The effect of wasps and ants on the reproductive success of the extrafloral nectaried plant Turnera ulmifolia (Turneraceae)


  • M. Cuautle,

    Corresponding author
    1. Departamento de Ecología Vegetal, Instituto de Ecología, A.C., Apo. 63, Xalapa, Veracruz 91070, México
      can be addressed to both authors. E-mail:
    Search for more papers by this author
  • V. Rico-Gray

    Corresponding author
    1. Departamento de Ecología Vegetal, Instituto de Ecología, A.C., Apo. 63, Xalapa, Veracruz 91070, México
      can be addressed to both authors. E-mail:
    Search for more papers by this author

can be addressed to both authors. E-mail:


  • 1. To assess the effectiveness of extrafloral nectaries (EFN) as a defensive mechanism of plants it is vital to use a multiple interactions approach and assess the contribution of all visiting species and their interactions.

  • 2. The effect of EFN-visiting ants (Camponotus planatus, C. abdominalis, Conomyrma sp., Crematogaster brevispinosa, Forelius sp., Pseudomyrmex sp.) and wasps (Polistes instabilis, Polybia occidentalis) on the reproductive success (estimated as the number of reproductive structures) of Turnera ulmifolia (Turneraceae) was experimentally evaluated. Herbivory effects were tested using Euptoieta hegesia larvae (caterpillars), which is the main herbivore of this plant. The study was done in a coastal sand dune scrub in Veracruz, México.

  • 3. Wasps and ants were selectively excluded using a two-factor design (Wasps, Ants) block design, both factors with two levels (absent, present). The response variables were an index of herbivory per branch and the number of buds, flowers, ripe and unripe fruit per plant, and the seed/fruit ratio per branch.

  • 4. After a week of placing the larvae on experimental plants, they were significantly more frequent on plants where wasps and ants had been experimentally excluded.

  • 5. Wasp presence was associated significantly with greater numbers of buds, flowers, ripe fruit and seeds. When acting separately, wasps and ants exerted a positive effect in decreasing herbivory levels and increasing the number of unripe fruit; when acting together, however, their effect was not additive.

  • 6. This is the first demonstration of a positive effect on the plant by wasps associated with EFN. The ecological implication of this finding is that the function of EFN and the ultimate effects on a plant will probably depend on the array of organisms visiting its EFN.


Extrafloral nectaries (EFN) are plant secretory glands usually found on leaves (blade, petiole), but also on reproductive structures (e.g. inflorescence spike, sepals, buds, fruit) (Bentley 1977). They are common structures in many plant species, and attract ants because they represent a predictable food source (Rico-Gray & Sternberg 1991; Whitman 1994). Many EFN-bearing plants are protected by nectar-foraging ants (e.g. Rico-Gray & Thien 1989a; Koptur, Rico-Gray & Palacios-Rios 1998a; Torres-Hernández et al. 2000). Moreover, it has been suggested that EFN evolved as a generalist plant defence, attracting ants that in turn will repel or remove herbivores (Koptur 1991; but see Becerra & Venable 1989). However, little is known about the evolution of ant–plant interactions (Beattie 1985; Pemberton 1992): the only fossil evidence of an EFN is for the extinct Populus crassa (Lesquereux) Cockerell (Pemberton 1992). This species was found in the 35-million-year-old Florissant Formation in Colorado, USA, where, besides ants, there was fossil evidence for Coleoptera, Diptera and other Hymenoptera, as well as spiders (Pemberton 1992). Similarly, extant ants are not the sole visitors to EFN; other arthropods actively forage for the nectar produced by these structures (e.g. spiders, wasps, bees, beetles, flies, mites) (Hespenheide 1985; Koptur 1985; Pemberton 1993; Pemberton & Vandenberg 1993; Pemberton & Lee 1996; Ruhren & Handel 1999; Torres-Hernández et al. 2000). Furthermore, the potential protective role of these nectar foragers has been questioned (Koptur 1985; Pemberton & Lee 1996), but their joint effect with ants on plant defence has not been directly addressed or evaluated. The study of systems that involve more than two interacting species (i.e. multiple interactions) is vital, because plants rarely interact with only one other species and the outcome is not usually additive (Juenger & Bergelson 1998; Herrera 2000). To evaluate the effectiveness of EFN as a defensive mechanism, the contribution of all visiting species should be considered, as well as their interactions (Pemberton et al. 1996).

