Forest fragmentation can alter herbivory on tree recruits with possible consequences for regeneration. We assessed effects of forest-fragment quality (tree diversity, vegetation complexity, relative abundance of pioneer trees) and matrix habitat on arthropods and herbivory in KwaZulu-Natal, South Africa. We compared arthropod abundances and herbivory on woody seedlings and saplings among four forest-fragment types differing in size and matrix (large fragments and small fragments surrounded by natural grassland, eucalypt and sugarcane plantations; nplots = 24) using analyses of covariance. We recorded 3385 arthropods and inspected 897 seedlings (71 species) and 876 saplings (91 species). Relative abundance of predators increased with fragment quality; that of herbivores decreased. Herbivory responses to fragment quality varied: seedling herbivory decreased with relative abundance of pioneers and sapling herbivory increased with vegetation complexity. Matrix effects were low with little variation in relative abundance of predators (0.39–0.53) and herbivores (0.22–0.32), proportion of seedling (8.3–11.0%) and sapling herbivory (12.4–14.3%) among the forest-fragment types. These findings indicate that herbivory on tree recruits is mediated by forest-fragment quality rather than matrix habitat. Future studies should evaluate whether contrasting effects of fragment quality on arthropods and herbivory are caused by weak trophic interactions and variable herbivore compositions.

Most seedlings and saplings remain in the forest understorey for decades before becoming adults or dying, and thus may be exposed to various sources of physical disturbance (Clark & Clark 1991, 2001). Because tree species vary in their ability to recover after physical damage (Gillman et al. 2003, Guariguata 1998), this damage can act as an ecological filter and influence the juvenile community structure and species composition (Peters et al. 2004). Studies have demonstrated the occurrence and magnitude of stem breakage in juveniles caused by falling branches and other canopy debris (Clark & Clark 1989, 1991; Gillman & Ogden 2001, Scariot 2000). Surprisingly, little is known about the magnitude and ecological consequences of physical damage to juvenile plants by mammals, particularly large ungulates, including herbivory, trampling and uprooting (Gillman & Ogden 2003, Roldán & Simonetti 2001).

A prediction of the hypothesis that niche processes control plant species distributions across edaphic gradients is that pairwise comparisons of related species on contrasting soils should show consistent patterns of trait divergence. This hypothesis was tested in lowland Bornean rain forest by combining measurements of leaf traits, soil nutrients and the distribution of tree species in the Dipterocarpaceae across a nutrient gradient. Nine species were studied, comprising four related pairs in Shorea (phylogenetic independent contrasts; PICs) and one habitat generalist (Dipterocarpus sublamellatus). Lamina area, specific leaf area (SLA), leaf nitrogen (N) and phosphorus (P) were measured for ∼10 saplings of each species; habitat associations were defined as a continuous function of soil nutrients and categorically in relation to substrate. Species distributions and traits varied significantly with soil nutrients. When all species were pooled, SLA and leaf P increased significantly with species' distribution across the nutrient gradient; leaf N showed the same trend. Trait shifts with habitat were stronger and more pervasive when habitat was defined categorically – in all four PICs, rich-soil specialists had larger leaves, higher SLA, leaf N and P, and lower N:P ratios. Trait shifts with habitat at least partly reflect intrinsic differences between species. Within Shorea, variation in lamina area and N:P ratio were more strongly related to phylogeny than habitat, whereas the reverse held for SLA and leaf P. Phylogeny also influenced the extent of trait divergence between related species on different soils, and patterns of trait correlation within lineages. Results support the hypothesis that niche processes influence the distribution of species and traits in lowland tropical tree communities, and highlight the value of phylogenetic information for increasing the power of comparative studies.

Fire is a common feature of tropical savannas and it has an important role in the ecology and evolution of the flora. Many woody species in tropical savannas are well adapted to fire and have a vigorous resprouting ability that enables them to survive recurring fire (Frost & Robertson 1987, Hoffmann & Solbrig 2003, Meyer et al. 2005). Studies on the interaction between fire and woody plants in Mediterranean-climate regions have considerably increased our understanding of the subject but relationships between plant attributes (stem diameter and height) and post-fire resprouting ability of southern African savanna woody species are poorly understood. Resprouting is a widespread mechanism by which many plants regenerate after partial or total defoliation (Bellingham & Sparrow 2000, Bond & Midgley 2001).

Light fluctuations and crown traits were studied for saplings of four tree species in a Costa Rican rain forest. Light fluctuations (1988–1994) were assessed by annual light estimations above saplings, using a visual crown position index. Crown traits in 1994 and growth between 1994 and 1995 were measured. Crown position values varied between 1.5 and 2.5. Of the 70 saplings only four were found at higher light levels (crown position of 3 or 3.5), but for 1–2 y only. Over the 6 y of investigation, 55–75 % of the saplings experienced no or only one light fluctuation, and 25–45 % two or three fluctuations. Crown traits in 1994 were either most strongly correlated with light levels in 1993 (Lecythis) and with light levels in 1991 and 1992 (Dipteryx and Simarouba), or they were not significantly correlated with light levels (Minquartia). It is hypothesised that: (1) the saplings require 1–3 y to establish a crown trait in response to light levels in the forest; and (2) and species that can economically produce a leaf can adjust crown traits more quickly in an environment dominated by low light levels than species that are less economic. Crown trait responses may track environmental changes in three of the four species, in particular in Lecythis. In this latter species, leaf area had no significant effect on growth and survival, whereas light level had a positive effect. Conversely, in the other three species, light levels had no discernible effect on growth and survival (due in part to low variation in crown position in two of these species), whereas leaf area had a positive effect on both.