Specialization is a repeated pattern in living systems, suggesting that there are general mechanisms underlying its evolution. Clearly, every species interacts with only a small subset of all other species and in only a subset of habitats. Thus, some constraints on associations have simple explanations: biogeography and range limits preclude pairs of species, or a species and a particular suite of environmental variables, from being in the same vicinity. We do not currently have precise information on the extent to which range limits enforce the levels of specialization that we see in nature [1,2]. Despite this large knowledge gap, we have clear instances where species have access to many potential mutualists or habitats, yet have evolved adaptations that restrict their proportional usage [3,4]. The evolution of these adaptations may themselves directly affect the speciation process  or simply be associated with macroevolutionary and macroecological patterns (e.g. diversification through correlated characters or reduced geographical extent ). This Special feature serves to highlight our current knowledge of how evolution has produced clades and communities notable in their variation in the level of specialization, despite a paucity of well-characterized pathways that might produce such variation . We focus especially on studies that employ a macroevolutionary perspective to investigate (i) the evolutionary success of specialists and (ii) how specialization is shaped by variation along multiple morphological and environmental axes.
(a) The evolution of specialization
Is specialization a dead-end? A once-held view was that the evolution of specialization was a one-way street, with transitions back to a broader niche breadth being restricted . Recently, however, a growing number of studies of specialization indicate that transitions from specialization to generalization are possible [9,10]. However, examinations of transitions in a phylogenetic context sometimes provide conflicting results. In birds, transitions in the level of specialization are rare . Yet in plants, there are many examples of reversals in specialization ; for example, pollinator breadth exhibits little phylogenetic signal . Phytophagous insects also exhibit equivocal results with transitions occurring in either direction [14,15]. Thus, evidence is mounting that specialization is not an absolute ‘dead-end’, even though a disproportionate number of transitions may be in the direction of generalist to specialist in some systems. However, the conditions and processes that lead to biases in transition rates in one direction or the other remain poorly understood.
The envisioned pathway that would produce biased transitions from generalization to specialization involves the idea of trade-offs, which however have proved elusive to demonstrate empirically. The trade-offs might operate at the physiological level through antagonistic pleiotropy . For instance, C4 plant photosynthetic pathways have greater efficiency in hot, dry environments, but lower photosynthetic rates in shaded, moist environments . Ecology could enhance the evolution of specialization through producing further genetic trade-offs. If a certain environment were more common (e.g. dry environments in the above example), deleterious mutations that affect the performance in other environments might accumulate . If these changes represented loss-of-function mutations, reversals towards a generalist state would be rare. Whether physiological or genetic trade-offs occur in tandem or independently is not well characterized, nor is there much compelling evidence of irreversibility .
Notably, the difficulty of defining and detecting specialization hampers our ability to pinpoint its pathways and trade-offs. In the evolution of specialization of plants on pollinators, for example, steep trade-offs (where adaptions improving the attraction or use of one pollinator decrease the attraction or use of others) will generally favour specialization . Studies on plant–pollinator specialization, in particular, have provided many examples showing how caution must be exercised in inferring fitness trade-offs from morphological traits (‘phenotypic specialization’, e.g. long corolla tubes ). Classically, floral corolla tubes were one trait conceived as an appropriate indicator of reduction in biotic partnerships (narrow corolla tubes suggesting efficient pollination by hummingbirds and poor pollination by bees and vice versa for wide corolla tubes ). Many cases of apparent specialization on certain functional groups of pollinators have been observed to be, in fact, ecologically generalized (e.g. flowers with wide corollas are visited by insect and birds [22–24]).
Characterization of trade-offs is further complicated because they can occur at different scales and involve different aspects of a species' biology  (see Axes of specialization). This complexity has led to the somewhat perplexing view that trade-offs are uncommon, at least in within-species comparisons . Intriguingly, the elusive nature of trade-offs spurred investigation of whether they were an essential condition—a theoretical study by Muchhala et al.  demonstrated that the selective cost of lost pollen alone is sufficient to drive specialization even in the absence of trade-offs. To date, there have been few phylogenetic comparisons, yet analyses of host–pathogen and plant–pollinator associations suggest that the ability to incorporate a certain plant species into diet breadth is correlated with phylogenetic distance [27–29]. This suggests that specialization on clades of hosts or mutualists is widespread and that the use of certain hosts is indeed lost over evolutionary timescales. Notably, this pattern is not only consistent with trade-offs, but also consistent with ecological models of the loss of selection for interacting with hosts or mutualists that are outside of a species' range.
From a macroecological perspective, there is one additional reason to expect the appearance of dead-ends with transitions to the specialist state . If specialists occupy a narrow niche, they often also occupy smaller ranges , and endemics are more susceptible to extinction [6,30]. While some data suggest that the evolution of specialization is associated with evolutionary success in plants , habitat specialization is correlated with increased extinction risk in birds and bumblebees [32,33]. If specialists go extinct more frequently, most specialists will appear as young lineages on phylogenetic trees  that have had less opportunity to transition to a generalist state . Species occupying a smaller range can also be less likely to speciate [35,36]. Both of these processes would produce a pattern whereby there are more extant generalist lineages, each with the potential to transition to a more specialized state. Yet there are examples that show empirical support for the opposite as well, with generalists exhibiting higher extinction risk, at least in Odonata .
