The study was conducted between 2019 and 2023 in the tropical montane humid forests in the Bolivian Andes, in the vicinities of Chulumani and Irupana localities, Sud Yungas, La Paz, Bolivia (16°24’37” S, 67°31’37” W and 16°27’24” S, 67°25’50” W), between 1,850 and 2,450 m asl ( Supplementary Figure S1 ). The mean annual temperature is 20.5°C and the mean annual precipitation is about 1,390 mm (Molina-Carpio et al., 2019). Due to frequent and uncontrolled anthropogenic fires and the expansion of Erythroxylum coca plantations, the montane forest have become highly fragmented (Killeen et al., 2005, 2008; Beck et al., 2024).

The only two large remaining patches of continuous primary montane humid forest in the area, located at the top of the mountains, cover approximately 1,200 and 4,000 ha, respectively (Lippok et al., 2014; Beck et al., 2024). These forests are dominated by tree species such as Alchornea brittonii, Beilschmiedia latifolia, Couepia sp. nov., Coussapoa david-smithii, Ficus crassiuscula, Hedyosmum cuatrecazanum, Ladenbergia oblongifolia, Miconia plumifera, Pectinopitys harmsiana, Piper bolivianum, Tetrorchidium andinum, Weinmannia rhoifolia and different tree ferns (Beck et al., 2024). Surrounding these forests are extensive areas dominated by the bracken fern Pteridium esculentum (Schwartsburd et al., 2018), wind-dispersed trees such as Clethra scabra and Weinmannia sorbifolia, shrubs like Baccharis spp., and some animal-dispersed tree species like Myrsine coriacea, Piper elongatum and Clusia trochiformis, and shrubs like Miconia spp., Gaultheria spp. and Rubus boliviensis (Lippok et al., 2013; Beck et al., 2024). Bracken covers more than 70% of the vegetation, reaching heights of 150–300 cm (Gallegos et al., 2015; Beck et al., 2024), whit its litter covering approximately 70% of the soil, with a depth between 10–50 cm (Lippok et al., 2013; López et al., 2024).

We selected eight bracken-dominated areas at 100 m from the forest edge ( Figure 1A ) that had last burned between five and 15 years prior to the experiment and were each at least 1 km apart. The selected sites had similar vegetation structure and slope aspect. At each site, 50 × 50 m plots were established and divided into four 25 × 25 m subplots using a randomized-block design with the following treatments: 1) fronds and litter intact (F+L+); 2) fronds intact and litter removed (F+L−); 3) fronds removed and litter intact (F−L+); and 4) both fronds and litter removed (F−L−) ( Figure 1B ). To maintain the treatments, fronds were carefully cut at ground level with a machete every four months, and litter was removed manually.

The environmental conditions varied across bracken management treatments (López et al., 2024). For instance, soil temperature in treatments without fronds and litter is approximately 8°C higher than in treatments where fronds and litter are present. Similarly, soil moisture is about 15% lower after bracken removal, and photosynthetically active radiation (PAR) at 20 cm above the ground is nearly 80 μmol m⁻² s⁻¹ higher in treatments without fronds compared to those with bracken cover.

Based on fruit availability in the study area, 24 common tree species at forest edges and in the forest interior were selected. To enhance response diversity in the experiment, we included species from different families and seed sizes (Palma and Laurance, 2015). (see Supplementary Table S1 ) for details of species, seed size and number of seeds sown). Seeds were collected from at least ten individuals per species and mixed. All seeds were manually cleaned and counted before sowing. Orthodox seeds were collected once for sowing at all sites, stored in paper bags, and sown within one month. Recalcitrant seeds were collected shortly before sowing at each of the eight sites, and stored in small flasks with damp paper to maintain moisture until sowing within the next two days. In each subplot, seven rows were cautiously delineated to avoid bracken disturbance ( Figure 1B ). Four rows were designated for the seed addition experiment. In each row, eight 1 × 1 m² quadrats were established and further divided into four 50 × 50 cm² sub-quadrats, leaving the lower right sub-quadrant free to facilitate the evaluation of the other three sub-quadrats. These quadrats were separated by 2 m vertically and 3 m horizontally, with a pathway between rows. Because seed germination and survival are positively related to seed size, and small-seeded species tend to produce more seeds (Westoby et al., 1996; Moles and Westoby, 2004), in each sub-quadrant, a different species was sown at different quantities, depending on their seed size and availability, approximately 50 seeds for small-, 15 for medium- and 3 for large-seeded species ( Supplementary Table S1 ). Seeds were dropped to simulate natural dispersion. A total of 46,640 seeds were sown, of which 32,320 (69.3%) were small seeds, 13,376 (28.7%) were medium-sized seeds, and 944 (2%) were large seeds.

