More than 50 years ago two ecologists, Janzen (1970) and Connell (1971) in Central America and in Australia respectively, independently proposed the hypothesis that seeds and seedlings suffered a distance- and density-dependent mortality. Empirical observations reported that seed-fall obviously concentrated under the parent fruiting trees, whereas seedlings and saplings recruitment did not match the seed-fall kernel, unexpectedly peaking at a certain distance from the source tree. Their descriptive model consisted of two curves: the first showing seeds dispersal around a mother tree and the second reporting the survival probability of the seedlings as a function of distance from the same tree. Accordingly, a species was found unable to recruit under adult conspecifics because of the formation of an “exclusion area” where the mortality was disproportionately higher and not matching with the large seed availability (Janzen, 1970). Such phenomenon, preventing the dominance of a single species, was also recognized to allow species coexistence and to promote the maintenance of biodiversity in the ecosystems (Levi et al., 2019). In other words, field observations showed that a mortality agent operated in a distance, density-dependent, and species-specific way, impairing only the recruitment of the focal tree, with no effects on seeds and seedlings of other species.

Early reports of the Janzen-Connell (JC) recruitment distribution suggested that insects were killing all seeds falling under the canopy of the studied trees (Janzen, 1971). However, most later studies assessed only the occurrence of some indirect insect damages, missing the identification of a particular causative pest species (Basset et al., 2019). Other studies blamed the activity of vertebrates and mammals, but in most cases the activity of such animals cannot be related to the distance-dependent effects, considering their mobility over the forest floor (Song et al., 2021). In any case, predation is still considered as the main causal mechanism of the JC effect also in theoretical studies (Smith, 2022).

The role of soilborne and airborne pathogens has been also considered based on the assumption that pathogens propagules, including spores and sclerotia, may accumulate under the focal tree. For instance, Packer and Clay (2000) reported that the oomycete Pythium spp. disproportionately killed the seedlings rooted under the canopy of conspecific trees. Then, plant pathogens have been considered as main key factors in maintaining coexistence by causing local plant-soil Negative Feedback (NF), especially in wet ecosystems where oomycetes and fungi thrive (Agrios, 2005). However, most of the available experimental and field studies only reported pathogens damage, e.g., seeds covered by mold, seedlings wilting, leaf discoloration and spots, often missing the identification of the causal pathogenic agent and always without any assessment of consistency between the spatial distribution of the NF occurrence and the pathogens dispersal behavior.

Less popular is the hypothesis of the “exclusion zone” being generated by the accumulation of autotoxic chemical compounds. The study by Webb et al. (1967) was one of the first reporting that some unidentified autotoxic compound, killing seedlings of Grevillea robusta, could be the cause of the JC patterns reported in Australian forests. Autotoxicity has been reported for hundreds of plant species, mostly crops, and associated with the release of phytotoxic compounds during the decay of leaf and root litter, including a range of different compounds such as aromatic phenols, saponins, coumarins, and organic acids, among others (Singh et al., 1999). However, the autotoxicity hypothesis has suffered from a relevant objective weakness because all the above-mentioned compounds produce a generic phytotoxicity, thus unable to explain the species-specificity of the JC distributions. Moreover, no studies were able to identify and quantify the accumulation of any specific toxin under and around the focal tree. Despite these major problems, the autotoxicity hypothesis has continued to be inconsistently associated to generic allelopathic phytotoxic effects (Inderjit et al., 2021). Differently, autotoxicity found a logical explanation in the discovery that accumulation of fragmented extracellular DNA from decomposing leaf litter did cause extensive seed germination impairment and root damages to a broad range of higher plants, with toxic effects specifically limited to conspecifics (Mazzoleni et al., 2015). The evidence of such inhibitory effect by extracellular self-DNA was then recently confirmed and further investigated by means of whole-plant transcriptome and metabolomic profiling on the model plant Arabidopsis thaliana (Chiusano et al., 2021; Lanzotti et al., 2022).