In this volume, we examine how ‘evolutionary success' in terms of speciation and extinction rates varies with specialization and document transitions between specialization and generalization. Specialization and generalization in the diversification of lepidopterans were examined for evidence of the musical chairs versus the oscillation hypotheses . Following a ‘musical chairs' model we might see that specialist clades were more often transitioning between hosts, but remaining specialized, whereas in the ‘oscillation’ model we would predict that niche-breadth shifts (e.g. in phytophagy in lepidopteran clades) from generalist to specialist would be more common. Rather than a pattern of unidirectional shifts to specialization in lepidopteran clades, Hardy and Otto find more support for the musical chairs hypothesis . In addition, they find a negative relationship between host-plant breadth and diversification rates, with generalists diversifying at lower rates because of their broad niches. This calls into question whether specialization can ever be considered a dead-end, at least in phytophagous insects. While transitions to a more generalist state might be rare, host switching within specialist clades is common and generates more species that are specialized, such that lowered diversification rates will not be apparent. These patterns held despite the finding that extinction rates were considerably lower in polyphagous lineages, suggesting that specialists could potentially appear as evolutionary dead-ends due to declines in persistence, but not due to trade-offs that prevent transitions back to a more generalized state.
The musical chairs hypothesis may be clade-specific, as other patterns have been seen in some plant–pollinator relationships. The shift from a specialized relationship (e.g. pollination by few resin-collecting bees) to a generalized relationship (e.g. pollination by many pollen-feeding insects) can be followed quickly by a reversal to a more specialized relationship (e.g. pollination only by ‘buzz-pollinating’ bees) . This last example is consistent with the oscillation hypothesis, which postulates that generalist lineages give rise to specialist daughter species, but over time specialists can gradually add functions and become more generalist. Similarly, in an analysis of pollinator breadth in passionflowers, Abrahamczyk et al.  find that shifts are not disproportionately from generalization to specialization. Instead, reliance on the sword-bill hummingbird (Ensifera ensifera) appears to have evolved early on in a clade that then generated many new species by allopatric isolation, some of which escaped from specialization by reducing their floral tubes, thereby being able to rely on a broader set of bird or bat pollinators. In contrast to the idea that shifts in specialization result in speciation (pollinator shifts), Abrahamczyk et al.  find more evidence favouring biogeographical shifts spurring the process of lineage splitting. In Tritoniopsis revoluta (Iridaceae) Anderson et al.  report that pollinators vary geographically across the plant's range and are closely associated with variation in floral traits, suggesting a strong role of distribution and range in how biotic specialization influences speciation (see Range extent, specialization, and diversification).
In summary, the studies in this Special Feature indicate that specialization is not a certain ‘dead-end’ from an evolutionary perspective. First, transition from specialization to generalization is possible and even prevalent in certain ecological contexts. Second, specialization in traits related to foraging or reproduction can be associated with increased evolutionary success of some specialist clades, especially in specialist clades that experience greater transition rates to different specialist states (‘musical chairs' pattern described above ). Specialization also need not by itself be the driver of speciation. In the sword-bill-pollinated clade of passionflowers, Abrahamczyk et al.  find that specialized pollination is not the driver of speciation but instead the precondition for successful species persistence in small populations, which then adapt locally and evolve into separate species.
Hardy & Otto  raise the interesting point that the question of whether specialization influences speciation depends on how specialization is defined: ‘One grey area is how to define the relevant niche with respect to diversification, as generalists along some axes (e.g. resource use) may be specialists along others (e.g. in host–pathogen interactions). While theoretical models have shown that speciation is more likely when phenotypes are multi-dimensional … this raises a challenge for empiricists who must identify the phenotypic axes exhibiting the strongest diversifying selection in order to detect relationships between niche breadth and speciation’. Other authors in this Special Feature also tackle the issue of multi-dimensional axes of specialization.
(b) The axes of specialization
Specialization can be defined in a number of ways, and there are many ways to expand the ‘Jack of all trades, master of none’ paradigm. One way to define specialization is the breadth occupied by a species on niche axes. Most species probably are a generalist on some axes and a specialist on others . For example, some species of Dalechampia (Euphorbiaceae) exhibit apparently compensatory specialization/generalization on two pollination niches axes: specialization on the temporal axis (shorter duration of blooming season) is associated with generalization in the number of pollinator species used and vice versa . A growing body of evidence suggests that, while physiological trade-offs are uncommon, constraints may act to allow for specialization along alternate facets of a species' life history, e.g. where an advantage with one biotic partner or in inhabiting one niche comes at the expense of dealing with another [43–45]. Limits on floral specialization may also accrue from conflicting selection generated by herbivores or by abiotic factors. For example, specialization on large bees may select for large petals or bracts, but this may be countered by selection by herbivores (that use the same cues to find host tissues) , selection for water conservation in xeric environments [47,48] or selection for rapid seed production in seasonal habitats [49–51].
Expanding the number of axes to include both biotic and abiotic specialization can also provide insight into the underlying forces that spur the evolution of specialization. For example, pollinators often select for larger corolla size, but such increases exert a large cost in terms of water loss in dry environments [47,48,50], as noted above. Without information on the physical-environmental niche, it would be hard to ascertain why more species do not display large flowers. Examining these trade-offs in a phylogenetic framework can be a powerful approach to understanding the constraints on the evolution of specialization. Litsios et al.  provide evidence in this Special Feature of a negative correlation between environmental tolerances (in temperature, salinity and pH) and host specificities in clownfish and anemone mutualisms, which would likely confound phylogenetic analyses of diversification along any single specialization niche axis. Further, if differential specialization across resource axes is widespread, it may be a large contributor to the local coexistence of specialist and generalist species , and provide insight into the puzzling observation that specialists often do not outcompete generalists .
Despite finding that multiple axes contribute to specialization and interact to influence its evolution , we have little information on whether abiotic or biotic factors are more likely to drive specialization, or whether dispersal and geographical range provide environmental heterogeneity to spur initial transitions to specialization. Muschick et al.  examine these questions in this volume using the radiation of cichlids in Lake Tanganyika, testing the idea that specialization along multiple niche axes occurs according to a common sequence of transitions. In these cichlids, subdivision of trophic traits occurs in the early stages of adaptive radiation, while sexual communication traits diversify late in the radiation. The phylogenetic analysis of Muschick et al.  also provides limited support that specialization along biotic niche axes (diet) precedes specialization along abiotic niche axes (macrohabitat).
