Conditional outcomes of interactions: the pollinator-prey conflict of an insectivorous plant

Regino Zamora

INTRODUCTION

The outcomes of interspecific interactions are affected in predictable ways by current ecological conditions (Bronstein 1994). Several biotic factors, such as age and size of individuals, density and spatiotemporal structure of populations, and the general community context, have been outlined as determining the nature of interactions (Thompson 1988, 1994). Moreover, abiotic conditions can also play a major role in determining the outcome of interactions (Dunson and Travis 1991). In this respect, site-specific effects in plant-insect relationships may be particularly important, because the physical environment governs the ecophysiology of both plants, due to their sessile lifestyle, and insects, due to their small size. Therefore, the ecological and evolutionary consequences of plant-insect interactions can vary between contrasting environments, even at a restricted spatial scale. Evidence supporting this possibility has recently been provided by Herrera (1995b, 1997) for plant-pollinator systems, and by Zamora (1995) for plant-prey interactions of insectivorous plants. A further possibility is that the full ranges of outcomes of these plant-animal relationships are conditional, and that the degree of interference between mutualistic and antagonistic systems depends on the abiotic characteristics surrounding the interaction. I have tested this hypothesis for an insectivorous plant, which faces possible antagonism in the diverse relationships between plant and pollinators, plant and prey, plant and herbivore and/or plant and kleptoparasite.

Despite the fact that plants experience, simultaneously or sequentially, mutualistic and antagonic interactions with animals, few studies have examined the combined effects of these opposing interactions on plant evolutionary ecology (see Strauss and Armbruster 1997 for a recent review). For example, studies commonly assume that pollinators behave exclusively as plant mutualists. However, pollinators may act also as floral herbivores (Faegri and van der Pijl 1979), seed predators (Pellmyr 1989), or as vectors of plant disease (Jennersten 1988). In this context, the interplay of plant-pollinator and plant-prey interactions has seldom been explored in the ecological literature, and only anecdotal references are available (Juniper et al. 1989). Here, I analyze the effect of local physical conditions on the possible conflict between plant-pollinator and plant-prey systems.

Ideally, an insectivorous plant would, independently, attract prey to traps and pollinators to flowers. However, many traps use traits that also attract pollinators, such as color, scent, nectar, and UV guides (Joel et al. 1985), thereby blurring the distinction between prey and pollinator attraction. This conflict of interests could be alleviated by spatial or temporal separation of traps and flowers. For example, the widespread aquatic genus Utricularia sets traps underwater, but flowers above the water surface. In most terrestrial carnivorous plants, the flowering stalk extends high above the trapping devices, separating flowers from traps (Givnish 1989, Juniper et al. 1989). This solution applies to most species of Pinguicula (Lentibulariaceae), in which flowers stand well above the rosette of leaves. However, in P. vallisneriifolia Webb, an endemic insectivorous plant of wet rocky habitats and cliffs in southeastern Spain, the flowers not only open near the adhesive leaves, but emerge as the leaves sprout. Thus, the possible pollinators of P. vallisneriifolia are also its potential prey.

Pinguicula vallisneriifolia produces two types of glandular leaves. The first 5-7 leaves form a rosette tight against the rock wall; these are followed by erect, distal leaves. A typical reproductive plant produces 14 spurred flowers at the top of leafless, 8-18 cm stalks (Zamora et al. 1996). The 10-30 cm long distal leaves occupy the same spatial zone as the flowers, both overhanging the rock wall on which the plant grows. The blooming period (May-June) precedes the period of maximum prey capture (July). However, a typical flowering individual has 2-4 fully functional distal leaves. Thus, the spatial and temporal coincidence of flowers and the first distal leaves creates the potential for interference between the plant's pollination and insectivory systems. This carnivorous Mediterranean species inhabits a broad range of irradiance regimes, so that the variability of interactions can be assessed in contrasting abiotic scenarios. Indeed, local abiotic conditions govern prey capture of P. vallisneriifolia, because both the abundance of flying insects and the retention capacity of leaf mucilage varies between habitats (Zamora 1995, Zamora et al. 1998).

In this paper, I assess possible conflicts between pollination and insectivory for P. vallisneriifolia under contrasting scenarios, and their consequences for the plant's reproductive success. To explore these issues, I first carried out a large-scale study to identify the pollinator and prey assemblages over the entire geographical range of P. vallisneriifolia. Later, I performed an intensive local study, combining observations and field experiments, to investigate the possible mechanisms of interference between the plant-pollinator and plant-prey systems. Specifically, I analyzed: (1) the plant breeding system and pollination ecology of this carnivorous plant; (2) the spatial distribution of pollinators and prey in relation to microclimatic conditions; (3) the identity and capture rate of insects trapped on leaves during the flowering period, in relation to the retention capacity of the mucilage; and (4) whether flowering plants captured more or different prey than did nonflowering ones.

