plant species share herbivores (Novotny et al. 2002, 2010; Odegaard et al. 2005) or pathogens (Gilbert & Webb 2007; Liu et al. 2012) is broadly associated with plant phylogeny, most likely as a result of conservatism of host use-determining traits at deep scales of phylogeny (Wink 2003; Barrett & Heil 2012). As a result, the exclusion of related heterospecific individuals through competition mediated by shared enemies (Holt 1977) may be par- tially responsible for observations of phylogenetically even (‘over- dispersed’ sensu Cavender-Bares et al. 2004) species assemblages at small spatial scales (Webb et al. 2006; Bagchi et al. 2010; Metz et al. 2010). On the other hand, detailed investigations of Asclepias milk- weeds (Agrawal & Fishbein 2006) and the tropical tree genera Bur- sera (Becerra 1997) and Inga (Kursar et al. 2009) suggest that plant defences can be quite evolutionarily labile at finer scales of phylog- eny, such as within a genus. Furthermore, communities of co-occur- ring Bursera (Becerra 2007) and Inga (Kursar et al. 2009) appear to exhibit a more even distribution of defences than by chance. These findings suggest that density-responsive insects and pathogens increase the diversity of the plant community with respect to defence traits, and potentially increase plant species richness in the process. Furthermore, by allowing phenotypically distinct plants to escape the enemy load of neighbouring heterospecifics, nonspecialist enemies may promote divergence in defence among closely related species, thus linking the Janzen-Connell (J-C) mechanism of diver- sity maintenance with the role of enemies in promoting plant line- age diversification posited by Ehrlich & Raven (1964). © 2013 John Wiley & Sons Ltd/CNRS Ecology Letters, (2013) 16: 995–1003 doi: 10.1111/ele.12130