(c) Range extent, specialization, and diversification
Environmental heterogeneity is a key factor both in the evolution of specialization and in the evolutionary success of the resulting lineage . For example, if spatial heterogeneity is such that the abundance of a commonly used host changes rapidly over space (beta-diversity is high), this should accelerate the evolution of specialization . Much of the work in this area is done with herbivorous insects, with some studies suggesting that generalization is positively associated with large range size  and others finding cases where a specialist can have a much larger range if its host species is widespread . Models of the evolution of specialization that incorporate environmental heterogeneity and associative mating indicate that these variables can result in a decrease in gene flow between environments and contribute to speciation . Anderson et al.  examined pollinators in different parts of a plant species' range and found a close association between floral traits and the traits of the pollinators in the region but did not find strong evidence that these patterns greatly influenced gene flow and dispersal. Presumably, if selection pressures were consistent for generations, speciation could occur, yet pollinators may be too variable between years . Further work on the interplay between dispersal, range, and beta-diversity would lend insight into how specialization evolves and persists as well as the propensity of these factors to lead to speciation.
Widely dispersing organisms are more likely to have opportunities to expand their geographical range . Species occupying large ranges should experience divergent selection pressures upon their constituent populations; heterogeneity of selection pressures may in turn provide selection towards generalization across the entire species (leading to its scoring as a generalist in a phylogenetic trait reconstruction that might use just one accession to represent the species), but selection for different specialists at the local population level. Bonetti & Wiens  find evidence in amphibians, however, that the causal arrow could point in the opposite direction, with species with wide climatic tolerances (e.g. generalists along a climate niche) having the ability to persist in a greater number of locations and thus be exposed to a greater number of conditions in another niche axis. Range size could then influence the heterogeneity in selection pressures from biotic partnerships, allowing specialization to evolve in other niche axes. Bonetti & Wiens  find trait associations consistent with these expectations, with species having broad tolerances for variation in temperature and precipitation rather than trade-offs in these tolerances. For example, specialization along the climatic niche can reduce range size and set-up conditions conducive to the evolution of specialization along other niche dimensions. Thus, we can also observe positive associations in the levels of specialization between different axes of specialization rather than trade-offs.
(d) Conclusions and future directions
Forister et al.  list a number of interesting unanswered questions in the evolutionary ecology of specialization. While they concentrated on plant–insect associations, we attempt here to examine the process in a range of invertebrate and vertebrate systems (butterflies, bees, hummingbirds, amphibians, fish lineages). The problem of how to define specialization remains. Generally, our view of trade-offs appears to be widening, and this broadened perspective has the consequence of making trade-offs more readily apparent. Whether or not trade-offs are observed depends on how widely we cast the net; trade-offs do appear to be an important characteristic of specialization if we allow that they may operate between very disparate facets of a species' life history (e.g. pollination and herbivory). Furthermore, the issue of dispersal and range size presents further complicating factors, influencing the number and combinations of conditions encountered (and therefore the trade-offs observed). Recent studies indicate that the association between range size and niche breadth may vary in its strength depending on niche position as well as the axes of the niche (dietary or habitat) examined , suggesting that the complexity of these factors will provide an active area of research for some time.
From a conservation perspective, specialists are some of our most charismatic species, with the sword-billed hummingbird and the ca 50 species of plants that completely depend on it for pollination being a striking example. Thus, specialist species often receive greater conservation attention than do generalists . Although there is evidence that specialists can exhibit superior competitive strategies in their element (for foraging and/or reproductive assurance) , there is also evidence that their greater reliance on a smaller subset of species puts them at greater risk of extinction . From a macroevolutionary perspective, specialist clades may play a particularly important role in generating additional species at high rates due to host switching (the musical chairs process; see ), and this process would tend to make many specialists species appear ‘young’ on phylogenies. With the current conservation focus on the phylogenetic uniqueness of a given species , one implication is that the ‘young’ nature of many specialists may put them at lower prioritization for conservation. Additionally, while there is little evidence to suggest that specialization is irreversible or associated with lower speciation rates, specialist clades can experience higher extinction rates. Elevated speciation rates may buffer specialist clades from being lost to extinction to a certain extent , but further research should examine which specialist clades may be at the limits of the compensatory effects of speciation and experiencing net declines in species richness.
Network studies are providing some valuable insight into how specialization varies among communities. However, while connectance (the number of links between trophic levels compared to the maximum possible) is often equated with stability, loss of specialists will appear as increased connectance in networks . Additionally, gain of a high proportion of weedy generalist species in numerous communities will result in lowered beta-diversity and more homogeneous community composition over larger spatial scales . While these two outcomes would suggest that we lose biodiversity despite increasing stability in networks, there are at least two reasons to suggest that specialists may be as robust as generalists to environmental perturbations. First, specialist species often rely on generalist partners (i.e. networks tend to be asymmetrical and nested) [70,71]. Second, as exemplified in clownfish in this Special Feature, generalist–specialist trade-offs across multiple resource axes will act as a buffering force, such that specialists in bipartite networks may be habitat generalists, thus providing a further balancing mechanism that allows for coexistence of species . Recent studies have incorporated macroevolutionary and phylogenetic approaches into network studies to reveal the influence of shared traits on forming network interactions [28,64], and the new metrics currently emerging  will likely further provide an important link between the influences of evolutionary history, traits, and environmental heterogeneity.