METHODS

I sampled pollinators in 11 populations distributed broadly throughout the range of Pinguicula vallisneriifolia in the Sierras de Cazorla, Segura y Las Villas, Spain, an orographically complex area with mountainous escarpments and narrow valleys. All populations grew on cliffs (5-80 m in height) with varying sun exposure. For each population, I recorded floral visitors for 15-min periods evenly distributed over the day (8 h of observation per population) during the peak of blooming (May-June) of 1992. These observations focused on a sector of the cliff representative of the predominant irradiance conditions in the population. Population size ranged from 150 to 600 flowering plants. All observations were conducted on cloudless days. Any insect seen moving among the flowers and making contact with the anthers and stigma was considered a likely pollinator. Some insects (e.g., Diptera) that landed and rested on the lower lip of the corolla were not considered pollinators because they did not touch the plant's sexual organs. In addition, I noted the presence/absence of small flower visitors (e.g., thrips) in the perianth of 40-60 flowers (individuals) per population. I later visited the 11 populations sampled for flower visitors during fruiting to record the percentage of flowers setting fruit, and to identify and count prey on a sample of 40-60 distal leaves per population (one leaf per reproductive plant), noting especially any floral visitors trapped on the leaves.

After collecting this broad-scale information, I made an intensive study in a population situated in the center of the geographical range of P. vallisneriifolia. This study site lies at the headwater of a spring surrounded by 50-m high cliffs, and harbors one of the largest populations of this endemic species (see Zamora 1995). To quantify fruit set and seed set under natural pollination conditions, I tagged flowering buds on different reproductive individuals in three sections of the vertical cliff. Although these sites were very close to each other, they received different irradiance levels and constituted three distinct habitats: (1) sunlight habitat, a northeast-facing section of the cliff receiving 2 h of direct morning sunlight; (2) shade habitat, a north-facing sector of the cliff that receives no direct sunlight and is intermediate in the irradiance gradient; and (3) deep shade habitat, a small cave situated at the bottom of the north-facing wall, the shadiest environment. The three habitats were located at the same altitude, with the shade and deep-shade habitats [approximately]10 m from each other and both [approximately]30 m from the sunlight habitat. I quantified several abiotic variables to characterize the environment of the three habitats. Total radiation was measured using a Li-Cor LI-200 sz pyranometer sensor connected to a Li-1000 data logger (Li-Cor, Lincoln, Nebraska, USA). Air temperature and air humidity were taken using a Rotronic YA-100 (Rotronic ag, Zurich, Switzerland) combined sensor. Readings were collected simultaneously in the three habitats, consistently using the same recording point (the center of the sample zone in each habitat). Sensors were placed near the flowers of P. vallisneriifolia.

To test pollen limitation under natural pollination conditions, I tagged flowering buds in the three habitats. After anthesis, I hand-pollinated each flower twice, with pollen from three different donors. To determine self-fertility, I covered flower buds with cellophane bags attached to the peduncle, excluding all pollinators. I then assigned these flowers to one of three pollination treatments: (1) spontaneous autogamy in the absence of insects; (2) self-pollination, and (3) cross-pollination with pollen from other plants in the same population. Each treatment involved 30 flowers (30 plants) in the shade habitat during 1994. I harvested mature capsules prior to dehiscence, counted the seeds and aborted ovules, and weighed all seeds produced by a capsule as a group.

As with the geographic survey, I sampled flower visitors during 15-min periods distributed evenly during the day over the entire flowering period (15 May-15 June 1992) in the three habitats, where plants flowered simultaneously. In the sunlight habitat, I noted whether plants received direct sunlight during the observation period. In each period, I noted all floral visitors to plants growing in 10 [m.sup.2] of cliff per habitat, and with similar flowering plant density. In the same quadrats, I also periodically counted small insects in flowers (for a total of 50 flowers on 50 plants per habitat) from the beginning to the end of blooming. Nocturnal and crepuscular observations of flower visitors were also made over the flowering period.

To test experimentally the potential role of thrips as pollinators, in 1994 I caged two groups of plants, with 1.5-mm mesh tents attached to the rock wall. This mesh size excluded all flying floral visitors except thrips. I tagged 30 caged buds belonging to 30 different reproductive individuals in the sunlight habitat, another 20 different flowering reproductive individuals in the sunlight habitat, and another 20 flowering buds in the shade habitat. Additionally, 20 caged floral buds were emasculated in the sunlight habitat to test cross-pollination by thrips. Further checks confirmed that the thrips were able to pass through the mesh, whereas the other larger species of floral visitors, including small beetles, were excluded.

To quantify insect captures by the leaves, I sampled 30 distal leaves per habitat belonging to different flowering plants at the end of blooming in 1994 and again in 1995. Later, in the laboratory, I measured leaf area and identified and counted any trapped insects.