Simultaneously, seeds from the same cohort were sown at the greenhouse from the Santiago de Chirca Biological Station (16°23’50.23” S, 67°34’53.53” W, 2080 m asl), with a 50% shade mesh and regular watering, and grown in separated bags. One year later, nursery-raised seedlings were transplanted to the plots in three rows adjacent to the rows with the seeds added ( Figure 1B ). Both experiments were implemented in the rainy season (November 2019–February 2020) to promote germination and survival (Paterno et al., 2016). Due to the high asynchrony of fruit production in the tropics (Usinowicz et al., 2012) and the lower germination and survival rates of some species (Moles and Westoby, 2004), it was not possible to obtain seedlings from all 24 species to transplant. Therefore, only 14 species were used for the transplant experiment ( Supplementary Table S1 ).

For the seed addition experiment, we assessed seedling establishment and recruitment success. Seedling establishment was calculated as the proportion of sown seeds that developed cotyledons or leaves. Recruitment success was calculated as the proportion of the established seedlings that survived to the final assessment (i.e., after 36 months) (Paterno et al., 2016; Barczyk et al., 2024). Measurements were taken at 6, 12, 24 and 36 months, with a focus on the cumulative outcome at 36 months. For transplanted seedlings, the first evaluation of height took place immediately after planting, with subsequent evaluations of seedling growth after 24 and 36 months. Growth was calculated as the difference between the final and initial height of the transplanted seedlings (Paterno et al., 2016).

To assess the changes in early ontogenetic stages of seedlings across the four bracken management treatments according to seed size, we first conducted separate generalized linear mixed-effects models (GLMMs) for each treatment, treating seed size as a continuous variable. Seedling establishment and seedling recruitment were analyzed using a beta-binomial error distribution to account for over-dispersion (Crawley, 2012), while growth was modeled with a Gaussian distribution. Secondly, we classified species by seed-size categories as follows: small-seeded species (n=8, seed size <2 mm), medium-seeded species (n=13, seed size 2–10 mm), and large-seeded species (n=3, seed size >10 mm). We then performed three separate GLMMs using the same response variables, with bracken treatments, seed-size category, and their interaction as fixed effects, and site as a random effect. Additionally, we conducted separate models for each of the 24 species, using only treatment as a fixed effect and site as a random effect.

To assess the intensity and direction of the interaction between bracken fronds and litter, we used the Relative Interaction Intensity Index (RII), which quantifies the relative effect of the species interactions, ranging from −1 (maximum competition) to +1 (maximum facilitation), while values near to 0 indicate neutral interactions (Armas et al., 2004). The RII was calculated based on the performance of the planted seedlings in the different bracken treatments, using the formula:

where B_W represents the performance of each ontogenetic stage and seed size category in the presence of bracken fronds, and B₀ represent their performance in the absence of bracken fronds. The same formula was used to assess the effect of litter. We pooled the data from treatments with fronds and litter, respectively. The differences in RII indices for seedling establishment, recruiting success and growth for each seed size category and treatment were tested using generalized linear mixed-effect models (GLMM). In all models we included treatment (fronds and litter), seed size category and their interaction as fixed effects, with sites and species as random effects, and a Gaussian error distribution.

All models were fitted using the ‘glmmTMB’ package (Brooks et al., 2017), and each model was validated analyzing residuals in ‘DHARMa’ package (Hartig, 2022). Post hoc Tukey comparisons among treatments were conducted with the ‘emmeans’ package and all figures were plotted using the ‘ggplot2’ package. Marginal R2 values, representing the proportion of variance explained by the fixed factors were calculated using the r.squaredGLMM function of the ‘MuMIn’ package (Barton, 2018). The significance of the fixed factors in each model was assessed using the ‘mixlm’ package (Liland, 2019). All statistical analyses were carried out using R, version 4.2.1 (R Core Team, 2022).

We recorded a total of 1,995 established seedlings (4.3%). Seedling establishment and recruitment success significantly and positive increased with seed size in all treatments except in the treatment without fronds and without litter, were both demographic variables remained low ( Supplementary Figure S2A, B ; Supplementary Table S2 ). Seedling growth tended to reduce with seed size in all treatments, but the differences were not significant ( Supplementary Figure S2A, B ; Supplementary Table S2 ).

Seedling establishment was significantly influenced by treatment (X ² = 158.6, p<0.001), seed size category (X ² = 170.9, p<0.001) and their interaction (X ² = 21.4, p<0.01). Establishment was the lowest in the absence of fronds and litter (F−L−, p<0.001), without an effect of seed size. In contrast, the highest seedling establishment was found in the treatment with fronds and litter for large-seeded species ( Figure 2A ; Table 1 ). A similar trend was observed in the treatment with fronds and without litter (F+L−) and in the treatment without fronds with litter (F−L+, Figure 2A ).

Recruitment success was significantly influenced by treatment (X ² = 228.4, p<0.001), seed size category (X ² = 19.6, p<0.001) and their interaction (X ² = 15.1, p<0.05). Recruitment success in both treatments without fronds (F−L+ and F−L−) was significantly lower (p<0.001) than in treatments with fronds (F+L+ and F+L−) for all seed size categories ( Figure 2B ; Table 1 ).