In summary, specialists can experience greater evolutionary success compared to their generalist counterparts, possibly due to the very trade-offs that often drive specialization. In cases where we observe specialization along a number of different niche axes, historical range size may provide insight into how suites of specialized traits arise in lineages. Some of these insights would be impossible to gain without using a macroevolutionary perspective, and the studies in this issue highlight how comparative phylogenetic analysis sheds light on general principles underlying the evolution and persistence of specialized interactions.
One contribution to a Special feature ‘Evolution of specialization: insights from phylogenetic analysis’.
- Received August 12, 2014.
- Accepted August 29, 2014.
- © 2014 The Author(s) Published by the Royal Society. All rights reserved.
Many recent studies have suggested that the majority of animal-pollinated plants have a higher diversity of pollinators than that expected according to their pollination syndrome. This broad generalization, often based on pollination web data, has been challenged by the fact that some floral visitors recorded in pollination webs are ineffective pollinators. To contribute to this debate, and to obtain a contrast between visitors and pollinators, we studied insect and bird visitors to virgin flowers of Hypoestes aristata in the Bamenda Highlands, Cameroon. We observed the flowers and their visitors for 2-h periods and measured the seed production as a metric of reproductive success. We determined the effects of individual visitors using 2 statistical models, single-visit data that were gathered for more frequent visitor species, and frequency data. This approach enabled us to determine the positive as well as neutral or negative impact of visitors on H. aristata’s reproductive success. We found that (i) this plant is not generalized but rather specialized; although we recorded 15 morphotaxa of visitors, only 3 large bee species seemed to be important pollinators; (ii) the carpenter bee Xylocopa cf. inconstans was both the most frequent and the most effective pollinator; (iii) the honey bee Apis mellifera acted as a nectar thief with apparent negative effects on the plant reproduction; and (iv) the close relationship between H. aristata and carpenter bees was in agreement with the large-bee pollination syndrome of this plant. Our results highlight the need for studies detecting the roles of individual visitors. We showed that such an approach is necessary to evaluate the pollination syndrome hypothesis and create relevant evolutionary and ecological hypotheses.
Citation: Padyšáková E, Bartoš M, Tropek R, Janeček Š (2013) Generalization versus Specialization in Pollination Systems: Visitors, Thieves, and Pollinators of Hypoestes aristata (Acanthaceae). PLoS ONE 8(4): e59299. https://doi.org/10.1371/journal.pone.0059299
Editor: Katherine Renton, Universidad Nacional Autonoma de Mexico, Mexico
Received: June 28, 2012; Accepted: February 13, 2013; Published: April 10, 2013
Copyright: © 2013 Padyšáková et al. This is an open-access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.
Funding: This work was supported by the projects of Czech Science Foundation P505/11/1617, Grant Agency of the University of South Bohemia 136/2010/P and 156/2013/P, institutional support RVO:60077344 and the long-term research development project no. 67985939. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.
Competing interests: The authors have declared that no competing interests exist.
Debates about the generalization or specialization of pollination systems have been a prevailing theme in pollination ecology for many years. During that time, the view has been that pollination systems permanently balanced on the specialization–generalization continuum . The original idea that co-evolution often resulted in the specialization of plants and their pollinators came firstly up with Darwin’s evolutionary theory  and then was extended in later works . The specialization has been discussed over a long period and is closely related to the concept of pollination syndromes –, which are defined as a set of traits that convergently evolved as adaptations to similar pollinators. Simultaneously, the pollination syndrome concept has been opposed by some pollination biologists who noted that the links between floral traits and observed visitors are much weaker than predicted ,  and that co-evolution is often diffuse . Whereas the existence of generalized pollination systems was firstly manifested only for some plant species –, the more recent community-wide studies have shown that flowers of most plants are visited by a relatively high diversity of visitors and that generalization is much more common than was previously expected –.
Nevertheless, this broad generalization hypothesis has been criticized by other researchers , ,  who argue that some floral visitors that are usually considered in pollination webs are actually ineffective pollinators. In fact, a broad spectrum of diverse floral visitors with positive, neutral, and even negative effects on plant reproductive success can be found –. Several different techniques can be used to test the effects of particular pollinators. Indirect techniques, such as estimating visitor frequency rates – or direct measuring the total amount of pollen grains brought onto the stigma during a single visit of a particular visitor , , –, may not sufficiently consider the real contribution of particular visitors to the plant’s reproduction . One possible way to detect the visitor’s actual contribution directly is by using estimates from single visits to virgin flowers –. However, the single-visit approach has several weaknesses. Although it allows positive contributions to plant reproduction (i.e. the contribution of effective pollinators) to be quantified, it is not possible to reveal any negative effects of other visitors, so those visitors are simply classified as ineffective pollinators. Since many studies have shown negative effects of floral visitors –, these should be considered whenever hypotheses on floral evolution are developed .
Here, we focus on the pollination system of a broadly distributed Afrotropical plant species, Hypoestes aristata. This species shows the pollination syndrome ,  associated with bee pollination. Typically, its flowers have nectar-guide markings and produce a small amount of highly concentrated nectar. However, according to previous studies it is visited by a much broader spectrum of potential pollinators, including long-proboscid flies in South Africa ,  and various sunbirds, bees, flies, butterflies, and moths in our study area in the Bamenda Highlands, Cameroon , . In this area, H. aristata is the most favoured food plant of the sunbird Cinnyris reichenowi. Although the H. aristata morphology suggests pollinator specialization, it is apparently visited by a variety of birds and insects. Thus, H. aristata is an ideal model plant species for testing the validity of the concept of pollination syndromes. Simultaneously, examining its pollination system can contribute to the current debate about the proportion of generalization and specialization in pollination biology. The aim of our study was to answer the following main questions: (1) What is the spectrum of floral visitors of H. aristata? (2) Which visitors are effective pollinators? (3) Which visitors have neutral or negative effects on the reproduction of H. aristata? (4) Is the pollination system of H. aristata rather generalized, as suggested by previous studies on its floral visitors, or more specialized, as predicted by its floral traits? and (5) Is the bee pollination syndrome a good predictor of effective pollinators?
