SEARCH

SEARCH BY CITATION

Keywords:

  • carnivorous plants;
  • evolutionary convergence;
  • exploitation of perceptual bias;
  • pitcher plants;
  • plant population and community dynamics

Summary

  1. Top of page
  2. Summary
  3. Mimicry in plants
  4. Olfactory mimicry in carnivorous plants
  5. Other examples of olfactory mimicry
  6. Visual mimicry
  7. Conclusion
  8. Acknowledgement
  9. References

1. Based upon similarity in visual or olfactory appearances, recent studies concluded that mimicry of plants or plant parts occurs in distinct systems, in carnivorous plants that mimic flowers to increase capture success, in thorny plants to defend themselves against predators and in parasitized ants to increase parasite dispersal.

2. Taking the example of the carnivorous plant Nepenthes rafflesiana emitting volatiles that insects find attractive, and that also occur in flowers, we argue here that two alternative explanations are more plausible than mimicry: exploitation of perceptual bias and convergence.

3. Exploitation of perceptual bias requires only the well-established phenomenon of generalization of rewarding (or unrewarding) experiences; and does not require mimicry’s more specialized conditions of sufficient similarity of the mimic to a specific model, as well as consistency in the overlaps of their phenology and distribution, to cause misidentification by animals.

4. Like most flowers, the pitchers of N. rafflesiana offer a nectar reward to visiting insects. Evolution acting on both flowers and pitchers may have converged on the use of similar volatiles entirely independently, simply because these volatiles are effective at attracting nectar-seekers.

5. We conclude that not only are there currently no demonstrations of mimicry of a flower’s scent among carnivorous plants, there is also no evidence that mimicry (in any sensory modality) plays any part in carnivory by plants.

6.Synthesis. Researchers should guard against prematurely accepting intuitively appealing explanations of mimicry that may hinder the search for the true mechanisms underlying the evolution of some fascinating insect–plant interactions.


Mimicry in plants

  1. Top of page
  2. Summary
  3. Mimicry in plants
  4. Olfactory mimicry in carnivorous plants
  5. Other examples of olfactory mimicry
  6. Visual mimicry
  7. Conclusion
  8. Acknowledgement
  9. References

Mimicry of flowers has always been a controversial issue, with an influential review by Roy & Widmer (1999) claiming that no previous studies had fully verified the existence of floral mimicry in any system. They argued that, in order to demonstrate mimicry, a study would need to show (i) overlaps between distributions and phenologies of model and mimic, (ii) that individual pollinators move freely between mimics and models, and (iii) that the similarity between mimic and model is adaptive; that is, relevant to fitness. In the last decade there have been a very small number of studies that provided these evidences of floral mimicry (e.g. Anderson, Johnson & Carbutt 2005; Benitez-Vieyra et al. 2007; Peter & Johnson 2008). However, a number of recent publications also continue to claim demonstrations of mimicry without eliminating (often more likely) alternative explanations for their data: we suggest that the recent Journal of Ecology paper by Di Giusto et al. (2010) is worthy of discussion in this regard. In doing so, we wish to emphasize that our decision to focus on this paper should certainly not be taken as suggestive that we feel that this manuscript is especially problematic. On the contrary, the issues that we raise are pervasive in studies of plant–animal interaction more generally (see review in Schaefer & Ruxton 2009). For example, mimicry has been assumed rather than shown in other recent studies of plant carnivory (e.g. Jürgens, El-Sayed & Suckling 2009), in studies suggesting defensive Müllerian mimicry between thorny plants (e.g. Lev-Yadun 2009), and in studies suggesting fruit mimicry by ants or Lepidoptera larvae (Yanoviak et al. 2008; Piskorski, Trematerra & Dorn 2010).