To test specifically whether plants in bloom captured more prey than did nonflowering plants, I cut all flowering buds from 30 tagged reproductive plants in the sunlight habitat at the onset of blooming in 1995. After the end of blooming, I harvested the longest functional distal leaf from each of these tagged plants for later prey identification and determination of leaf area. In addition, I harvested other functional leaves from 30 nonreproductive plants (one leaf per plant) in the same habitat. The capture rate of distal leaves of flowering plants in the sunlight habitat was used as a control for comparing prey capture.

Statistical analyses were performed using the computer software JMP 3.1.5 (SAS Institute 1995) for Macintosh. ANOVAs were performed using type III sums of squares. Prior to all parametric analyses, I arcsine-transformed the ratio variables and log-transformed the remaining variables (Zar 1996). Throughout this paper, data are expressed as means [+ or -] 1 SE.

RESULTS

Large-scale study

During 88 h of observation, 36 floral visitors were observed at flowers of Pinguicula vallisneriifolia in 11 populations. The presence of medium-to-large flower visitors varied significantly with insolation, being almost absent from the shadiest populations ([r.sub.s] = 0.697, P [less than] 0.05, df = 10; [ILLUSTRATION FOR FIGURE 1A OMITTED]). Furthermore, even in the sunniest populations, the presence of flower visitors was restricted to the period of sunlight on the plants. In fact, censuses grouped by insolation regime instead of by population (sunlight on flowers, presence vs. absence), revealed that 32 (89%) floral visitors were observed during the 15 h (17% of the total time) of observation when the plants were in sunflecks (12 Bombylius sp., eight Lasioglossum sp., five Anthophora sp., three Halictus sp., three Bombus terrestris, and one Macroglossum stellarum). By contrast, during the remaining 65 h of observation, when no direct sunlight reached the plants, I observed only three Bombus terrestris and one unidentified small bee visiting the flowers. Fruiting success was also positively associated with the amount of insolation ([r.sub.s] = 0.783, P [less than] 0.025, df = 10); that is, the shadiest populations had the lowest fruiting success, probably because of the scarcity of pollinators [ILLUSTRATION FOR FIGURE 1B OMITTED].

Unlike the medium-to-large floral visitors, thrips (Taeniothrips meridionalis Priesner, Thysanoptera) and beetles (Eusphalerum scribae Schlaufuss, Staphylinidae) appeared at flowers in all but one population [ILLUSTRATION FOR FIGURE 1C, D OMITTED]. Taeniothrips meridionalis is a common visitor of many other Mediterranean plant species (Bournier 1983). In some sunny populations, thrips appeared in most flowers [ILLUSTRATION FOR FIGURE 1C OMITTED]. However, the presence of thrips and beetles did not vary significantly with insolation (thrips: [r.sub.s] = 0.322, NS; beetles: [r.sub.s] = -0.196, NS).

Prey counts indicated that the same species of thrips (T. meridionalis) and beetles (E. scribae) that visited P. vallisneriifolia flowers were also frequently trapped on the leaves. Furthermore, the abundance of these species on leaves correlated positively with their abundance in flowers (thrips: [r.sub.s] = 0.98, P [less than] 0.005, df = 10; beetles: [r.sub.s] = 0.74, P [less than] 0.05, df = 10; [ILLUSTRATION FOR FIGURE 1C, D OMITTED]). In contrast, the most common medium-to-large floral visitors, Bombylius sp., Lasioglossum sp., and Bombus terrestris, were never found trapped on the leaves, although some rare floral visitors, such as hoverflies and butterflies, were sporadically found adhering to the leaves by their appendages. Thus, only thrips and beetles were both common flower visitors and common prey in most P. vallisneriifolia populations.

Intensive study

Habitat differences. - The three habitats differed markedly in mean irradiance because of the differences in exposure and degree of cliff coverage. Only the sunlight habitat received direct morning sunlight, for [approximately]2 h, coinciding with the flowering peak of P. vallisneriifolia plants. Differences between habitats in mean irradiance (sunlight, 128.2 [+ or -] 26.5 W/[m.sup.2]; shade, 29.2 [+ or -] 4.8 W/[m.sup.2], deep shade, 6.5 [+ or -] 0.6 W/[m.sup.2]; F = 91.8; df = 2, 32; P [less than] 0.0001) were relatively greater than differences in air temperature (sunlight, 19.8 [+ or -] 1.6 [degrees] C; shade, 18.4 [+ or -] 0.7 [degrees] C; deep shade, 17.1 [+ or -] 0.5 [degrees] C; F = 3.48; df = 2, 32; P [less than] 0.05) and relative air humidity (sunlight, 50.0 [+ or -] 4.4%; shade, 56.3 [+ or -] 3.7%; deep shade, 63.6 [+ or -] 3.4%; F = 3.72, df = 2, 32; P [less than] 0.05). These values are averages from sunrise to sunset collected at 1-h intervals in the three habitats on 8 June 1994 during a sunny day. Diurnal temperatures registered during observations of flower visitors ranged from a minimum of 8 [degrees] C in the deep shade to a maximum of 23 [degrees] C in the sunlight habitat, during direct sunlight. Thus, the sunlight site was the warmest and driest, the deep-shade habitat the coolest and wettest, and the shade habitat intermediate.