The study site was situated in the Mendong Buo area (6°5′26′′N 10°18′9′′E; 2100–2200 m a.s.l.), ca. 5 km southeast from Big Babanki (Kedjom-Keku community), in the Bamenda Highlands, North-West Province, Cameroon. This area is a mosaic of extensive pastures, frequently burned forest clearings dominated by Pteridium aquilinum, shrubby vegetation along streams, and remnants of species-rich tropical montane forests with a frequent occurrence of Schefflera abyssinica, Schefflera manii, Bersama abyssinica, Syzygium staudtii, Carapa procera, and Ixora foliosa. There is a single wet season from March to November, with annual precipitation ranging from 1780 to 2290 mm/year (For more details see: Cheek et al., Reif et al. & Tropek et al. –).
Our research was permitted by the Ministry of Scientific Research and Innovations of the Republic of Cameroon (permit no. 93/MINRESI/B00/C00/C10/C12) and the Ministry of Forestry and Wildlife of the Republic of Cameroon (permit no. 2306/PRBS/MINFOF/SG/DFAP/SDVEF/SC). Voucher insects were exported with the permission of the Ministry of Agriculture and Rural Development of the Republic of Cameroon (permit no. 15347/A/PPP/LBE). Our research was also permitted by Benjamin Vubangsi, the local chief of the Kedjom-Keku community, which owns the study area. The study was not conducted in any of the protected areas or on any protected species.
Our target plant species, Hypoestes aristata (Vahl) Sol. ex Roem. & Schult var. aristata (family Acanthaceae), is a clonal herb that grows up to 1.5 m high and is native to tropical sub-Saharan Africa , . The plant has hermaphroditic, zygomorphic flowers that are crowded into verticillate inflorescences. Dark purple blossoms with white nectar-guide markings on the upper lip have a pistil and 2 stamens long exerted from the corolla (Fig. 1). H. aristata produces a low volume (1.27 µl per flower) of hexose-dominant nectar of highly variable concentration (62.21% ±24.13; mean concentration ± [SD] w/w; i.e. sucrose equivalent mass/total mass; ). Nectar is accumulated in its 1-cm-long, narrow corolla-tube. Individual flowers last for about 5 days and can be found throughout the dry season. After pollination, a flower turns into a dehiscent capsule with up to 4 seeds (pers. obs.). H. aristata forms dense clumps, with several shoots flowering more or less simultaneously, which increases its local attractiveness for visitors. Usually, the plant dominates locally in disturbed montane forests, at their edges, in shrubby vegetation along streams, and in successionally older clearings. Experimental hand-pollinations during a preliminary study showed that H. aristata cannot effectively reproduce via autonomous selfing or parthenogenesis, and thus, is fully dependent on its pollinators (File S1; Fig. A in File S1; Table A in File S1).
Figure 1. The visitors of Hypoestes aristata: (A) Xylocopa cf. inconstans; (B) Xylocopa lugubris; (C) Cinnyris reichenowi; (D) Megachile sp.; (E) Bombyliidae; (F) Apis mellifera.
Photos (A)–(E) by R. Tropek, (F) by Š. Janeček.
Flower Visitors and Pollination Effectiveness
The flower visitors were studied from November to December 2010, when the plants of H. aristata are in full bloom. Ten shoots in 10 patches of flowering H. aristata were chosen within the whole study area. Shoots with several target flower buds were bagged individually with a fine mesh and the buds were marked. The bags were large enough to allow the flowers to completely open inside the netting. The following day, all open marked flowers on a shoot (5.3±1.29 per shoot; mean ± standard deviation [SD]) were observed simultaneously for a 2-hour session (i.e. one shoot with several open flowers was observed in one session) and all flowers were bagged again immediately after the observation. During each observation session, all animals that visited the marked flowers were recorded and identified to morphotaxa (Table 1, Movie S1). Observations of individual shoots were equally distributed throughout the day (between 0700 and 1800) to include all possible diurnal visitors and were limited to suitable weather conditions (sunny or partly cloudy). Fruits were harvested after maturation and their seeds were counted and weighed.
Due to many zero values, the data on seed production were not normally distributed. We thus analysed the effects of particular flower visitors on seed production using non-parametric permutation models. Seed numbers produced by individual flowers served as a dependent variable and visits of individual visitors as explanatory variables (i.e. each visitor represents one explanatory variable in each analysis). These explanatory variables contained either abundance data (i.e. numbers of visits to individual flowers during 2-hour observations – see Model 1 below) or presence-absence data (i.e. the information if the visitor at least once visited or did not visit the flower – see Model 2 below). Note that we also considered the value of zero at the unvisited flowers for abundance data in Model 1. To avoid the variability in seed production that can be explained by having more than one visitor to a flower during the 2-hour session we used the Type II sums of squares approach for a given explanatory variable –. In this way, the sum of squares for each visitor (explanatory variable) was calculated as the increase of the model sum of squares (and equivalently the decrease in the error sum of squares) due to adding this visitor into a model that already contained all of the other visitors . Thus, only the variability that could not be explained by other than just the tested visitor was considered. Two models with different biological predictions were established. Model 1 assumed that the number of developed seeds increases or decreases with visitation frequency (e.g. visitors continuously saturate the stigma with pollen grains or continuously consume nectar from the flower and decrease the attraction of the flower by this way). Model 2 assumed that the most important is whether the visitor visit the flower or not (e.g. flower receives enough pollen to produce the maximum number of seeds after a single visit from each pollinator or the nectar is completely depleted during the single visit). Following these approaches, the log (x+1) transformed numbers of visits by individual visitors to each flower were used in the first model as an explanatory variable, whereas binary coded visits (i.e. at least one visit = 1, no visit = 0) to each flower were analysed in the second model. In both the models, the factors (visitors) with high p-values and negligible contribution to total variation in seed set among flowers indicated by negative estimates of the component of variation were stepwise excluded from the model , , . After exclusion of the term with the lowest negative value of the component of variation, the models were recalculated. Consequently, only visitors with positive values of components of variance remained in the models . The spatial autocorrelation effect (i.e. the term ‘shoot’) was considered in the models as a random variable. This term was always significant (i.e. individual shoots differed), and we have not shown the results for this term in Table 1. Except for the above described whole models, where all visitors were considered, we calculated marginal tests for each of the visitors. These tests demonstrate how visits of each visitor are related to seed production when each visitor is taken alone, ignoring others. Permutation tests were run with PERMANOVA+ for PRIMER .