Ant–plant associations in tropical sand dunes are abundant relative to their frequency in temperate semiarid or humid mountain sites. However, research on plant defence by ants in tropical sand dunes is scarce and mainly restricted to four systems: Myrmecophylla christinae (Orchidaceae) (Rico-Gray & Thien 1989a,b), Paullinia fuscescens (Sapindaceae) (Rico-Gray & Castro 1996), Opuntia stricta (Cactaceae) (Oliveira et al. 1999) and Turnera ulmifolia (Turneraceae) (Torres-Hernández et al. 2000), which, however, represent a wide range of mutualistic interactions between ants and plants (Rico-Gray et al. 2003). A previous study showed that T. ulmifolia individuals associated with Camponotus planatus Roger and C. abdominalis (Fabricius), the largest ants visiting this plant, produced more fruit (Torres-Hernández et al. 2000). Interestingly, however, plants where ants had been excluded produced more fruit than plants with ant species other than C. planatus and C. abdominalis. It was hypothesized that ant exclusion allowed visits by wasps [Polistes instabilis (Olivier) and Polybia occidentalis (Saussure)]. Some authors have suggested that wasps are important in plant protection, owing to their predator and parasitoid activity (Hespenheide 1985; Domínguez, Dirzo & Bullock 1989; Pemberton et al. 1996). Domínguez et al. (1989) showed that wasps that feed on flower nectar defend plants, and can reduce the number of herbivores. However, the effect of wasps associated with EFN has not been quantified. The objective of this work was to evaluate the effect of EFN-visiting wasps and ants (jointly or individually) on the reproductive success of T. ulmifolia, in a coastal sand dune scrub in Veracruz, Mexico. Two questions were addressed: (1) Do wasps have a beneficial effect on the reproductive success of T. ulmifolia, similar to that reported by ants? (2) Is there a significant interaction (non-additivity) between the wasp and the ant?

Study site and methods

The study was done in the coastal sand dune scrub at Centro de Investigaciones Costeras La Mancha (CICOLMA), located along the coast of the state of Veracruz, Mexico (19°36′N, 96°22′W; elevation < 100 m). The climate is warm subhumid, a rainy season occurs between June and September, total annual precipitation is 1100–1500 mm, mean annual temperature is 24–26 °C and minimum temperature is 15 °C (Moreno-Casasola et al. 1982).

the plant and study system

Turnera ulmifolia L. (Turneraceae) is a polymorphic polyploid complex of herbaceous, perennial weeds, bearing extrafloral nectaries, and native throughout much of the neotropics (Gama, Narave & Moreno 1985; Barrett & Shore 1987). Turnera ulmifolia inhabits a variety of vegetation associations, exhibiting two contrasting patterns of floral morphology, where populations are either dimorphic or monomorphic for a range of floral traits (e.g. style length, stamen height, pollen size; Barrett & Shore 1987). At the study site T. ulmifolia grows on the semistabilized and stabilized sand dunes, is monomorphic, self-compatible with long styles and a range of stamen heights. It flowers and fruits all year round, with a peak during the summer (rainy season) (Torres-Hernández et al. 2000). Branches grow continuously from an apical meristem, producing leaves regularly. Flowers are axillary, and one to three flowers open per day (Gama et al. 1985; Torres-Hernández et al. 2000). Flowers last less than a day, and the associated leaf remains throughout fruit development, which lasts 2–3 weeks. EFN are located at both sides of the petiole, close to the insertion of the floral pedicel. Ants (Camponotus planatus, C. abdominalis, Conomyrma sp., Crematogaster brevispinosa Mayr, Forelius sp., Pseudomyrmex sp.), wasps (Polistes instabilis and Polybia occidentalis) and honey bees (Apis mellifera) forage for the nectar produced by the EFN (M. Cuautle and V. Rico-Gray, personal observation). The main leaf herbivore is a caterpillar (Euptoieta hegesia Cramer, Lepidoptera: Nymphalidae), which is most active from June to August, although it can be found year round. The maximum number of E. hegesia larvae per plant in the study site was three; larvae/plant density during the study was 0·084 (n = 71). Larval and pupal development takes 2 and 1 week, respectively. Larvae can reach 50 mm in length and remove a significant amount of foliar tissue; for example, three larvae can totally defoliate a T. ulmifolia individual (Schappert & Shore 1998).

