Presence points of C. elatior in Mexico were taken from the compiled and revised database of herbarium material from Villar-Morales (2020). Central American distribution in Guatemala and Honduras was obtained from the Global Biodiversity Information Facility (GBIF 2024). The total distribution map was visualized in R v.4.3.1 (R Core Team 2024).

We sampled six populations of C. elatior in five localities from Mexico and Guatemala corresponding to both solitary and cespitose morphotypes (Fig. 1; Suppl. material 1). A single locality, El Mirador, Veracruz, Mexico had both morphotypes growing in syntopy, while in the others only one of them was found. Morphotypes were recognized and assigned on the field by their habit, either solitary or cespitose with above-ground branching. We collected leaves from different individuals, with an average of 10 plants sampled per population. Leaves of C. elatior show significant morphological changes as the plant grows and acquires its scandent habit as an adult (Standley and Steyermark 1958). The typical, reflexed apical leaflets are seen in mature individuals with fully pinnate leaves, and as such only these leaves were sampled to avoid variation regarding the age of the individuals and leaf maturity as much as possible.

To avoid confusion, we have used the term “morphotype” in contrast to “form” used by previous authors (Hodel 2013; Dowe and Hodel 2021). We consider this term to reflect better the framework of this study since sampling and evaluations were based on morphological, specifically habit, differences. Our cespitose morphotype corresponds to the previously mentioned cespitose form and our solitary morphotype would compare to the first solitary form of Hodel (2013). Although we were unable to visit the locality of the remaining solitary form described by Hodel (2013) from the isthmus of Tehuantepec, we consider it to be a rare, local variant of our solitary morphotype for a number of reasons that will be discussed afterwards. Field exploration in the area where this form inhabits has been made problematic due to recent fires and an uprise in violence.

We could not consistently find reproductive structures, either inflorescences or infructescences, during the sampling field trips. Most solitary plants had already lost their male inflorescences or were in advanced state of decomposition to properly measure them, while only a few individuals in some populations had immature fruits. Similarly, we did not observe measurable flowering and fruiting structures in the cespitose populations, except for very few fruiting plants in La Esperanza. As a result, due to the small and inconsistent sample size comparative statistical analyses could not be performed with reproductive characters.

To evaluate the morphological variation among populations and morphotypes, we measured 13 morphological characters from the leaves, most of which have been used in previous studies (Atria et al. 2017; Santos-Hernández et al. 2022): petiole length (PL), petiole width (PW), rachis length (RL), width of the rachis at its middle length (mRW), width of the rachis between the most distal leaflets (aRW), leaflet number on one side of the rachis (LN), basal leaflet length (bLL), basal leaflet width (bLW), basal leaflet insertion length (bLI), median leaflet length (mLL), median leaflet width (mLW), median leaflet insertion length (mLI), and distance between median leaflets (mLD). All variables were log-10 transformed before the analyses. All statistical analyses were performed in R v.4.3.1 (R Core Team 2024) using RStudio v.2023.12.1 (RStudio Team 2023).

Univariate methods were used to study the variation between both morphotypes. Student’s t-test was performed for all characters considering both morphotypes using the stats base package in R. For a multivariate approach using Q-type analyses, we first explored the structure of the data and summarized its variation by applying a non-metric multidimensional Scaling (NMDS) on a distance matrix obtained with the Euclidean measure. Clustering of the data was also analysed via a non-hierarchical k-means cluster analysis which does not form nested clusters iteratively as in hierarchical clustering, instead finding all clusters simultaneously as partitions of the data (Jain 2010). To assess the optimal number of clusters, we used the Calinski-Harabasz index (Caliński and Harabasz 1974). Both NMDS and k-means cluster analyses were done using the R package vegan v.2.6-8 (Oksanen et al. 2024).