During the observations of 539 flowers, 1979 flower visits, involving fifteen visitor morphotaxa, were recorded (Table 1). On average, 198 (±68.52) visits per patch and 3.67 (±2.61) visits per flower were detected. Although more than 95% of the flowers were visited at least once, less than 15% of the visited flowers produced fruit with viable seeds.
The total visitor community was highly dominated by two carpenter bees: Xylocopa cf. inconstans (Fig. 1A; including X. inconstans and X. caffra, which are hardly recognisable from each other in the field) and Xylocopa lugubris (Fig. 1B); followed by the honeybee Apis mellifera (Fig. 1F) and the northern double-collared sunbird Cinnyris reichenowi (Fig. 1C; Fig. 2). Nevertheless, the visitors’ abundances and community composition differed considerably among patches (Figure S1). All the studied patches had a similar pattern of visitor distribution, with one or a few highly abundant taxa, while most other visitors were rarely observed. X. lugubris was the only visitor taxon observed at all studied patches.
Figure 2. Total visitation frequencies.
Abbreviations: CinBou = Cinnyris bouvieri, CyaOri = Cyanomitra oritis, CynRei = Cinnyris reichenowi, Bom = Bombyliidae, Syr = Syrphidae, Dipt = other dipterans, Lep = Lepidoptera, ApiMel = Apis mellifera, AntSp = Anthophora sp., MegSp = Megachile sp., Api = other bees, XylInc = Xylocopa cf. inconstans, XylLug = Xylocopa lugubris, XylNig = Xylocopa nigrita, XylEry = Xylocopa erythrina.
Although 5 visitor taxa significantly affected seed production, if the other visitors were not considered (marginal tests for models 1 and 2, Table 1), only three visitor taxa were able to explain the variability in the reproductive success of H. aristata when the variability which could be explained by more visitors was eliminated (whole models 1 and 2, Table 1). Both the whole models indicated that the carpenter bee X. cf. inconstans and the leafcutter bee Megachile sp. (Fig. 1D) increased plant reproductive success, whereas the honeybee A. mellifera was related to fruit abortion (Table 1). According to the estimated values in the first model, X. cf. inconstans is three times more important pollinator than Megachile sp. Most of the variability in the second model was explained by the visits of A. mellifera.
The majority of the flowers were visited repeatedly during our observations, usually by more than one visitor taxon, but 79 observed flowers were visited just once. These single visits were made by the four most frequent visitors, but flowers produced seeds only after a single visit of either X. cf. inconstans or X. lugubris, not of A. mellifera or C. reichenowi (Table 2). Although the flowers visited once by these four visitors did not significantly differ in seed production (permutation ANOVA; d.f. = 3; F = 1.98; p = 0.114), Xylocopa spp. differed from A. mellifera and C. reichenowi which were indicated by the models (Table 1) as visitors with rather negative influence on the seed production (permutation ANOVA; d.f. = 2; F = 5.07; p = 0.039). Although a honeybee might receive a pollen load from the anthers, it rarely deposits the pollen because it is too small to touch the stigma when inserting its head into the flower to forage on nectar (see Fig. 1F). Similarly, sunbirds, while visiting, introduced their bills partially or totally into the floral tube in a space between the upper lip and both sexual organs.
Summarizing all the analyses performed, the carpenter bee X. cf. inconstans seemed to be the main pollinator of the plant in the study area. The importance of the other carpenter bee, X. lugubris, followed from its total high frequency of visits. X. lugubris equally visited the successfully and unsuccessfully pollinated flowers, which means that, in total, it contributed to pollination of the flowers only occasionally. Its high frequentness, however, guarantees a relatively bigger contribution to seed production than the less frequent visitors. The leafcutter bee Megachile sp. positively affected seed production of H. aristata (Models I and II in Table 1), nevertheless its visitation rate was too low (Fig. 2) to be crucial to H. aristata’s reproduction in the study area.
We have described the reproductive and pollination system of H. aristata, and have shown that the apparently generalized pollination system is actually highly specialized in the study area and that the effective pollinators are in agreement with the pollination syndrome of this plant.
Due to our experimental approach, we were able to determine not only the pollinator effectiveness but also the negative impact of visitors on the studied plant’s reproduction. Interestingly, single visits from 2 frequent visitors, the honeybee A. mellifera and the sunbird C. reichenowi, did not result in any seed production, and visits of A. mellifera even decreased the reproduction success of H. aristata.