wasp and ant exclusion

To test the hypothesis that wasp activity increases the reproductive success of T. ulmifolia (estimated as greater number of reproductive structures), and to establish similarities and/or differences between the effect of ants and wasps, we selectively excluded ants and wasps. Owing to differences in exposure to wind and salt spray, and distance to the water table for T. ulmifolia individuals, we used a two-factor [wasps(W), ants(A)] block design, both factors with two levels (absent = 0, present = 1). The four treatments (W0A0 = without wasps and ants; W0A1 = without wasps, with ants; W1A0 = with wasps, without ants; W1A1 = with wasps and ants) were randomly assigned within a block (four plants/block). The number of blocks were 24 (n = 24 plant per treatment). The response variables were an index of herbivory (IH) per branch and the number of buds, flowers, ripe and unripe fruits per plant, and the seed/fruit ratio per branch. Ant and wasp species were not specifically selected within the blocks; the experiments were done with ants and wasps naturally visiting the EFN.

Field work was done during the rainy season (May–September 2000), when herbivores are more active, establishing and sampling two to three blocks per week, until all 24 blocks were established and sampled. To be sure that herbivores were present, we placed one caterpillar of E. hegesia (3–6 mm in length) on each plant within a block. Wasps were excluded by completely covering plants with a cloth/wire cage (0·70 × 0·70 × 0·75 – 1·50 m3), which, however, was not fixed to the soil to allow freedom of movements to the caterpillars. Ants were excluded from plants by applying a band of tree tanglefoot® (The Tanglefoot Co., Jackson, MS) on mature plant tissue at the base of branches. The numbers of buds, flowers, ripe and unripe fruit (fruits were marked after counting), and seeds were counted at five intervals: t0, prior to assigning treatments to plants; t1 and t2 were, respectively, 2 and 3 weeks after treatments were assigned (E. hegesia larvae were present and consuming foliar and floral tissues); t3 was at week four (usually larvae had pupated) and exclosures were then removed; and t4 was at week six.

pollinator exclusion

One of the main reasons why the role of other EFN-visitors has been underestimated is the difficulty implied in the experimental design: wasp exclusion also excludes potential pollinators and certain herbivores (e.g. ovipositing female butterflies). The fact that T. ulmifolia is self-compatible and sets fruit autogamously (without pollinators) allows fruit production in the absence of pollinators. However, a decrease in fruit production due to an endogamic effect could not be disregarded. Thus, using a pollinator exclusion experiment, we evaluated the effect of pollinators fruit and seed production. We marked 25 T. ulmifolia individuals (different from those used above) and selected four branches per plant, two were bagged and two were left untouched (free access to pollinators); the response variables were the number of ripe and unripe fruits and the seed/fruit ratio. Counts were done at the beginning of the experiment, and after 2, 4 (branches were then unbagged) and 6 weeks. For the analysis, we considered data only from the counts done at the second and fourth week (when the pollinators were excluded).

Data for the two time periods were pooled, transformed [y = v(x + 0·5)] for ripe and unripe fruit, and analysed using nested anovas. The number of unripe fruits was not significantly different between bagged and unbagged branches (mean ± SD = 1·69 ± 0·49 vs 1·83 ± 0·49, respectively; F = 2·16, df = 1,50, P = 0·148, N = 100), whereas the number of ripe fruits (mean ± SD = 1·42 ± 1·29 vs 2·46 ± 1·48, respectively; F = 16·9, df = 1,50, P < 0·001, N = 100) and the seed/fruit ratio (mean ± SD = 25·6 ± 15·6 vs 41·9 ± 11·2, respectively; F = 23·0, df = 1,30, P < 0·001, N = 77) were significantly different between treatments. Based on the latter, we only used the t4 data when analysing fruit number and the seed/fruit ratio. As mentioned above, treatments were removed at the end of t3, allowing visits by pollinators. Because fruits were marked, it was possible to determine which fruits were formed after exclusions were removed. If we recorded a herbivory event during at t4, we still expected its influence on the number of ripe fruit and seeds.