From the 1,285 transplanted seedlings, 948 (73.7%) survived the first 36 months. Growth was higher in the treatments with fronds (F+L+ and F+L−) than in the treatments without fronds (F−L+ and F−L−, Figure 2C ), and this pattern did not change according to seed size ( Figure 2C ; Table 1 ; Supplementary Figure S3 ). Species specific trends can be visualized in Supplementary Figures and Tables ( Supplementary Tables S3 – S5 ).

For all three early recruitment processes and seed-size categories, the relative interaction intensity index (RII) indicated different patterns in both the intensity and direction of interactions. The intensity of facilitative effects was particularly strong for fronds, as the RII values were consistently positive for seedling establishment, recruitment success and growth, for all seed-size categories, indicating beneficial effects ( Figure 3 ).

On the other hand, the intensity of negative interactions in response to bracken litter was only pronounced for the seedling establishment of small-seeded species, indicating competition, while recruitment and growth showed neutral effects for small-seeded species. In contrast, medium- and large-seeded species showed a strong positive interaction with bracken litter in seedling establishment, indicating facilitative effects. Large-seeded species maintained a positive interaction with litter during recruitment, whereas the interaction was neutral during growth ( Figure 3B, C ; Table 2 ).

As expected, bracken fronds and litter influenced the early development of tree species based on seed size. Seed size was positively related to seedling establishment in all treatments except without fronds and without litter. This pattern is consistent with findings where plants with small seeds have lower probabilities of seedling establishment compared to those with large seeds (Leishman et al., 2000; Wang et al., 2022) in shaded conditions, typical of bracken-dominated areas. This occurs because light requirements for germination and establishment decrease as seed size increases (Milberg et al., 2000). Furthermore, large seeds can produce robust seedlings with thicker stems and larger cotyledons, which not only provide nutrients until the first true leaves develop but also, in some species, contribute to photosynthesis, providing an additional energy source that is advantageous in shaded environments (Kitajima, 2003).

In our seed addition experiment we observed significantly higher recruitment of small-seeded species in treatments with fronds and litter compared to those where fronds and litter had been removed. This pattern might become even more evident with an increase in sample size. For instance, Soliz et al. (in preparation) found similar results using the same species, comparing seedling emergence under natural conditions and in a nursery setting. This result can be explained by the higher temperatures and lower humidity in the absence of fronds and litter (Gallegos et al., 2015; López et al., 2024), conditions that may be too harsh for seedling survival. This is consistent with previous findings showing that fronds ameliorate the harsh conditions of open areas by reducing soil temperature and photosynthetically active radiation, thus improving seedling establishment and recruitment (Gallegos et al., 2015; López et al., 2024).

Bracken-dominated areas are characterized by low species diversity (Günter et al., 2007; Ssali et al., 2017; López et al., 2024). However, in addition to Pteridium, there is a dominance of some pioneer and small-seeded shrub species, particularly from the Asteraceae, Ericaceae and Melastomataceae families (Lippok et al., 2013). Our results suggest that this lower diversity could be due to the fact that only a few small-seeded species can thrive in these ecosystems, and those that could succeed do not arrive, mainly due to the lack of seed dispersers (Saavedra et al., 2015). Among the small-seeded species in our study, Miconia hygrophila (Melastomataceae) exhibited high seedling establishment and recruitment. These results show the variability within small-seeded species, suggesting that other functional traits such as cotyledon type, bark thickness, leaf area, specific leaf area, and wood density, may also play an important role in determining a species ability to thrive in these challenging environments (Green and Juniper, 2004; Moles and Westoby, 2006; Baraloto and Forget, 2007; Poorter et al., 2008), and should be investigated in future studies.

Our study showed that all seed-size categories benefited from the presence of bracken fronds in the early stages of tree development. All the species, showed improved seedling development under bracken fronds, which highlights the role of bracken as a facilitative species rather than a competitor for resources. These results aligns with recent research has found that ferns promote community assembly by increasing soil stability and other soil properties and reducing competitive pressures (Azevedo-Schmidt et al., 2024).

The competitive effects of bracken through litter on small-seeded species are consistent with previous findings suggesting that small-seeded species are particularly sensitive to litter barriers (Moles and Westoby, 2004; Jessen et al., 2023). This sensitivity may be related to their limited reserves, which make it difficult for them to penetrate or tolerate litter layers (Aud and Ferraz, 2012; Wang et al., 2022). During the establishment phase, bracken litter particularly disadvantages small-seeded species, which often grow rapidly and require a lot of light (Pereira De Souza and Válio, 2001). In addition to the shade provided by the bracken fronds, the litter creates an even shadier environment that is difficult for small-seeded species to overcome (Molofsky et al., 1992; Loydi et al., 2012). Moreover, seeds can become trapped in the litter layer, preventing them from coming into contact with the soil and reducing their probabilities of germination (Donath and Eckstein, 2012).

Although litter showed a facilitative effect for medium- and large-seeded species during seedling establishment, these effects tend to became neutral over time. This ontogenetic shift is consistent with previous research showing that the intensity and direction of interactions could change during ontogeny (Paterno et al., 2016). These findings emphasize the need for long-term studies to better understand the dynamics of facilitation and competition in ecological restoration efforts.