The effectiveness of both the above mentioned carpenter bees in the H. aristata pollination system is in accordance with statements of other researchers, showing the Xylocopa species as extremely important pollinators in various tropical systems –. The honey bee A. mellifera is commonly considered to be a generalist forager, visiting many plant species . Although it usually visits flowers more frequently than other flower visitors –, its effectiveness as a pollinator is likely to differ, depending upon its foraging behaviour ,  and the morphology of the flowers . Our finding that A. mellifera had a negative impact on H. aristata seed production might be because of a combination of both of the above-mentioned factors. We assume that, as has been shown by other studies , , A. mellifera acted as a floral thief, removing a substantial part of the available nectar or pollen and thus making the flower unattractive for other visitors.
Among the three sunbird species visiting H. aristata, C. reichenowi was the most frequent visitor , , but it did not effectively pollinate the flowers. Its ineffectiveness could be related to the relatively small and specialized flowers of H. aristata that do not fit the birds’ heads (Fig. 1C). Thus, the anthers and stigma contacted the lower part of the bird’s bill, which seems to be inappropriate for pollen transfer. In bird-pollinated flowers, pollen grains typically attach firmly to a bird’s crown when the bird inserts its bill into the perianth to extract nectar , . On the basis of our results, we consider C. reichenowi to be a nectar thief, although there was no obvious negative effect on H. aristata reproduction, in contrast to that by A. mellifera. In accordance with our observations (Fig. 1), we agree that ‘trait-matching’ between flowers and their visitors plays an important role in pollination interactions , , –.
A limitation of our study is the relatively small study area size and short time in which the study was performed. It has been shown that diversity, abundance, and the importance of individual visitors may differ depending on the time and place –. Conversely, H. aristata in South Africa is also visited by carpenter bees ; thus, there is a high possibility that they are the main pollinators in that region. Moreover, our findings are in accordance with the expectations from ‘trait-matching’; i.e. the honeybee A. mellifera rarely reaches the stigma to deposit pollen and the sunbird carries pollen on its lower bill. Therefore, neither of these species should be an effective pollinator. Nevertheless, similar studies conducted in different African regions would substantially contribute to this debate.
Choosing the right field technique for measuring the pollination or plant reproductive success is important since there are several possible methods with various weaknesses and benefits . Because of the shortcomings of using the single-visit method to estimate pollination effectiveness , , we chose the approach based on 2-hour observation periods. Basing observations on time-defined periods is more suitable to detect the potential effects of the whole spectrum of floral visitors, including occasional visitors; and to reveal both positive and negative effects of individual visitors. This method is, moreover, less laborious than bagging flowers after each single visit. If the length of the observation period is well chosen the dataset can also include single-visit data, at least for the more frequent pollinators. A drawback of this method follows the fact that the seed set is usually formed after multiple visits from the same or different visitors.
The analyses of the pollination system of H. aristata show different roles for individual visitors. Our finding that the two carpenter bees were the only important pollinators among the wide spectrum of floral visitors is in accordance with the bee pollination syndrome of H. aristata and with the concept of pollination syndrome , . Nevertheless, as much as successful pollination is highly dependent on ‘trait-matching’ between flowers and their visitors , , –, we agree that the visitor’s body size plays an important role in the assessment of the pollination syndrome validity. The large bees were effective pollinators whereas the relatively smaller bee A. mellifera had a negative effect on H. aristata reproduction. This assumption supports the idea that the bee pollination syndrome should be divided further into large-bee and small-bee syndromes , . Our results are also in accordance with the most effective pollinator principle , supposing that the plant traits evolved as a response to the most effective pollinators. In contrast to the expected generalization of this system, we found a high degree of specialization. This would be even more apparent if we followed the ideas of Fenster et al.  and considered the similarly large bees Xylocopa spp. and Megachile sp. as one functional group exerting similar selection pressures. Moreover, we also observed visitors with negative or potentially negative effects on plant reproductive success. As shown in other studies , these visitors can create different selection pressures on various floral traits. If they are overlooked or even considered as pollinators, then our understanding could lead to a total misinterpretation of the pollination systems. Our conclusions would be completely different if we considered all visitors as pollinators as is typical in plant-pollinator web studies (Fig. 3). It also clearly demonstrates why pollination networks frequently show flowers to be phenotypically specialized but ecologically generalized .
Figure 3. Interactions between H. aristata and its visitors.
(A) Binary interactions showing just the visitor-plant interaction - the approach commonly used in pollination networks. (B) Quantitative interactions showing the frequencies of visits by individual visitors - the less frequently used approach in pollination networks. (C) Quantitative interactions indicating the role of individual visitors: yellow = important effective pollinators, green = pollinators with a marginal effect on H. aristata reproduction, red = nectar thieves with a negative impact on H. aristata reproduction; brown = nectar thieves with a potential negative effect on H. aristata reproduction; and black, visitors with no effect on H. aristata reproduction. Abbreviations: CinBou = Cinnyris bouvieri, CyaOri = Cyanomitra oritis, CynRei = Cinnyris reichenowi, Bom = Bombyliidae, Syr = Syrphidae, Dipt = other dipterans, Lep = Lepidoptera, ApiMel = Apis mellifera, AntSp = Anthophora sp., MegSp = Megachile sp., Api = other bees, XylInc = Xylocopa cf. inconstans, XylLug = Xylocopa lugubris, XylNig = Xylocopa nigrita, XylEry = Xylocopa erythrina.
Although we assume that the progress from studies on simple pollination systems (often including just one pollinator and one plant species) to community level studies is the right direction for pollination biology, we must urge, together with other researchers , , that without any knowledge of the roles of individual visitors, we cannot confirm the validity of the pollination syndrome hypothesis, determine the degree of generalization, nor create a relevant evolutionary hypothesis.