Olfactory mimicry in carnivorous plants

  1. Top of page
  2. Summary
  3. Mimicry in plants
  4. Olfactory mimicry in carnivorous plants
  5. Other examples of olfactory mimicry
  6. Visual mimicry
  7. Conclusion
  8. Acknowledgement
  9. References

The study reported by Di Giusto et al. (2010) concerned the carnivorous pitcher plant Nepenthes rafflesiana. This plant shows dimorphism of pitchers, with aerial (upper) pitchers being different in morphology from ground-resting (lower) pitchers. Upper pitchers emit more volatiles, producing a stronger smell. Consistent with insect attraction being mediated by scent emissions, Di Giusto et al. demonstrated that upper pitchers attracted a greater quantity and diversity of insects than the lower pitchers; including a guild of flower-visiting species that were absent from lower pitchers. They also demonstrated that, in the absence of visual cues, insects found the scent of the upper pitchers attractive. Finally, they identified terpenoids and benzenoids in the blend of upper-pitcher scent that are also characteristic of many floral scents. On the basis of this evidence, these authors conclude that the upper pitchers are flower mimics, duping pollinators through olfaction; as can be inferred from the paper’s title ‘Flower-scent mimicry masks a deadly trap in the carnivorous plant N. rafflesiana’. Here, we argue that (as in the other studies mentioned above) the case for mimicry is not yet sufficient to be compelling; and that alternative, evolutionarily simpler, explanations for similarity (in scent or colour) among plant species or plant organs exist and should be considered before mimicry can be established.

Mimicry of a flower by an insect-trap that the insect would benefit from avoiding implies that the insect is duped because of similarity (in terms of stimulation of the insect’s sensory system) between individuals of the mimic (the trap in this case) and the model (the flower). The key requirement here is that there is a specific model that the insect finds attractive. The study of Di Giusto et al. does not identify what the model for mimicry by pitchers might be. As such, there is no quantification of the overlap in distribution and phenology between pitchers as mimics and flowers as models. Indeed, the lack of identifying a model precludes that similarity and phenological overlap between mimic and model can be quantified. Note that the model can be a single species (or plant organ) or several species as, for example, in the case of Müllerian mimicry. However, even in Müllerian mimicry the most common species often evolves into being the model of the others (Joron & Mallet 1998). Possible models that the pitchers could mimic are the flowers of the carnivorous plant itself, or the flowers of a species that reliably co-occurs in the same places as the pitcher plant.

Why is it important to specify a model? It is important because – as Di Giusto et al. state – the odour profiles emitted from distinct plant organs such as flowers and vegetative parts overlap widely (e.g. Knudsen & Tollsten 1993; Knudsen, Tollsten & Bergstrom 1993; Pichersky & Gershenzon 2002; Knudsen et al. 2006; Pichersky, Noel & Dudareva 2006a; Pichersky, Sharkey & Gershenzon 2006b). For example, terpenoids, one of the compound classes supposedly involved in mimicry in N. rafflesiana, are emitted from a wide range of plant tissues (Theis & Lerdau 2003). Many terpenoids have defensive functions such as the amelioration of stress induced by high temperatures and the reduction of damage caused by oxidative stress (Loreto et al. 2004). These defensive functions may provide a functional explanation for why odour profiles overlap between distinct plant organs at a given location. Furthermore, overlaps in odour profile may reflect historical constraints. One scenario for the evolution of flower odours is that they were originally employed in defence against herbivorous insects and only secondarily co-opted for the attraction of pollinators (Pellmyr & Thien 1986). This scenario would also expect an overlap between the odour profiles of plant organs that – like flowers and pitchers – originated from leaves. Consequently, the overlap between odour profiles among plant parts (pitcher and flowers in this case) is not sufficient evidence to distinguish the hypothesis of mimicry from other scenarios.

Quantifying whether the overlap between odour profiles of putative mimic and model exceeds the confidence intervals of overlap between species or plant organs that co-occur at the same times would represent a first step to document that similarity between two species or plant organs is higher than expected by alternative functional explanations. In other words, mimicry requires (i) convergence and (ii) that this convergence is driven by the sensory or cognitive system of animals.