Pollination and seed production. - Pinguicula vallisneriifolia is self-compatible, but spontaneous autogamy does not occur due to herkogamy (30 bagged flowers set no fruit). All experimental hand-crosses resulted in a high percentage of flowers setting fruit (selfing treatment, 81.5% of flowers set fruits; cross-pollination treatment, 95.8% of flowers set fruit). Furthermore, self- and cross-pollinations resulted in similar numbers and mass of seeds (for selfing, 145.43 [+ or -] 18.12 seeds per capsule, n = 24; for xenogamy, 165.43 [+ or -] 16.31 seeds per capsule, n = 27; F = 0.417; df = 1, 49; NS; mean seed mass for selfing, 12 [+ or -] 0.1 [[micro]gram]; for xenogamy, 13 [+ or -] 0.1 [[micro]gram]; F = 2.104; df = 1, 49; NS). Thus, P. vallisneriifolia depends totally on insect pollinators for reproduction, but natural seed set could represent a mixture of selfing and cross-fertilization. Individual [TABULAR DATA FOR TABLE 1 OMITTED] flowers bloomed for 12.6 [+ or -] 2.8 d (n = 26 flowers) under natural pollination conditions, and 20.6 [+ or -] 3.7 d (n = 25) for unpollinated bagged flowers.

Naturally pollinated plants in the sunlight habitat had a greater fruit set than did shade and deep-shade plants (Table 1). Nearly 100% of the flowers subjected to supplemental hand pollination set fruit in the three habitats. These flowers produced more and larger seeds per capsule than did naturally pollinated flowers. Thus, plant reproductive success was pollen limited in all habitats, regardless of the component of plant reproduction considered (Table 1).

The presence of medium-to-large flower visitors varied between habitats. In the shade habitat during 1992 and 1993, I observed only three Bombus terrestris, one B. pascuorum, and one unidentified small bee visiting flowers of P. vallisneriifolia during 34 h of diurnal observation. Crepuscular and nocturnal observations in the three habitats (over 10 h) revealed no floral visitor. During 1994, I observed two B. terrestris and one B. pascuorum in the shade habitat during a total of 14 h of observation, whereas no large pollinators were observed over 10 h in the deep shade. In contrast, during 8 h of observation when the sunlight habitat was exposed to sunlight (0900-1100), I counted 47 flower visitors (17 Bombylius sp., 12 Lasioglossum sp., four Calliphoridae, three Halictus sp., three Pararge aegeria, two Osmia cornuta, two Empididae, one Macroglossum stellarum, one Pieris napi, one Andrena sp., and one B. terrestris). In the same sector, but without sunlight (1100-2000), the abundance of flower visitors fell drastically: during a similar period (8 h), I observed only seven visitors (three B. terrestris, two Bombylius sp., one Macroglossum stellarum, and one Lasioglossum sp.). Thus, even in the sunlight habitat, large insects visited flowers during the brief period when sunlight fell directly on the flowers. Medium-to-large pollinators were scarce and visited few flowers (normally 1-5 flowers) per foraging trip, probably because the flower of P. vallisneriifolia produces only traces of nectar (0.03 [[micro]liter] per bagged flower per day, n = 30 flowers). These pollinator species are common flower visitors to other plants in the same study area (personal observation).

Thrips and small beetles also visited P. vallisneriifolia flowers. Thrips (T. meridionalis) were very common in flowers in the sunlight habitat (Table 2), less abundant in the shade, and almost absent from the deep-shade habitat. On the other hand, beetles (Eusphalerum scribae) were less abundant in the sunlight habitat than in the shade and deep-shade habitats (Table 2). The abundance of thrips and beetles did not vary temporally. Thrips and beetles colonized recently opened flowers of P. vallisneriifolia at the beginning of blooming. Only adult thrips were seen at flowers. Observations indicated that thrips and beetles remained active at flowers of P. vallisneriifolia all day long, irrespective of the irradiance level. Thrips and beetles foraged actively on the anthers, thus carrying pollen grains of P. vallisneriifolia on their bodies (5.5 [+ or -] 6.6 grains per thrips, up to 24, n = 20; 13.5 [+ or -] 9.8 grains per beetle, up to 30, n = 12). [TABULAR DATA FOR TABLE 2 OMITTED] Both species crawled repeatedly over the upper part of the corolla tube (site of anthers and the stigmatic surface), transferring pollen to the stigma. Furthermore, the exclusion experiments demonstrated that thrips provide pollination: 22% of flowers set fruit (Table 1) despite a cage effect (thrips abundance with the netting was clearly less than without). There were 1.26 [+ or -] 0.21 thrips per caged flower in the sunlight habitats vs. 0.36 [+ or -] 0.14 thrips per caged flower in the shade habitats, in comparison with 1994 data for the same habitat (Table 2; differences in thrips abundance between caged and uncaged flowers: F = 93.15; df = 1,393; P [less than] 0.0001, two-way ANOVA). The number and mass of seeds produced by thrips pollination resembled the results for naturally pollinated flowers (Table 1). Most thrips-mediated pollen movement was probably autogamous. However, some emasculated [TABULAR DATA FOR TABLE 3 OMITTED] flowers excluded from large pollinators also set seed, thus indicating outcrossing by thrips (Table 1).