Preliminary study on the breeding system of Hypoestes aristata. The breeding system was studied by emasculation and pollen supplementation in five experimental treatments. The results showed that the experimental treatments differed in the reproductive success of H. aristata; i.e. in the number and total weight of seeds per fruit. Table A, Results of the hand-pollination experiment done by permutation mixed models. Fig. A, Seed number per flower (Means and Standard Errors) of Hypoestes aristata in five experimental treatments.
The video file attached shows the representative visitors of Hypoestes aristata while foraging for the nectar. Shots were taken at the study site by the small hand camcorder during the field studies in 2010 and 2012. Some of the presented shots were intentionally slowed to better show the visitorś behaviour. High definition of the video file was converted to fit the size limit given by the journal.
We are grateful to the entire Kedjom-Keku community and particularly Ernest Vunan Amohlon from SATEC NGO for their kind reception in the Big Babanki village. We also thank J. Bartošová for the field equipment preparation; J. Straka, J. Halada, and A. Vlašánková for their help with the insect identification; M. Sweney and Editage editor for English proofreading; and the 3 anonymous reviewers for their useful comments.
Conceived and designed the experiments: EP MB RT SJ. Performed the experiments: EP MB RT. Analyzed the data: EP SJ. Contributed reagents/materials/analysis tools: EP MB RT SJ. Wrote the paper: EP RT SJ.
- 1. Willmer P (2011) Pollination and floral ecology. Princeton and Oxford: Princeton University Press. 778 p.
- 2. Darwin C (1859) On the origin of species by means of natural selection, or the preservation of favoured races in the struggle for life, 1st edn. London: Murray John.
- 3. Stebbins GL (1970) Adaptive radiation of reproductive characteristics in angiosperms. Annual Review of Ecology and Systematics 1: 307–326.
- 4. Faegri K, van der Pijl L (1979) The principles of pollination ecology, third revised edition. Oxford: Pergamon Press xi. 244 p.
- 5. Hodges SA, Arnold ML (1994) Columbines - a geographically widespread species flock. Proceedings of the National Academy of Sciences of the United States of America 91: 5129–5132.
- 6. Renner SS, Feil JP (1993) Pollinators of Tropical Dioecious Angiosperms. American Journal of Botany 80: 1100–1107.
- 7. Manning JC, Goldblatt P (1997) The Moegistorhynchus longirostris (Diptera: Nemestrinidae) pollination guild: Long-tubed flowers and a specialized long-proboscid fly pollination system in southern Africa. Plant Systematics and Evolution 206: 51–69.
- 8. Waser NM, Chittka L, Price MV, Williams NM, Ollerton J (1996) Generalization in pollination systems, and why it matters. Ecology 77: 1043–1060.
- 9. Ollerton J, Alarcon R, Waser NM, Price MV, Watts S, et al. (2009) A global test of the pollination syndrome hypothesis. Annals of Botany 103: 1471–1480.
- 10. Strauss SY, Sahli H, Conner JK (2005) Toward a more trait-centered approach to diffuse (co)evolution. New Phytologist 165: 81–89.
- 11. Primack RB (1979) Reproductive-biology of Discaria-Toumatou (Rhamnaceae). New Zealand Journal of Botany 17: 9–13.
- 12. Bullock SH, Delrio CM, Ayala R (1989) Bee visitation rates to trees of Prockia-Crucis differing in flower number. Oecologia 78: 389–393.
- 13. Beattie AJ (1971) Pollination mechanisms in Viola. New Phytologist 70: 343–360.
- 14. Memmott J (1999) The structure of a plant-pollinator food web. Ecology Letters 2: 276–280.
- 15. Memmott J, Waser NM (2002) Integration of alien plants into a native flower-pollinator visitation web. Proceedings of the Royal Society of London Series B-Biological Sciences 269: 2395–2399.
- 16. Olesen JM, Jordano P (2002) Geographic patterns in plant-pollinator mutualistic networks. Ecology 83: 2416–2424.
- 17. Jordano P, Bascompte J, Olesen JM (2003) Invariant properties in coevolutionary networks of plant-animal interactions. Ecology Letters 6: 69–81.
- 18. Olesen JN, Dupont YL, Ehlers BK, Hansen DM (2007) The openness of a flower and its number of flower-visitor species. Taxon 56: 729–736.
- 19. Niemirski R, Zych M (2011) Fly pollination of dichogamous Angelica sylvestris (Apiaceae): how (functionally) specialized can a (morphologically) generalized plant be? Plant Systematics and Evolution 294: 147–158.
- 20. Kishore K, Kalita H, Singh M, Avasthe R, Pandey B, et al. (2011) Pollination studies in large cardamom (Amomum subulatum Roxb.) of Sikkim Himalayan region of India. Scientia Horticulturae 129: 735–741.
- 21. Inouye DW (1980) The terminology of floral larceny. Ecology 61: 1251–1253.
- 22. Maloof JE, Inouye DW (2000) Are nectar robbers cheaters or mutualists? Ecology 81: 2651–2661.
- 23. Roubik DW (1982) The ecological impact of nectar-robbing bees and pollinating hummingbirds on tropical shrub. Ecology 63: 354–360.
- 24. Janeček Š, Patáčová E, Bartoš M, Padyšáková E, Spitzer L, et al. (2011) Hovering sunbirds in the Old World: occasional behaviour or evolutionary trend? Oikos 120: 178–183.
- 25. Herrera CM (1987) Components of pollinator quality - comparative analysis of a diverse insect assemblage. Oikos 50: 79–90.
- 26. Herrera CM (1989) Pollinator abundance, morphology, and flower visitation rate - analysis of the quantity component in a plant-pollinator system. Oecologia 80: 241–248.
- 27. Stone JL (1996) Components of pollination effectiveness in Psychotria suerrensis, a tropical distylous shrub. Oecologia 107: 504–512.