Irrespectively of the underlying mechanisms, several studies explored the consequences of the JC hypothesis using simulation models. Hubbell (1980) argued that disproportionately high seed densities under the parent tree would overcome the lower survival, thus resulting in monotonic recruitment patterns independent of distance from the focal plant. Later, Nathan and Casagrandi (2004) made a first systematic modelling exploration of the JC hypothesis. By using a mathematical model of distance and density-dependent seed mortality, the authors demonstrated that the net balance between seed dispersal and recruitment survival could generate all observed recruitment patterns, including both the hump-shaped typical JC distribution and monotonically decreasing (Hubbell) patterns ( Figure 1 ). Later, Vincenot et al. (2017) presented an individual based model, at the scale of one focal plant, reporting that strong NF under a conspecific tree may overtake the seed dispersal kernel, thus creating an “exclusion area”. Moreover, the study also demonstrated that NF could produce an outward shift of the recruitment peak from seedlings to saplings, during a longer assessment of the recruitment process. More recently, Levi et al. (2019) changed the perspective from a single focal plant to ecosystem scale and, using high-performance computing and analytical models, demonstrated that distance-responsive natural enemies can maintain tropical forest diversity nearly indefinitely by favouring rare species. Moreover, the effect of NF at ecosystem level has been modelled, clearly explaining species coexistence (Bonanomi et al., 2005), its consistent relationship with biodiversity levels in different ecosystems in association with the rates of litter decomposition producing autotoxicity (Mazzoleni et al., 2010) and the formation of vegetation spatial patterns (Iuorio and Veerman, 2021)

Besides these robust modelling studies, the importance of the JC recruitment pattern is largely supported by many publications of empirical data from a broad range of ecosystems, including tropical (Mangan et al., 2010; Comita et al., 2014) and temperate forests (Fox, 1977; Packer and Clay, 2000), as well as shrublands (Bonanomi et al., 2008; Teste et al., 2017; Idbella et al., 2022) and grasslands (Petermann et al., 2008). However, despite such strong base, scepticism still persists especially in the community of forest ecology (Terborgh, 2020). The main doubts about the actual relevance of the JC hypothesis are caused by the variable results observed in multispecies studies reporting JC distributions for some species but Hubbell patterns for others, also named as reverse JC, and interpreted as a positive distance-dependent process due to accumulation of symbiotic microbiota (Zahra et al., 2021). Although it cannot be denied that JC patterns do exist in many species, their absence in other coexisting species raised a strong debate on the effective generality and magnitude of JC effects (Song et al., 2021). Explaining the reasons of such variability of occurrence of JC distributions would shed light on the relevance of this ecological phenomenon and on its effect on species coexistence and diversity maintenance.

In this regard, a relevant issue is whether the distance-dependent mortality factors, preventing the recruitment near the parent tree, are affected by the surrounding density of neighboring conspecific adults. This point has been mostly neglected by previous studies that focused only on the focal tree concept, overlooking the possible role of the surrounding landscape of both conspecific and heterospecific trees. Only the recently published work by Smith (2022) recognized that conspecific density may affect JC effects on species coexistence relating this to putative changes of predation levels. Theoretically, the density of conspecific adults may affect the behavior of invertebrate and vertebrate predators (Janzen, 1971), the spread of airborne and soilborne pathogens (Sapoukhina et al., 2010), as well as leaf and root litter distribution and the associated self-inhibitory factors produced during decomposition (Bonanomi et al., 2017). For example, an isolated tree accumulates litter under its canopy, thus creating a pattern associated with the concept of “island of fertility” when interpreted in terms of positive soil conditions for plant growth (Facelli and Brock, 2000), but also generating a round shaped exclusion zone around its trunk and within its own crown projection by NF. However, as the surrounding density of adult trees of the same species increases, the spatial distribution of litter progressively overlaps among individuals, generating a complex patchiness in terms of exclusion zones created by the compenetrating conspecific “litter islands”.