A further issue is that attraction of insects to pitchers emitting flower-like volatiles does not necessarily indicate that they misidentify pitchers as flowers. Alternatively, they could be attracted to terpenoid and benzenoid blends because of inherent sensory or cognitive biases towards these odour components. Arguably, cognitive biases may arise from rewarding experiences with flowers where insects have been selected to allow a generalized pollinator to respond to a wide range of flowers. This could be compatible with generalized mimicry. However, a generalized mimicry system where the models converge on the putative mean phenotype of many models is very difficult to disentangle from the exploitation of perceptual biases (EPB) (Schaefer & Ruxton 2009). Crucially, the exploitation of pre-existing biases is quite different from mimicry, and the key difference is that widely accepted definitions of mimicry require the insect to misidentify the mimic as an example of specific models (e.g. Pasteur 1982). As an alternative to mimicry, perceptual biases towards terpenoid and benzenoid blends could arise because these stimulate the odour receptors more effectively than other components. Likewise, insects may be attracted (and hence be biased) towards common floral compounds such as monoterpenes and aromatics because these are compounds that insects use in their own communication with conspecifics (Schiestl 2010). Consequently, showing attraction towards compounds and showing that these compounds overlap with those of flowers are two aspects that are not unique to the mimicry hypothesis but are also expected by alternative hypotheses.

The EPB is less restrictive than mimicry. There is no specific model, there is thus no need for overlap between phenologies and no misidentification is involved. Rather, EPB requires that the pitcher plant is able to exploit the generalization that most animals show in being attracted to objects that stimulate the senses in a similar way to previously encountered rewarding objects. Such generalization can still be selected in the face of exploitation by pitchers, provided (i) the insects benefit from exploitation of a range of flower types, and (ii) pitcher plants are not too common. Both these conditions will commonly be met. Many individual pollinating insects are known to visit dozens of different flowers in the course of a day (Waser & Ollerton 2006) and pitcher plants are not necessarily the dominant species in environments where they occur. Thus, we consider EPB as a plausible alternative explanation to mimicry for the observations reported by Di Giusto et al. (2010). Further, EPB requires only the very well-established phenomenon of generalization of rewarding experiences, and does not require mimicry’s more specialized conditions of sufficient similarity of the mimic to a specific model, as well as consistency in the overlaps of their phenology and distribution, to cause misidentification. Thus, we suggest that EPB will evolve more readily than floral mimicry, and Occam’s razor would suggest that it might be more prudent to adopt EPB as the working explanation in the light of the currently available information.

How can one disentangle the mimicry hypothesis from the EPB model? Support for mimicry (either Batesian or Müllerian) would be greatly strengthened; however, if it were to be demonstrated that (i) the upper pitchers of this species coincide (in space and time) with the flowers of a specific species; (ii) the flowers and pitchers share insect visitors and (iii) the overlap in insect visitors can be reduced by experimental manipulation of pitchers to reduce similarity (in appearance and/or volatiles) to that specific type of flower. Support for the EPB model would be greatly strengthened if attraction to the volatile blends was innate, i.e. present in insects that had not experienced flowers before.

There may be an even simpler explanation for the similarity of volatiles between these pitchers and some flowers: simple evolutionary convergence on an effective strategy. There are countless examples where evolution has found very similar solutions to a similar problem arising from different starting points (e.g. Coyne 2009; Futuyma 2009). Flowers are attractive to many insects because they offer a nutritional reward in the form of nectar. The pitchers of N. rafflesiana also offer a nectar reward (Di Giusto et al. 2010 and references therein). This nectar has been shown to enhance prey capture rates (Bauer, Bohm & Federle 2008). It may be that nectar-seeking insects do not have to be duped into visiting pitchers; they may be selected to actively seek out pitchers as a food source. It should be noted in this regard that pitchers are not quite the ‘deadly trap’ that Di Giusto et al.’s title implies, and mortality rate of insects visiting pitchers can be low. Merbach et al. (2001) estimated mortality rates of 0.34%–1.6% in a study of capture rates of pitchers of five Nepenthes species. Similar low values have been found in other studies of carnivorous plants: Sarracenia purpurea 1–2% (Newell & Nastase 1998), Darlingtonia californica 1.3% (Dixon, Ellison & Gotelli 2005). The nutritional rewards offered by the pitchers may be sufficient to still attract insects in the face of such a low mortality risk. It is worth remembering in this regard that flowers can be places of heightened predation risk too; with predators waiting specifically on or near flowers, to ambush the insects drawn by the flower (Heiling et al. 2005). If flowers and pitchers are both attractive to insects, then it would be no surprise at all if evolution acting on both flowers and pitchers converged on the use of similar volatiles entirely independently, simply because these volatiles are the most effective and/or more economically produced. In this scenario the similarity of volatile blends occurs simply because independent evolutionary trajectories converged on a similar effective solution to the challenge of producing a volatile signal of nectar reward that enhances arrival rate of nectar-seeking insects. However, this scenario remains very speculative since the nectar provided by pitchers may serve to attract mutualistic ant species rather than (or as well as) prey, and the cues used by ants (or any other insects) to locate such extra-floral nectaries – and in particular whether volatiles are involved – are unknown.