Insect capture. - No medium-to-large pollinators (e.g., Lasioglossum sp., Bombylius sp., and Bombus sp.) were found adhering to the leaves of P. vallisneriifolia, although a few butterflies (Pararge aegeria and Pieris napi) were found trapped in the sunlight habitat. Captures of small flower visitors by the leaves reflected visitor abundance: many thrips were captured in the sunlight habitat, few thrips in the shade habitat, and almost none in the deep shade (Table 3). The capture of beetles and insects that did not visit flowers (mainly mosquitos, flies, and aphids) also differed between habitats, but not between years (Table 3).

In addition, I analyzed whether flowering plants captured more insects than did nonflowering individuals in the sunlight habitat. Plants with their flowers removed captured only 1.7 [+ or -] 0.24 thrips/10 [cm.sup.2] of leaf surface, compared to 12.3 [+ or -] 1.75 thrips/10 [cm.sup.2] captured by plants with intact flowers (F = 52.55; df = 2, 78; P [less than] 0.0001; [ILLUSTRATION FOR FIGURE 2 OMITTED]). Moreover, nonreproductive plants also captured significantly fewer (2.5 [+ or -] 0.56 thrips/10 [cm.sup.2]) than did flowering ones. Both flower-removed plants and nonreproductive plants captured equivalent numbers of insects that did not visit flowers (mainly mosquitos, flies, and aphids) as did flowering plants (F = 3.08; df = 2, 78; NS; [ILLUSTRATION FOR FIGURE 2 OMITTED]).

The interference between plant-pollinator and plant - prey interactions depended on pollinator size. Thrips were the smallest floral visitors of P. vallisneriifolia (0.98 [[micro]meter] 0.02 mm body length, n = 75), followed by the beetles (2.4 [+ or -] 0.3 mm, n = 32). In contrast, Lasioglossum sp. (measuring 7.3 [+ or -] 1.3 mm, n = 20), Bombylius sp. (8.7 [+ or -] 1.3 mm, n = 15), and Bombus sp. ([greater than]10 mm) were strong enough to escape from the leaves in case of contact. However, entrapment also depended on the mode of locomotion between flowers. Medium-to-large pollinators, such as Lasioglossum sp. and Bombylius sp., flew between flowers, approaching the plant from the front of the corolla and rarely touching the leaves. Beetles usually left the flower by descending the peduncle, often touching a leaf and thus becoming trapped. Thrips became entangled on the leaves when leaving the senescing flowers by means of erratic jumps or short flights.

Although thrips remained immersed in the glandular surface immediately after being trapped because of their tiny body size, beetles sometimes escaped from plants growing in the deep-shade habitat. However, in the sunlight and shade habitats, any beetle that made contact with a leaf invariably become entangled. I tested the retention capacity of the mucilage by placing living beetles (E. scribae) on functional leaves. The experiment involved 10 plants per habitat (one beetle on one leaf per plant) in the sunlight, shade, and deep-shade habitats. Beetles were placed in a natural landing position. All beetles placed on sunlight and shade plants remained fixed for [greater than] 1 h, whereas 80% of beetles escaped from the leaves of plants growing on the deep-shade habitat, because of the low retention capacity of the mucilage there. By escaping from the traps, beetles had a greater chance to visit more flowers in the deep-shade habitat than in the sunlight and shade habitats. In fact, beetles were more abundant as flower visitors than as prey in the deep-shade habitat, whereas the reverse occurred in the shade habitat (Tables 2 and 3).

DISCUSSION

Pollination and seed production

Seed set in Pinguicula vallisneriifolia is pollen limited in all habitats, both in terms of fruit production and in the percentage of ovules developing into seeds (Table 1). Results of the intensive study agree with fruit set data obtained from the 11 populations of the broad study [ILLUSTRATION FOR FIGURE 1B OMITTED]. This suggests that failures to set fruit are due to the absence of pollinator visits, a fact corroborated by the results of pollinator censuses. Because flower longevity is shortened by pollination, the prolonged flower life also suggests a prevailing scarcity of pollinator visits. The scarcity of floral visitors of P. vallisneriifolia is also characteristic of other Mediterranean (P. mundi and P. submediterranea; Zamora et al. 1996) and boreal Pinguicula species (P. vulgaris, P. alpina, and P. villosa; Molau 1993), being thus a widespread phenomenon of the Pinguicula genus.