The aim of this work is to explore the connection between adult density, either conspecifics or heterospecifics, on the probability of occurrence of JC distributions. In detail, using an Individual-Based Modeling (IBM) approach (DeAngelis and Grimm, 2014), we simulated the formation of exclusion zones due to the build-up of NF in proximity of conspecific adult plants. The specific hypotheses of our study were:

Assuming that NF is species-specific, localized under an individual tree, and with limited horizontal diffusion, our simulations show that tree density is critical to understand the observed variability of tree recruitment patterns. Our model highlights the complex interconnection between NF intensity, stand density, and recruitment patterns explaining where and why the JC distribution occurs, and clarifying the relevance of this ecological phenomenon in different plant community frameworks.

Our initial hypothesis that JC distribution is very common in the case of an isolated tree was partially supported by the model simulations. Indeed, we found that JC distribution was very frequent for isolated trees when NF was strong and capable to form an exclusion zone under the parent tree. However, with decreasing NF intensity, both JC and decreasing patterns co-occurred and were recorded with similar frequencies. A prevalence of the decreasing pattern was also observed at very low NF, because under such conditions the inhibitory effect due to NF was unable to overcome the clustering effect of the seed dispersal kernel, with resulting concentrated recruitment under the parent trees. JC distribution in isolated individuals has been previously reported for both shrubs and trees, but in far less cases compared to tropical and temperate forests (reviewed in Bonanomi et al., 2010; Comita et al., 2014; Song et al., 2021). A notable example is the study of Clark & Clark (1981), reporting a clear distance-dependent recruitment limitation for isolated trees of Bursera graveolens in arid ecosystems with discontinuous vegetation. Moreover, a recruitment distribution consistent with the JC model has been reported for woody plants belonging to Fabaceae, a plant family forming fertility islands under individual canopies, associated to the accumulation of organic matter, nitrogen and phosphorus (Facelli & Brock, 2000). Under this scenario, in order to observe a distance-dependent inhibition, the generating factors of NF, attributed either to soilborne pathogens or soil autotoxicity, must overwhelm the positive effects of both nutrients and beneficial microbes of the fertility islands under the canopy of woody plants. In this context, it is well established that plants belonging to Fabaceae suffer greatly from NF in both agroecosystems and natural plant communities (Cesarano et al., 2017). Accordingly, several studies reported the presence of intense NF and JC recruitment distributions in woody perennial plants, including Medicago marina in Mediterranean sand dunes (Bonanomi et al., 2008), Medicago sativa in US old field (Jennings & Nelson, 2002), Genista aetnensis over volcanic lavas (Stinca et al., 2015), and several Acacia species in South Africa (Ben-Shahar, 1991).

Regarding the second hypothesis, i.e., the decreasing occurrence probability of JC distribution with high density of conspecifics, our model demonstrated a complex scenario dependent on the intensity of the NF. Indeed, when the NF intensity is low, the JC frequency increases linearly with the density of adult conspecifics. However, if NF is strong, the peaks in JC frequency are still observed at intermediate stand densities while suddenly decrease in stands with a continuous conspecific cover of 200 trees in the plot. So, counterintuitively, our model shows that a plant suffering from strong NF in monospecific stands can rarely exhibit a recruitment pattern fitting the JC model. This seemingly paradoxical result is due to the progressive expansion of the exclusion zone surrounding all trees in the forest stand. In other words, as individual trees become more clustered and denser, their exclusion zones progressively overlap, leaving no safe place for an effective recruitment in the stand. This is consistent with the lack of JC evidence in many monospecific stands in temperate and boreal forests, including Fagus sylvatica in Mediterranean forests (Rita et al., 2021), as well as monodominant tropical forests (Hart et al., 1989; Richards, 1996). In general terms, our model demonstrates the association between strong NF and lack of distance-dependent inhibition in dense, monospecific stands. This result reconciles NF with forest composition and should reduce the scepticism of many forest ecologists towards the JC model (Terborgh, 2020).