If flowers and pitchers are both attractive to individuals, then flower-like features in a pitcher could be selected by convergence, or EPB or Müllerian mimicry. For authors to argue for Müllerian mimicry over the more parsimonious alternatives they must identify the specific species whose pitchers and flowers are involved in the mimicry complex. Müllerian mimicry requires that specific species in the mimicry complex exert selection on each other. This is essentially the same requirement as the need to identify the flower model for exploitative (Batesian mimicry) discussed above.

The issues that we raise in this paper apply to other papers on mimicry by carnivorous plants. For example, Jürgens, El-Sayed & Suckling (2009) also report on the volatile emission of three pitcher plant species, and draw attention to overlap between these volatiles and those found in many flowers and fruits. Although they speculate in the abstract (p875) that ‘these traps are possibly flower or fruit mimics’, they subsequently adopt a more cautious interpretation in the main text (p884) that suggests that no specific model species is mimicked and so perceptual exploitation is a more likely explanation.

We suggest that there is currently no evidence that mimicry (in any sensory modality) plays any part in carnivory by plants. Because pitchers of some carnivorous plants often sport striking colour patterns, this has caused several authors to suggest that such pitchers mimic flowers in their visual appearance (e.g. Wickler 1968; Wiens 1978; Pasteur 1982; Williamson 1982). However, Joel (1988) argued against this theory, suggesting that (i) there was no evidence that any insects visit a pitcher because they have been duped, (ii) no specific flower models have been identified, (iii) there seems no selection pressure for pitchers to be rare relative to flowers and (iv) pitchers provide a nectar reward that insects can exploit. We know of no work that provides evidence that requires revision of Joel’s conclusions.

Other examples of olfactory mimicry

  1. Top of page
  2. Summary
  3. Mimicry in plants
  4. Olfactory mimicry in carnivorous plants
  5. Other examples of olfactory mimicry
  6. Visual mimicry
  7. Conclusion
  8. Acknowledgement
  9. References

Di Giusto et al. (2010) cite two other publications that they consider to provide examples of scent mimicry of flowers. In both these cases, we again consider that alternative explanations are more likely. The study of Raguso & Roy (1998) concerns the scent produced by pseudo-flowers of the rust fungi Puccinia monoica. These authors specifically tested for similarity of the scent blend produced by the pseudo-flower to those of the flowers of the host plant and of other co-blooming plants. The pseudo-flower blend was quite different from that of any flowers of any plant whose flowers might be found spatially close to the pseudo-flowers. Indeed, the authors suggested that fungi do not mimic flower scent but that the difference in scent profiles might be adaptive for fungi ‘in reducing gamete loss by re-enforcing constancy among foraging insects’. It should be noted in this regard that the pseudo-flowers offer nectar rewards to insect visitors that might spread the fungal spores. Indeed, the pseudo-flowers provide considerably higher levels of nectar than co-occurring flowers (Roy 1993, 1994). In this case, it seems clear that Batesian mimicry can actually be excluded as an important evolutionary factor in selection for pseudo-flower scent, but that EPB and/or convergent evolution remain highly plausible candidates.