statistical analyses

Bud, flower and ripe and unripe fruits were counted for each of five marked branches per plant. Counts per branch were pooled per plant. The mean number of branches for the plants used in the experiment was 10·2 (SD = 5·86, n = 92 plants). Bud, flower and ripe and unripe fruit data were transformed [y = v(x + 0·5)] for normality (Zar 1996); seed/fruit ratios were not transformed. Bud, flower and unripe fruit data were analysed using manovas because they do not need to fulfil the sphericity assumption (Zar 1996), whereas, ripe fruits and the seed/fruit ratio were analysed using anovas. The homoscedasticity assumption was verified using a graphic analysis of the residuals (fitted value vs residual) (Crawley 1993), and showed better results when data for the five branches were pooled, although for certain times the variables showed an undefined pattern. Despite the latter, the consistency of the results when analyses were done either using untransformed, transformed or ranked data gave us sufficient confidence to compute the manovas using transformed data.

We estimated herbivory using an index of herbivory per branch, IH = Σni(i)/N, where ni is the number of leaves in the category i, and N is the total number of leaves sampled (Domínguez et al. 1989). Using the same T. ulmifolia individuals and branches as above, we sampled 10 leaves per branch, and estimated the percentage of leaf area removed by herbivores using the following categories: 0, no damage; 1, 1–5% of leaf area removed; 2, 6–12%; 3, 13–25%; 4, 26–50%; 5, 51–100% of leaf area removed. The response variable was the difference between the IH per branch, at the beginning of the experiment and that 2 weeks later. These data were ranked (rank type 1: whole data set ranked from the smallest to the largest) to fulfil the normality and variance homogeneity assumptions, and analysed using a two way-anova. This procedure using the F distribution as approximation is similar in power and robustness to the Friedman test, and allowed us to test the effects of the interaction (Conover & Iman 1981). All figures were prepared using untransformed data. All computations were done using Systat (1993).


One week after E. hegesia larvae had been placed on experimental plants, they were significantly more frequent on plants where wasps and ants had been experimentally excluded (Table 1). In some plants where wasps and/or ants had been excluded, more than one larva was found. These emerged from eggs laid before the experiment began. Predation by wasps was observed only on early larval instars. Whenever larvae were larger, they escaped predation.

Table 1.  Frequency of E. hegesia larvae on experimental plants, 1 week after establishing the treatments. All plants received one caterpillar each at the beginning of the experiment. Larval presence was not independent of the treatment (Fisher exact test, P < 0·001; W= wasp, A= ant, 0 = absent, 1 = presence); n = 24 plants per treatment. W0A0= without wasps or ants; W0A1= without wasps, with ants; W1A0= with wasps, without ants; W1A1= with wasps and ants
Number of plants with larvae16754
Mean ± SD larvae per plant1.38 ± 2.080.33 ± 0.640.33 ± 0.640.42 ± 0.83

Herbivory (IH) was greater in branches from plants where wasps and ants had been excluded than in those from plants with wasps, with ants or both (Tukey test, P < 0·001). Wasps and ants acting separately produced a similar decrease in herbivory (IH) compared with the control (W0H0). However, their effect when acting together was not additive (Fig. 1). This was caused by the significant interaction between wasps and ants (Table 2), indicating that the signs and/or magnitude of the effect of either wasps or ants on the IH depended on the other: wasps reduced the IH, only when ants were absent, and vice versa.

Figure 1.

Significant Wasp*Ant interaction on the index of herbivory (IH).

Table 2.  Results of the anova testing for the effects of wasps and ants, made with ranked data, on the index of herbivory per branch (IH) (n = 118, 113, 119 and 112 branches for W0A0, W0A1, W1A0 and W1A1, respectively) (see text, Study site and Methods)
Wasp10·911,435   0·001
Ant10·991,435   0·001

More buds and flowers per plant were significantly associated with wasp presence (Table 3; Fig. 2a and b). The effect of time was significant, and the Time*Wasp interaction was significant for the number of buds (Table 3). Plants did not differ in mean number of buds and flowers at the beginning of the experiment, i.e. before applying treatments, t0 (Fig. 2a and b); whereas significant differences among treatments were observed for buds, at t1 and t2 (Fig. 2a), and for flowers at t2 (Fig. 2b). Differences were lost close to the end of the experiment.