Under a R-type multivariate approach, we first examined all variable’s contribution to total variance in a simplified manner with a Principal Component Analysis (PCA). Components with eigenvalues greater than 1.0 were extracted and plotted in a scatterplot. A priori grouping (solitary vs cespitose morphotypes) was first evaluated with a MANOVA to assess differences in group centroids. We calculated the partial eta-squared measure (η²) to quantify the effect of the morphotype on the variance of the leaf characters, using the R package effectsize v.1.0.0 (Ben-Shachar et al. 2024).

We then used a Linear Discriminant Analysis (LDA) to further test differences in the morphotype grouping and find the variables with the higher discrimination power. Our a priori classification of the individuals was evaluated using the confusion matrices produced by this analysis, testing its accuracy without cross-validation and with cross-validation via two methods: jack-knifing and Monte Carlo (1,000 repeats using 40% of the data as test). LDA and jack-knife cross-validation was done with the R package MASS v.7.3-60 (Ripley et al. 2024), while Monte Carlo cross-validation was done with the R package caret v.7.0-1 (Khun et al. 2024). Additionally, we repeated the discriminant analysis with population grouping as to compare its results with the morphotype grouping. The same evaluation of the accuracy via confusion matrices was done as in the previous grouping.

All but two of the analysed morphological characters of leaves showed significant differences between the solitary and cespitose morphotypes of C. elatior, and both morphotypes were significantly distinct when all variables were compared in the MANOVA analysis. In the similarity and ordination analyses, the morphotypes could be distinguished separately with only minor overlap between them. Specimens were classified to their previously assigned morphotype with high accuracy in the discriminant analyses. If populations were considered for the discriminant analyses instead of morphotypes, the classification still separated the latter with high accuracy. The variables that most distinguish between morphotypes are related to the overall size and robustness of the leaves (Table 5), with the solitary form’s being larger with bigger leaflets and longer insertions to the rachis. The cespitose morphotype, though smaller in leaves and leaflets, has longer petioles.

Certain morphological overlap was observed, mainly between some cespitose Totontepec and solitary specimens, both in the similarity analyses (NMDS and k-means clustering) and in the PCA. According to our raw data and the PCA coefficients, leaves of Totontepec specimens were the most robust and with the largest leaflets within the cespitose populations, but not so as in the solitary populations. This would explain their similarity with solitary specimens and their resulting position and clustering in the Q-type analyses. The discriminant analyses, however, group this population clearly with the other cespitose ones.

The most striking character used to recognize morphotypes in the field, the branching of the cespitose morphotype, was not considered in the analyses because of its lack of variance and evident segregation power. Mature individuals of the cespitose morphotype can produce branches well above ground, one on every other node, forming a dense clump above the ground and nearby smaller vegetation. No other Chamaedorea species are known to produce proper aerial branches, other cespitose (clustering) species form new stems (branches) from buds on proximal nodes below or just above the ground (Fisher 1974; Hodel 1992). Nevertheless, much work is still needed in characterizing the branching of many palm species.

The only other species in the genus known to also present a solitary and cespitose habit is C. tepejilote; however, the variation in this species has been associated to its cultivation history, the solitary habit a result of domestication (Castillo Mont et al. 2017). Although C. elatior is used in some areas in Mexico for handcrafting or as ornamental, its extraction is irregular, and no serious domestication effort has been recognized (Contreras Cortés et al. 2018; Rendón-Aguilar et al. 2022). Considering this, the presence of two habits in C. elatior does not seem to be a consequence of cultivation nor related to human activity. In view of the uniqueness of its branching habit, and pending proper development studies, this difference should be considered an important character that could enhance species delimitation of both morphotypes.

Chamaedorea elatior’s morphotypes are also segregated geographically and possibly ecologically (Table 1). According to previous knowledge and herbaria records, the solitary morphotype inhabits lowland tropical humid forests usually below 800 m a.s.l., with some populations extending up to about 1,200 m. In contrast, the cespitose morphotype is found in tropical montane cloud forests from 1,000 to 1,900 m a.s.l. (Hodel 2013; Villar-Morales 2020). In certain localities in Oaxaca and Veracruz, Mexico, the solitary morphotype can be found in tropical forests on the foot of the mountains and lowlands extending beyond, while the cespitose morphotype has been registered a few hundred meters uphill in montane forests on the slopes of the mountains. Though geographically very near, the difference in altitude and thus vegetation is notorious.