The other example of putative flower-scent mimicry cited by Di Giusto et al. (2010) is the study of Dufaÿ, Hossaert-McKey & Anstett (2003) of the dwarf palm Chamaerops humilis and its species-specific pollinating weevil Derelomus chamaeropsis. In this palm, the flowers offer a nectar reward to the weevil, but the odours that the weevil finds attractive are emitted by the leaves rather than the flower. The emissions of the leaves include compounds that are also common in floral scents. Here again, there is no suggestion that the weevil is exploited by causing it to misidentify examples of a mimic (the leaf) as examples of a specific model (a flower), and thus we can be confident that floral mimicry is not a good description of this system. Since the weevil is a plant specialist that has no difficulty obtaining the reward from the flowers (and transferring pollen between conspecific flowers) there is no suggestion that this system is any more exploitative than any other example of animal-based pollination. Thus, the most plausible explanation for this situation is convergent evolution, with leaves using similar compounds to flowers, not because of any advantage arising from that similarity but because these compounds are effective at enhancing pollination and evolution has found its way to similar good solutions from different starting points.

Visual mimicry

  1. Top of page
  2. Summary
  3. Mimicry in plants
  4. Olfactory mimicry in carnivorous plants
  5. Other examples of olfactory mimicry
  6. Visual mimicry
  7. Conclusion
  8. Acknowledgement
  9. References

Our arguments carry over to other recent papers on mimicry more generally where we feel that alternative explanations to mimicry have not been sufficiently evaluated. Recently, Yanoviak et al. (2008) describe a striking case where fruits may be the model for parasite-induced changes in ants. Ants parasitized by a nematode developed ‘conspicuous red gasters’ that contained hundreds of yellowish nematode eggs. Comparing predation on infected ants with that of uninfected ants, in a clever experiment where both were experimentally restrained, the authors found that infected ants were removed more quickly. In a second experiment, they found that red and pink artificial fruits were more quickly removed than artificial fruits of other colours. Yanoviak et al. (2008) suggest that their paper presents the first case of fruit mimicry by an insect and one of the rare cases where model and mimic belong to different biological kingdoms.

However, key assumptions of mimicry such as the spatial and temporal overlap between ants and the model fruit species, Hyeronima alchorneoides, have not been quantified. Likewise, it has not been established whether ants and fruits are consumed by the same animals. It is thus unknown whether animals are duped into misidentifying infested ants as fruits. The latter is unlikely because the colour similarity between infested ants and the model red fruits is not strong. When analysed according to a model of avian vision (Vorobyev & Osorio 1998), the colour contrast (ΔS = 10.6) is large enough to indicate ease of discrimination for birds (see Siddiqi et al. 2004). Imperfect mimicry may flourish given that the penalties for misidentification are low, owing to the reduced defences of parasitized ants. However, mimicry is not consistent with data on avian behaviour in other experiments. Omnivorous and frugivorous birds distinguish between insects and fruits and adjust their behavioural responses to colour stimuli accordingly even if insects are immobile (Gamberale-Stille 2001; Gamberale-Stille & Tullberg 2001). Finally, discrimination between living insects and fruits is not only based on colour but also involves shape and movement. Thus, the contention that colour change in parasitized ants represents an example of fruit mimicry depends on a number of untested assumptions.

Rather than duping frugivores into misidentifying the ant’s gaster as a fruit, the colour change of parasitized ants may simply increase their conspicuousness to potential predators of ants. This alternative hypothesis requires none of the untested assumptions above. It does not impose temporal or spatial restriction – like the co-occurrence between model and mimic – on the effectiveness of transport of the parasites. Further, this hypothesis is not frequency-dependent on the relative number of models and mimics, whereas the effectiveness of fruit mimicry is frequency-dependent.