Despite specialized floral traits of P. vallisneriifolia (large, spurred flowers with zygomorphic shape, contrasting colors and presence of floral guides) that might indicate a bee pollination syndrome, the paucity of medium-to-large pollinators increases the importance of small insects as pollen vectors of P. vallisneriifolia. Given that a typical ovary contains between 150 and 360 ovules, repeated floral visits of small flower visitors are likely to be valuable to the fruit set (Silander and Primack 1978, Mcdade and Dadivar 1984). In this respect, the long-lasting flowers of P. vallisneriifolia both increase the chance of a visit by large pollinators (Primack 1985) and favor repeated pollen transfers by thrips and beetles over the blooming period. Because P. vallisneriifolia reproduces by both selfing and outcrossing, any pollen transported by small flower visitors would be potentially suitable for seed production.

The low individual efficiency of thrips in transporting pollen is somewhat counterbalanced by their abundance at flowers (up to 20 thrips per flower in the sunlight habitat) and their constant presence over the day (up to 100% of the flowers had thrips in the sunlight habitat; Table 2) from the beginning to the end of blooming, thus enabling repeated pollen transfer. In fact, the experimental exclusion demonstrates that thrips are responsible for a proportion of both the fruit set and seed set of flowers that are not visited by large pollinators (Table 1; see Baker and Cruden 1991). In this respect, thrips are quantitatively important as pollinators, because thousands of thrips living in hundreds of flowers can transfer a considerable bulk of pollen, not only within flowers (self pollen), but also between flowers (outcross pollen; see also Ananthakrishnan 1993). Pollination by beetles also appears to be highly influential, because their abundance increased in the shaded habitats (Table 2). They are almost the only pollen vector present in the deep-shade habitat and, consequently, the only potential agent of fruit set and seed set in P. vallisneriifolia in this habitat (Table 1).

Even though only a few meters separate the three habitats, pollinator assemblages differ markedly in both taxonomic composition and abundance, spanning the variability evident over the entire geographical range of P. vallisneriifolia. The environmental patchiness appears to determine the main differences in pollination regimes. The cliffs where the plants grow cast shadows where sunlight barely reaches, and where temperatures remain low during blooming. This shading can strongly limit occupancy by many ectothermal insect species, which cannot remain active below a specific temperature threshold (Corbet 1990, Herrera 1995a). In fact, both the large-scale and the intensive studies indicate that most medium-sized pollinators, such as Bombylius sp. and Lasioglossum sp., forage only during the short period when sunlight falls directly on the flowers. On the other hand, large insects that maintain elevated body temperatures while they are active, such as bumble bees and syphingid moths (Heinrich 1993), forage indiscriminately across the thermal range (see also Herrera 1990, 1995a, b). In contrast to medium-sized pollinators, the small pollinators visit all habitats (thrips mainly in the sunlight, and beetles mainly in the shade and deep shade), and remain active all day long, indicating that their presence does not directly depend on scattered sunlight. These patterns of habitat distribution persist at both the spatial and temporal scale, for large pollinators and especially for thrips and beetles (Table 2). Therefore, although there is a within-population spatial mosaic in the pollinator assemblage of P. vallisneriifolia, there is site-specific predictability in the interaction between P. vallisneriifolia and its pollinators. As a result, pollen dispersal appears to be concentrated within habitats, being strongly restricted between patches a few meters apart that differ with respect to irradiance and thermal regimes (e.g., sunlight and deep-shade habitats) and, hence, pollinator assemblages. Furthermore, this spatial variability in pollinator assemblage translates into differential female reproductive success (Table 1), and can result in spatially variable selective pressures of pollinators upon the plants.

Interference between pollinator and prey systems

Pinguicula vallisneriifolia leaves have no obvious mechanism for differentially attracting prey, although the size, form, and spatial distribution of the leaves, overhanging from the cliff wall, increase the probability of capturing flying insects (Zamora 1995). Because leaves do not mimic flower characteristics, as occurs in pitcher plants (Joel 1988, Moran 1996), the key traits favoring the capture of small flower visitors are the spatial proximity of flowers and traps, and the way in which the pollinators move between flowers. On the other hand, flowering and nonflowering plants appear to catch nonpollinating taxa, such as mosquitos, flies, and aphids, which are the most common prey of P. vallisneriifolia (Zamora 1995), with equal frequency. Captures of these insects during the blooming period are also dependent on habitat (Table 3). This spatial pattern of prey capture is reinforced during the summer, when rising temperatures encourage many flying insects to concentrate in the wet, shaded places, avoiding the generally warm and dry conditions (Zamora 1995).

The long-lived flowers of P. vallisneriifolia, which are produced sequentially, allow a typical flowering individual with 2-3 flowers to attract floral visitors and, hence, potential prey, for 20-40 d. Because plants in bloom capture some flower visitors [ILLUSTRATION FOR FIGURE 2 OMITTED], flowers increase the prey capture of visitors smaller than [approximately]5 mm long (Zamora 1995), creating a positive association between the percentage of flowers and leaves visited by thrips and small beetles [ILLUSTRATION FOR FIGURE 1C, D OMITTED]. Thus, flowers attract insects that subsequently may become prey, increasing prey capture in reproductive plants (see also Karlsson et al. 1994 for P. vulgaris). This enhanced prey capture could further promote vegetative and reproductive success, because the performance of P. vallisneriifolia is prey limited under natural conditions (Zamora et al. 1997, 1998).