Our third hypothesis supposed that the JC distribution should not be frequent in the case of rare species immersed in a matrix of heterospecific adults. This was largely confirmed by our numerical simulations demonstrating that a species with only 25% stand cover showed lower frequency of JC distribution compared to stands in which the species occurrence was at 50% and 75% cover. This effect was observed in both low and strong NF conditions, and reflected the fact that the abundant presence of heterospecific adult neighbors provides a suitable place for recruitment overlapping with the exclusion zones by conspecifics and thus reducing the NF effect. Rare species are indirectly protected by non-host, neighboring heterospecifics, as predicted by the herd-immunity hypothesis, which reduces the probability of contact with propagules of the virulent plant enemies (Wills et al., 1997). In the case of the autotoxicity theory, the presence of leaf litter and root debris from heterospecifics likely results in the dilution of conspecific plant residues, thus providing soil patches free of NF even near conspecific mother plants.

With respect to forest dynamics, our model simulations are consistent with robust field data on alternative species replacement reported in temperate and boreal forests around the globe, in stands co-dominated by two tree species (Fox, 1977). Notable examples include Fagus grandifolia with Acer saccharum, Picea rubens with Abies fraseri, Fagus grandifolia with Tsuga canadensis, and Picea engelmannii with Abies lasiocarpa (Whittaker & Levin, 1977; Woods, 1979; Runkle, 1981; Waters & Savill, 1992). In all these studies, the recruitment of tree species was found to be significantly more abundant and healthy under heterospecific adults. In this context, on one hand our model shows that localized NF is able to explain species replacement in forest ecosystems. On the other hand, the observation of decreasing JC recruitment patterns in co-dominated mixed forests reflects a reduced NF effect related to departure from monospecificity.

Future model simulations can focus on multispecies systems to test the effect of adult density on distance-dependent inhibition also in forest stands with high tree diversity and also assess the fate of rare species having different levels of sensitivity to NF compared to the most abundant plant in the community (Van der Putten et al., 2013). Spatial comparison of numerical simulations with real data obtained from long-term forest censuses for tropical forests such as Barro Colorado (Condit, 1998), other tropical forests (Lewis et al., 2004), and temperate forests (Král et al., 2018) could be particularly useful to this goal.

From a modelling point of view, future work could address the following points: (i) explicit representation of the germination and establishment processes separately to help disentangle the effect of different causal mechanisms due to either chemical autotoxicity or action of soil-borne pathogens. Specifically, the effect of chemical autotoxicity is reported to affect both germination and early seedlings’ growth, whereas soil-borne pathogens mostly affect the establishment phase. (ii) In order to provide a more general description of plant-soil interactions, the inclusion of facilitation by heterospecific biomass can be also explicitly considered to evaluate its relevance in the emergence of seedlings’ JC patterns. (iii) To test the effect species-specific characteristics like crown shape and seed dispersal strategies on the emergence of seedlings’ establishment patterns, different formulations of the biomass distribution and seed dispersal kernels could be implemented. (iv) Moreover, future studies could investigate the impact of the priority effect by simulating different colonization timings, as occurs in ecological succession, and the consequent effect on the recruitment of conspecifics.

Finally, a challenge for future studies will be a spatially explicit definition of the exclusion zone in forests with different tree density and diversity. New-generation sequencing techniques may enable the production of fine-scale metagenomic maps coupled with an assessment of the conspecific extracellular DNA accumulated in the soil where the NF effects are observed. Such in-depth investigations on spatial information associated to tree recruitment distribution will be relevant to support the discussion on the putative mechanisms of the JC effect and to disentangle between the hypotheses of soilborne pathogens and self-DNA inhibitory effects, thus providing a better understanding of the spatial and temporal patterns of this important phenomenon.

The data that support the findings of this study are available from the corresponding author upon reasonable request.

GB conceived the work and performed data analysis. AB implemented the model and performed numerical simulations. FC designed and prepared the figures and designed the numerical simulations. SM conceived the work and designed the model and data analysis. FG designed the model, the numerical simulations and data analysis and team coordination. All authors contributed to the article and approved the submitted version.