Seasonal variation in the availability of the model provides an easily manipulated setting to distinguish between mimicry and alternative hypotheses. If ant predation occurs when no red fruits are available, the fruit mimicry hypothesis would not be supported. More specifically, the mimicry hypothesis predicts that per-capita predation risk of infected ants would decline if their density increases, but would probably increase with increasing fruit density. Conversely, the conspicuousness hypothesis is consistent with predation risk of infected ants increasing with the number of infected ants (because predator learning about the resource would probably have a stronger effect than predator satiation), and being only very weakly affected by the density of red fruits.

Mimicry of fruits is perhaps more likely to occur in fruit-infesting species that have an obligate association with fruits as host organs. For example, larvae of the Lepidoptera Cydia pomonella develop commonly in apple fruits (Malus domestica) but some populations develop in walnut (Juglans regia). Interestingly, the hydrocarbon profile of larvae resembles those of their host plants (Piskorski, Trematerra & Dorn 2010). While mimicry is a plausible scenario, the acid test for mimicry would be to show that the similarity is adaptive, i.e. that predators of the larvae indeed cue on hydrocarbons to identify their prey rather than on some other cue.

In another recent paper, Lev-Yadun (2009) concludes that “Müllerian mimicry rings seem to be very common in plants”. In support of this, he suggests that the defensive structures of many thorny, spiky and prickly plants sport conspicuous coloration. He lists a number of putative mimicry rings in which a group of spiny plants co-occur in space and share similarity of appearance. Whilst mimicry is compatible with these observations, so is the alternative hypothesis of convergence discussed above. In particular, convergence may occur because of developmental or mechanical constraints on spines and thorns. Moreover, it has yet to be demonstrated that the coloration of the defences of any of these species influences the behaviour of potential herbivores (i.e. that coloration acts as an aposematic warning signal), let alone that such herbivores generalize across species.

Conclusion

  1. Top of page
  2. Summary
  3. Mimicry in plants
  4. Olfactory mimicry in carnivorous plants
  5. Other examples of olfactory mimicry
  6. Visual mimicry
  7. Conclusion
  8. Acknowledgement
  9. References

We think it is not advisable to interpret similarity between distinct species or plant organs as mimicry as long as plausible alternative explanations have not been considered and excluded. In particular, alternative hypotheses often require fewer restrictive conditions and are thus likely to be easier to evolve. In this article we have discussed studies on very different putative mimicry systems to show that the lack of eliminating alternative hypotheses is a common limitation in the current field on plant mimicry. We do not argue that mimicry is implausible, neither generally in plants nor in the specific systems where it has been proposed. Rather, we would like to encourage researchers to more clearly focus on those predictions that distinguish mimicry from alternative hypotheses. We feel that such a focus would allow the field to progress more quickly.

Critically, demonstrations of mimicry must focus on the misidentification of a mimic as a specific rewarding model that is at the heart of mimicry. We suggest that researchers interested in the adaptations of plants on the sensory systems of animals take care to consider convergent evolution and exploitation of sensory biases as alternative explanations to mimicry for similarity between aspects of different plant species. Care must also be taken in the definition of these terms, and we recommend the definitions used in Schaefer & Ruxton (2009). No matter which definitions a researcher uses, it would help readers if they give this definition in full or point the reader to a previous definition in the literature. More generally, researchers must guard against prematurely accepting intuitively appealing explanations of mimicry that may hinder the search for the true mechanisms underlying the evolution of some fascinating insect–plant interactions. Such an approach should improve our understanding of the selection pressures that sensory and cognitive systems of animals place on the plants with which they interact.

Acknowledgement

  1. Top of page
  2. Summary
  3. Mimicry in plants
  4. Olfactory mimicry in carnivorous plants
  5. Other examples of olfactory mimicry
  6. Visual mimicry
  7. Conclusion
  8. Acknowledgement
  9. References

We thank an anonymous referee, Laurence Gaume, Bruno Di Giusto and the Handling Editor for useful comments on a previous version.

References

  1. Top of page
  2. Summary
  3. Mimicry in plants
  4. Olfactory mimicry in carnivorous plants
  5. Other examples of olfactory mimicry
  6. Visual mimicry
  7. Conclusion
  8. Acknowledgement
  9. References