Because seed production by P. vallisneriifolia is pollen limited in all habitats (Table 1), the capture of thrips and beetles by the leaves depletes an already limiting resource for plant reproduction: the number of free (untrapped) pollen vectors. The value of such insects as pollinators may decrease, because trapped insects cannot transfer pollen loads. Thus, this conflict of interests can represent a greater negative effect on plant sexual reproduction, through reduced pollination, than a positive effect on plant growth, given the small size of beetles and, especially, thrips as nutrient packets.

Moreover, the prey-pollination conflict is environment specific, because a plant's habitat determines both the availability of floral visitors and the retention capacity of the mucilage. In the sunlight habitat, flowers attract thrips, resulting in pollen transfer. In addition, thrips also provide a substantial reward of nutrients, because the great number trapped by flowering plants (up to 100 thrips per plant) offsets their small size (mean dry mass 7 [[micro]gram]). Flowers receiving exclusively thrip-borne pollen could have lower fitness than flowers cross-pollinated by medium-to-large pollinators available in this habitat. In this situation, trapping thrips would benefit reproduction directly (i.e., reducing the number of this comparatively lower quality pollen vector), as well as indirectly through nutrition, because of the great number trapped. In contrast, the deep-shade environment filters out the presence of most large pollinators and thrips, and beetles become the only pollinators found on the flowers. Because leaves are less adhesive in the deep-shade habitat (Zamora 1995), beetles are less likely to remain fixed on a leaf at first contact and, consequently, they could escape and pollinate several flowers before becoming trapped (Tables 2 and 3). On the other hand, flowering plants in the shade suffer the greatest interference because there are very few large pollinators, and few thrips per flower. All beetles are liable to be trapped once they make contact with the leaves after exiting from the flower, as demonstrated with the experimental placements of beetles on leaves. Thus, they could visit fewer flowers, resulting in the lowest plant reproductive success in the shade habitat (Table 1). In this habitat, pollination proves to be more clearly affected by insectivory than in the sunny and deep-shade habitats.

In conclusion, both the survey of 11 populations and the more intensive field observations and experiments in a single population indicate that P. vallisneriifolia does not efficiently separate the mutualistic plant-pollinator interaction from the antagonistic plant-prey interaction. Flowers and traps of P. vallisneriifolia compete for the same limiting resource: the small flower visitors. Thus, carnivory imposes a cost to the mutualism. This cost, however, is dependent on the local environment.

In a more general context of plant-animal interactions, it should be emphasized that the plant-pollinator and plant-prey interactions reported here depend on species-specific ecophysiological responses of insects and sessile plants to the abiotic characteristics of the environment (see also Herrera 1995b, 1997, and Zamora 1995). When plants are locally distributed throughout contrasting microenvironments, plant and animal populations may not interact homogeneously, creating a site-specific mosaic of mutualistic and antagonistic outcomes. Because of the geographical coincidence between P. vallisneriifolia populations and their small and large pollinator species throughout the plant's range, there is a limited possibility for interspecific outcomes and their potential variation to be geographically structured (sensu Thompson 1994). This plant-insect relationship appears to be microclimatically structured, with the local environment acting as the driving force conditioning the nature, strength, and variability of interspecific interactions. These results are relevant in the larger context of plant-animal interactions, clearly indicating that the physical environment can predictably affect the outcomes of interactions and, ultimately, the balance between mutualistic and antagonistic systems.

ACKNOWLEDGMENTS

We thank the Consejeria de Medio Ambiente, Junta de Andalucia, for permission to work and for providing housing facilities in the Parque Natural de Cazorla, Segura y Las Villas. Drs. Carlos Herrera, Felipe Pascual, Alfredo Lacasa, and Lee Rogers helped with insect identifications. The English version of the text was examined by David Nesbitt. Drs. Staffan Karlsson, Paul Wilson, W. Scott Armbruster, and an anonymous reviewer provided useful critical comments. Jose A. Hodar and Jose M. Gomez offered helpful suggestions for data presentation. This work was supported by CICYT grants PB90-0852 and AMB95-0479. This work is dedicated to Jose M. Gomez, Jose A. Hodar, Daniel Garcia, and Jorge Castro, my beleaguered colleagues of the Terrestrial Ecology group (grupusculo) of the University of Granada.

LITERATURE CITED

Ananthakrishnan, T. N. 1993. The role of thrips in pollination. Current Science 65:262-264.

Baker, J. D., and R. W. Cruden. 1991. Thrips-mediated self-pollination of two facultatively xenogamous wetland species. American Journal of Botany 78:959-963.