El Mirador is the only locality to our knowledge where both morphotypes have been found living in syntopy. Their coexistence in this site has been registered historically, as we located specimens collected by Liebmann in the 1840s (at C herbarium) corresponding to both morphotypes in the same area. This locality is geographically near the species’s northern limit, and only very few populations of both morphotypes are known to exist beyond it. Furthermore, El Mirador is on the upper boundaries of altitudinal range of the solitary morphotype and on the lower range of the cespitose one. This could explain their presence in this area, probably being a contact zone where both morphotypes can inhabit given the proper altitudinal range for each of them. The vegetation type there has been registered in diverse collected specimens as montane cloud forest (or synonyms sensu Gual-Díaz and González-Medrano 2014) though it is at a lower altitude and less dense and humid as the others sampled here in La Esperanza and Totontepec. Although montane cloud forests are variable in environmental characteristics and composition (Gual-Díaz and González-Medrano 2014), a proper compositional and ecological study of the site could lead to a better understanding of its uniqueness regarding this syntopy.

Chamaedorea species are known to occur sympatrically in many areas (Hodel 1992). However, very few research has been done on how species limits are maintained in these instances. Differences in pollinators and flowering time have been proposed in other species in the genus (Luna et al. 2005; Bacon and Bailey 2006), as well as in other understory palms (Knudsen 1999; Borchsenius 2002; Borchsenius et al. 2016). Different reproductive timing could also be keeping the morphotypes distinct in El Mirador, since immature fruits were only found in some solitary plants. However, records of the cespitose morphotype both in El Mirador and other sites are very limited and incomplete to properly assess its phenology and compare it with its solitary counterpart. Ecological studies pending, our results suggest that the variation found in the leaf characters has little correlation with environmental factors and could be evidence of underlying genetic divergence. This could perhaps be kept through certain means of reproductive isolation that can only be seen in this locality due to their cohabitation.

As stated previously, we have considered only two morphotypes in this study based on their habit differences, a solitary and a cespitose one. Hodel (2013) described a third form that resembles the solitary morphotype but differs in size and developmental characteristics. We believe this to be a rare variant of the solitary morphotype, rather than a completely different form, based on our current knowledge of C. elatior and the following reasons. First, bifid juvenile leaves of similar size were seen in other localities visited in Veracruz, Mexico. Second, the more robust size is only noted in its bifid leaves with no reference to a distinction in its pinnate leaves (Hodel 2013). Third, both its habitat and altitudinal range fall in the range of our solitary morphotype. Fourth, we have seen herbarium material of C. elatior from Uxpanapa, Veracruz and Los Chimalapas, Oaxaca that border the reported site and all are quite similar in appearance. Though we did not include herbarium material in this study due to its incompleteness and difficulty in assessing homology between leaflets, no material could be accurately referred to this form since all specimens were pinnate.

The central difference of this form is that it retains its bifid, otherwise juvenile, leaves for a certain period of time after attaining maturity (i.e. flowering) but will eventually produce its typical pinnate leaves (Hodel 2013; Dowe and Hodel 2021). This distinction is thus best interpreted as developmental rather than morphological. The difference between juvenile and mature leaves in palms has been described by Tomlinson (1960), but no studies have been done to understand the developmental processes responsible for it. In only one experiment in the palm Caryota mitis Lour., juvenile leaves were induced in a mature individual by administering gibberellin (Fisher 1976). If such a change is mediated by gibberellin or similar hormones in Chamaedorea is unknown. Although rare, other species in the genus can also flower before their leaves attain their maximum number of leaflets or even develop pinnate leaves after years of flowering and bifid leaf production (Hodel 1992).

