Summary1 Regeneration in forest canopy gaps is thought to lead invariably to the rapid recruitment and growth of trees and the redevelopment of the canopy. Our observations, however, suggest that an alternate successional pathway is also likely, whereby gap-phase regeneration is dominated by lianas and stalled in a low-canopy state for many years. We investigated gap-phase regeneration in an old-growth tropical forest on Barro Colorado Island (BCI) in Panama to test the following two hypotheses: (i) many gaps follow a pathway in which they remain at a low canopy height and are dominated by lianas and (ii) the paucity of trees in this pathway is a function of liana density. 2 We surveyed a total of 428 gaps of varying ages (c. 5, c. 10, and 13 years old) and identi®ed those which followed the conventional pathway of regeneration and others that remained stalled in a low-canopy state for many years and were dominated by either lianas or palms. Each of these pathways will likely have dierent successional trajectories that will favour the growth of a distinct suite of mature species and ultimately result in contrasting species composition. 3 The successional pathway of liana-dominated, stalled gaps is common throughout the forest. We estimate conservatively that 7.5% of the gaps that form each year will follow this pathway, probably due to the suppression of tree regeneration by lianas, and that many of these stalled gaps will persist for much longer than 13 years. Consequently, a high proportion of gaps in the forest at any given time will be stalled. Furthermore, liana tangles, which persist in the tropical forest understorey for extended periods of time, almost certainly originate in these gaps. 4 Liana abundance was positively correlated with pioneer tree abundance and diversity while negatively correlated with non-pioneer tree abundance and diversity. Thus, lianas appear to inhibit non-pioneer tree survival while indirectly enhancing that of pioneer trees. 5 Lianas are abundant in many types of tropical and temperate forests and a successional pathway involving liana-dominated, stalled gaps may therefore be frequent and widespread.

The mechanistic basis underpinning forest succession is the gap-phase paradigm in which overstory disturbance interacts with seedling and sapling shade tolerance to determine successional trajectories. The theory, and ensuing simulation models, typically assume that understory plants have little impact on the advance regeneration layer's composition. We challenge that assumption by reviewing over 125 papers on 38 species worldwide that form dense and persistent understory canopies. Once established, this layer strongly diminishes tree regeneration, thus altering the rate and direction of forest succession. We term these dense strata recalcitrant understory layers. Over half of the cases reviewed were linked to increases in canopy disturbance and either altered herbivory or fire regimes. Nearly 75% of the studies declared that competition and allelopathy were the likely interference mechanisms decreasing tree regeneration, yet only 25% of the studies used manipulative field experiments to test these putative mechanisms. We present a conceptual model that links the factors predisposing the formation of recalcitrant understory layers with their interference mechanisms and subsequent impacts on succession. We propose that their presence constricts floristic diversity and argue for their explicit inclusion in forest dynamics theory and models. Finally, we offer management suggestions to limit their establishment and mitigate their impacts.

The maintenance of species diversity by treefall gaps is a long-standing paradigm in forest ecology. Gaps are presumed to provide an environment in which tree species of differing competitive abilities partition heterogeneous resources. The empirical evidence to support this paradigm, however, remains scarce, and some recent studies even suggest that gaps do not maintain the diversity of shade-tolerant species. Although there is evidence that gaps maintain the diversity of pioneer trees, most of this evidence comes from studies that did not make comparisons between gaps and intact forest sites (controls). Further, nearly all studies on the maintenance of diversity by gaps have ignored lianas, an important component of both old-world and neotropical forests. We tested the hypothesis that treefall gaps maintain shade-tolerant tree, pioneer tree, and liana species diversity in an old-growth forest on Barro Colorado Island (BCI), Panama. We compared the density and species richness of these guilds between paired gap and non-gap sites on both a per-area and a per-individual (per capita) basis. We found no difference in shade-tolerant tree density and species richness between the gap and non-gap sites. Both pioneer tree and liana density and species richness, however, were significantly higher in the gap than in the non-gap sites on both a per-area and a per-individual basis. These results suggest that gaps maintain liana species diversity and that this effect is not merely a consequence of increased density. Furthermore, our data confirm the long-held belief that gaps maintain pioneer tree species diversity. Because lianas and pioneer trees combined account for 43% of the woody plant species on BCI, and in other forests, our results are likely to be broadly applicable and suggest that gaps play a strong role in the maintenance of woody species diversity.

Treefall gaps are hypothesized to maintain diversity by creating resource-rich, heterogeneous habitats necessary for species coexistence. This hypothesis, however, is not supported empirically for shade-tolerant trees, the dominant plant group in tropical forests. The failure of gaps to maintain shade-tolerant trees remains puzzling, and the hypothesis implicated to date is dispersal limitation. In central Panama, we tested an alternative Ôbiotic interferenceÕ hypothesis: that competition between growth forms (lianas vs. trees) constrains shade-tolerant tree recruitment, survival and diversity in gaps. We experimentally removed lianas from eight gaps and monitored them for 8 years, while also monitoring nine un-manipulated control gaps. Removing lianas increased tree growth, recruitment and richness by 55, 46 and 65%, respectively. Lianas were particularly harmful to shade-tolerant species, but not pioneers. Our findings demonstrate that competition between plant growth forms constrains diversity in a species-rich tropical forest. Because lianas are abundant in many tropical systems, our findings may apply broadly.

No abstract