Table 3.  Results of the manovas and anovas for the effects of wasps (W) and ants (A), on buds, flowers, unripe and ripe fruits per plant (n = 24 plants per treatment), and seed/fruit ratio per branch (n = 68, 78, 88 and 96 branches for W0A0, W0A1, W1A0, W1A1, respectively). Only t4 was used for ripe fruit and the seed/fruit ratio (NS = P > 0·051). T*A and T*W*A were not significant for any factor (0 = absent, 1 = presence)
 BudsPFlowersUnripe fruitRipe fruitSeed/fruit
Block 8·11523,69<0·0012·73423,690·001 3·21423,69<0·001 3·32523,69<0·0014·18323,69<0·001
Wasp 7·38 1,69   0·0085·70 1,690·020 6·17 1,69   0·01522·95 1,69<0·0015·0013,03   0·026
Ant  NS  NS 6·98 1,69   0·010  NS  NS
W*A  NS  NS 3·93 1,69   0·051  NS  NS
Time16·23 4,66<0·0013·68 4,660·00931·28 4,66<0·001    
T*W 0·83 4,66   0·016  NS  NS    
Figure 2.

Mean number of buds (a), flowers (b) and unripe fruits per plant (c) for the four treatments by sampling time: without wasps and ants (W0A0); without wasps, with ants (W0A1); with wasps, without ants (W1A0); with wasps and ants (W1A1). Counts were performed at five different times (t0t4). At the beginning of the experiment one E. hegesia caterpillar was added (t0). Differences in treatments were significant at P < 0·05 for wasps (*), ants (+) or their interaction (*/+), for the sampling time indicated (Hypothesis Test).

For unripe fruits, the Wasp*Ant interaction was significant (Table 3), and the same tendency was maintained for IH: the presence of wasps increased unripe fruit production, but had no effect when ants were present (Fig. 3). These results reflect field observations, namely that (i) C. planatus prevents visits by other insects to EFN and (ii) P. occidentalis flys among EFN until it finds an unoccupied one or one with nectar. The effect of time was also significant: plants did not differ in mean number of unripe fruit per plant at the beginning of the experiment (Fig. 2c), whereas significant differences among treatments were observed in the rest of the times (except at t3).

Figure 3.

Mean numbers of ripe fruit per plant (a) and seed/fruit ratio per branch (b) of the four treatments at time t4: without wasps and ants (W0A0); without wasps, with ants (W0A1); with wasps, without ants (W1A0); with wasps and ants (W1A1). Treatments with the same letter were not significantly different at P < 0·05 (Tukey test).

A greater number of ripe fruits per plant and seed/fruit ratio per branch, was also significant associated with wasp presence (Table 3). The ant factor was almost significantly associated with a greater number of ripe fruit (F = 3·83, df = 1,69, P = 0·054). For ripe fruit, 1·8 more fruits were produced in those treatments with wasps (W1A0, W1A1), than in those with none (W0A0). Treatment with just ants (W0A1), exhibited an intermediate production between the above two groups (Fig. 4a). For seeds, 1·2 more seed/fruit were produced in branches from the treatment with wasps, without ants (W1A0), that in the treatment without wasps, with ants (W0A1) (Fig. 4b).

Figure 4.

Significant Wasp*Ant interaction on the number of unripe fruit per plant.


Our results show that ant presence (when wasps were absent) had a positive and significant effect on the plant, reflected as a decrease in herbivory and an increase in the production of unripe fruit. This supports similar results by other authors (e.g. Rico-Gray & Thien 1989a; Koptur et al. 1998a; Torres-Hernández et al. 2000). Predation by ants was never observed on larvae or eggs of E. hegesia. However, the patrolling activity of the ants could delay the foraging activity of the larvae. Wasps also had a positive and significant effect on the production of reproductive structures by T. ulmifolia. This is the first demonstration of a positive effect for the plant by wasps associated with EFN. The effect on plant herbivores of non-ant EFN visitors had been evaluated only indirectly (Stephenson 1982; Hespenheide 1985; Koptur 1985; Pemberton 1993; Pemberton & Vandenberg 1993, 1996; Torres-Hernández et al. 2000). For example, jumping spiders associated with the EFN of Chamaecrista nictitans exerted a positive effect on the plant (Ruhren & Handel 1999). Ants, however, the most common EFN visitors, were not included in these experiments.