Bournier, A. 1983. Les thrips. Biologie et Importance Agronomique. INRA, Paris, France.

Bronstein, J. L. 1994. Conditional outcomes in mutualistic interactions. Trends in Ecology and Evolution 9:214-217.

Corbet, S. A. 1990. Pollination and the weather. Israel Journal of Botany 39: 13-30.

Dunson, W. A., and J. Travis. 1991. The role of abiotic factors in community organization. American Naturalist 138: 1067-1091.

Faegri, K., and L. van der Pijl. 1979. The principles of pollination ecology. Pergamon Press, Oxford, UK.

Givnish, T. J. 1989. Ecology and evolution of carnivorous plants. Pages 242-290 in W. G. Abrahamson, editor. Plant - animal interactions. McGraw-Hill, New York, New York, USA.

Heinrich, B. 1993. Hot-blooded insects. Springer-Verlag, Berlin, Germany.

Herrera, C. M. 1990. Daily patterns of pollinator activity, differential pollinating effectiveness, and floral resource availability in a summer-flowering Mediterranean shrub. Oikos 58:277-288.

-----. 1995a. Floral biology, microclimate, and pollination by ectothermic bees in an early-blooming herb. Ecology 76:218-228.

-----. 1995b. Microclimate and individual variation in pollinators: flowering plants are more than their flowers. Ecology 76:1516-1524.

-----. 1997. Thermal biology and foraging responses of insect pollinators to the forest floor irradiance mosaic. Oikos 78:601-611.

Jennersten, O. 1988. Insect dispersal of fungal disease: effects of Ustilago infection on pollinator attraction in Viscaria vulgaris. Oikos 5: 163-170.

Joel, D. M. 1988. Mimicry and mutualism in carnivorous pitcher plants (Sarraceniaceae, Nepenthaceae, Cephalotaceae, Bromeliaceae). Biological Journal of the Linnean Society 35: 185-197.

Joel, D. M., B. E. Juniper, and A. Dafni. 1985. Ultraviolet patterns in the traps of carnivorous plants. New Phytologist 101:585-593.

Juniper, B. E., R. B. Robins, and D. M. Joel. 1989. The carnivorous plants. Academic Press, London, UK.

Karlsson, P. S., L. M. Thoren, and H. M. Hanslin. 1994. Prey capture by three Pinguicula species in a subarctic environment. Oecologia 99: 188-193.

Mcdade, L. A., and P. Dadivar. 1984. Determinants of fruit and seed set in Pavonia dasypetala (Malvaceae). Oecologia 64:61-67.

Molau, U. 1993. Reproductive ecology of the Nordic Pinguicula species (Lentibulariaceae). Nordic Journal of Botany 13: 149-157.

Moran, J. A. 1996. Pitcher dimorphism, prey composition and the mechanisms of prey attraction in the pitcher plant Nepenthes rafflesiana in Borneo. Journal of Ecology 84: 515-525.

Pellmyr, O. 1989. The cost of mutualism: interactions between Trollius europaeus and its pollinating parasites. Oecologia 78:53-59.

Primack, R. B. 1985. Longevity of individual flowers. Annual Review of Ecology and Systematics 16:15-37.

SAS Institute. 1995. JMP introductory guide. SAS Campus Drive, Cary, North Carolina, USA.

Silander, J. A., and R. B. Primack. 1978. Pollination intensity and seed set in the evening primrose (Oenothera fruiticosa). American Midland Naturalist 100:213-216.

Strauss, S. Y., and W. S. Armbruster. 1997. Linking herbivory and pollination - new perspectives on plant and animal ecology and evolution. Ecology 78:1617-1618.

Thompson, J. N. 1988. Variation in interspecific interactions. Annual Review of Ecology and Systematics 19:65-87.

-----. 1994. The coevolutionary process. University of Chicago Press, Chicago, Illinois, USA.

Zamora, R. 1995. The trapping success of a carnivorous plant (Pinguicula vallisneriifolia): the cumulative effects of availability, attraction, retention, and robbery of prey. Oikos 73:309-322.

Zamora, R., J. M. Gomez, and J. Hodar. 1997. Responses of a carnivorous plant to prey and inorganic nutrients in a Mediterranean environment. Oecologia 111:443-451.

Zamora, R., J. M. Gomez, and J. Hodar. 1998. Fitness responses of a carnivorous plant in contrasting ecological scenarios. Ecology 79: 1630-1644.

Zamora, R., M. Jamilena, M. R. Rejon, and G. Blanca. 1996. Two new species of the carnivorous genus Pinguicula (Lentibulariaceae) from Mediterranean habitats. Plant Systematics and Evolution 200:41-60.

Zar, J. H. 1996. Biostatistical analysis. Third edition. Prentice-Hall, Englewood Cliffs, New Jersey, USA.

COPYRIGHT 1999 Ecological Society of America
COPYRIGHT 2000 Gale Group