No research has been done on the habit of scandent species of Chamaedorea. Still, field observations on C. elatior have shown the solitary morphotype to have a more vigorous climbing habit than the cespitose one, as it can reach higher heights in the canopy and has stronger supports. The latter was noticed during this study, since pulling on stems from below to reach the leaves proved harder in the solitary form. The cespitose morphotype, on the other hand, has more of a scrambling habit, covering a wider area and entangling with more plants but not reaching as high. The solitary morphotype’s larger leaves might have an adaptive value and could be related to its habit, as shown in other climbing palms. When comparing leaf morphology between Desmoncus orthacantos Mart. (a strong climber) and D. polyacanthos Mart. (a weaker climber), Isnard et al. (2005) mention that larger leaves provide D. orthacantos with a larger range to encounter support points in the vegetation and can ensure its attachment to wider support points (i.e. branches). Although C. elatior’s climbing habit is different in various ways from that of species in Desmoncus Mart., considering the first lacks more specialized structures as spines or cirri, certainly trends in leaf size appear to be similar.

With no spines nor proper acanthophylls in a cirrus, climbing habit in C. elatior seems to be facilitated by its reflexed leaflets, especially distal ones, and their hardened, callose bases acting as support (Acevedo-Rodríguez 2020). During sampling, we noticed that leaves would lose some leaflets, at the insertion to the rachis, when pulled down with force from under the canopy. This suggests that the main point of support is indeed the hardened base of the leaflets. The length of the leaflet insertion to the rachis was seen in situ to be proportional to the size of this callose base, longer insertions had larger bases. This could amount to the previous hypothesis in that the solitary morphotype, as a stronger climber than the cespitose one, would have a more sizeable support system based on its significantly longer leaflet insertions.

Traditional morphometrics have been shown to be a reliable tool in delimiting taxa in palm species (Henderson 2006; Laubengayer et al. 2012; Santos-Hernández et al. 2022). Our analyses on the morphological variation of the leaf show that the solitary and cespitose morphotypes of C. elatior are significantly distinct from each other. Though not properly included in our analyses for reasons mentioned previously, the branching habit of the cespitose morphotype is unique and is often used as the main difference to segregate them (Hodel 2013).

Though variation in reproductive structures could not be properly studied between morphotypes, we consider that the observed differences in vegetative characters are indicative of probable evolutionary distinctiveness between them, especially considering their cohabitation in the same habitat while maintaining their unique morphology (in El Mirador). Patterns observed when analysing vegetative characters were comparable when including reproductive characters in Geonoma Willd. (Borchsenius 1999) and Gaussia H.Wendl. (Santos-Hernández et al. 2022). Differences in leaf morphology were likewise confirmed by genetic divergence between similar species Chamaedorea glaucifolia and C. plumosa (Ruíz-Castillejos 2011). A targeted comparative exploration of reproductive structures, phenology, and other lines of evidence would add to our growing knowledge of variation in the morphotypes of C. elatior and permit a more satisfactory morphological delimitation.

Reproductive morphology has been used in Chamaedorea mainly to delimit artificial subgenera based on floral characters (Hodel 1992, 1999); however, these characters seem to not diverge substantially between closely related species. Similarly, floral variation has been proposed to be useful in delimiting or recognizing certain monophyletic groups within the genus (Askgaard et al. 2008), but no research has been done in intraspecific variation within these groups. In the segregation of C. tacanensis from C. elatior, only certain quantitative differences were observed in characters pertaining to the inflorescences’ axes (Pérez-Farrera et al. 2021). The main distinction was in the longer peduncles of C. tacanensis, but all other inflorescence characters overlapped to a certain degree (Pérez-Farrera et al. 2021). According to initial and incomplete observations during field work, we expect a similar pattern to occur among the morphotypes studied, mostly on size and number of the axes and probably not in flower or fruit morphology.

Based on the current evidence, the recognition of these morphotypes as separate species seems valid; however, we consider that more lines of evidence are desirable to properly delimit these forms within C. elatior with a more robust, integrative approach (Dayrat 2005; Padial et al. 2010). Preliminary climatic niche divergence and genetic results on nuclear loci of the same populations sampled in this study have shown noteworthy variation that can deepen our understanding of C. elatior and corroborate the findings presented here.