Interestingly, however, when acting together, the effect of wasps and ants on T. ulmifolia was not additive. An antagonistic interaction between parasitoid wasps and ant-tending honeydew-producing insects has been reported (Pierce & Mead 1981; Gibernau & Dejean 2001), and, in the case of EFN, the role of non-ant visitors may be limited by an increase in ant abundance (Koptur 1985; Pemberton et al. 1996). These results exemplify the importance of studying multiple interactions, rather than a pairwise interaction approach, which will rarely reflect the complexity of natural systems and the array of outcomes possible in nature. If non-additivity occurs frequently, then outcomes of single interaction studies will be strongly context-dependent (Herrera 2000).

Depending on the presence or absence of the wasps, ants had an effect only on herbivory and on the number of unripe fruits produced. Oliveira (1997) also found that ants increased initial fruit production (juvenile fruits) in Caryocar brasiliense, but had no effect on mature fruit production. He suggested that the ant might provide other kinds of benefits to the plant, i.e. increase pollen donation or selective fruit abortion. However, we cannot reject the possibility of a positive effect on ripe fruits and the seed/fruit ratio because we were able to use data only from the last time period sampled (t4). For variables evaluated at all the time periods, only wasps had a positive effect on the number of buds and flowers. This suggests that, despite the effect of the ant–wasp interference found for herbivory and unripe fruit, this interference could decrease in other reproductive stages if both participants exerted a function in this system. Similar to other species, it is possible that T. ulmifola could produce different types of nectar depending on the different requirements of its visitors, either induced by herbivory or synchronized with the degree of vulnerability and/or activity of its herbivores (Stephenson 1982; Koptur 1989; Smith, Lanza & Smith 1990; Wäeckers et al. 2001). Furthermore, ants and parasitoid wasps can discriminate among different sugars (Koptur & Truong 1998b; Wäeckers 1999). EFN in T. ulmifolia are active between bud development and fruit maturation; however, we did not measure the quantity, quality or variation in nectar. We suggest that nectar volume could increase from bud to fruit development, with buds being less visited by ants, allowing more visits by wasps. An accumulation in secretion over time is implied, because when ants were excluded from EFN during fruit development, nectar runs down the branch and is colonized by fungi (Torres-Hernández et al. 2000). It is also possible that herbivory by caterpillars may increase EF-nectar production and/or even produce leaf volatiles to attract more wasps/ants (F.L. Wäeckers, personal communication), i.e. an induced response. As mentioned above, changes in EFN production induced by herbivory have been reported, and induced responses to herbivory may be common in ant–plant systems (Agrawal & Rutter 1998). These changes in plants that provoke an increase in ant-patrolling, might be detected by wasps and produce a corresponding response from them.

Both of the EFN-visiting wasps that we commonly observed (P. instabilis and P. occidentalis) are efficient caterpillar predators (Domínguez et al. 1989; Raveret Richter 2000), particularly because they learn quickly and return repeatedly to sites where they have been successful (Stamp & Bowers 1988). The wasps prey on caterpillars directly or force them to abandon the plant. Both mechanisms could help to regulate the activities of E. hegesia on T. ulmifolia. The lack of caterpillars on plants where wasps were not excluded was mostly caused by the indirect effect of wasps, which is similar to that reported for ant–guard systems (Messina 1981).

In summary, we have demonstrated that wasps play a role that had been almost exclusively ‘reserved’ for ants (i.e. EFN visitors and/or plant defence), and that the study of the role of other EFN visitors is essential to understand multiple interaction systems.


We thank C. Díaz-Castelazo, E. Priego, D. Hernández-Conrique and M. Ferrer-Ortega for their help during field work, and V. Parra-Tabla, S. Koptur, P. S. Oliveira, C. M. Herrera, J. Maloof, and a anonymous referee for their comments and suggestions to the manuscript. Wasps were identified by Alicia Palafox (UNAM). Research was supported by Instituto de Ecología, A. C. (902–16) and CONACYT (118945 to MC).