Floral humidity as a reliable sensory cue for profitability assessment by nectar-foraging hawkmoths

Abstract

Most research on plant–pollinator communication has focused on sensory and behavioral responses to relatively static cues. Floral rewards such as nectar, however, are dynamic, and foraging animals will increase their energetic profit if they can make use of floral cues that more accurately indicate nectar availability. Here we document such a cue—transient humidity gradients—using the night blooming flowers of Oenothera cespitosa (Onagraceae). The headspace of newly opened flowers reaches levels of about 4% above ambient relative humidity due to additive evapotranspirational water loss through petals and water-saturated air from the nectar tube. Floral humidity plumes differ from ambient levels only during the first 30 min after anthesis (before nectar is depleted in wild populations), whereas other floral traits (scent, shape, and color) persist for 12–24 h. Manipulative experiments indicated that floral humidity gradients are mechanistically linked to nectar volume and therefore contain information about energy rewards to floral visitors. Behavioral assays with Hyles lineata (Sphingidae) and artificial flowers with appropriate humidity gradients suggest that these hawkmoth pollinators distinguish between subtle differences in relative humidity when other floral cues are held constant. Moths consistently approached and probed flowers with elevated humidity over those with ambient humidity levels. Because floral humidity gradients are largely produced by the evaporation of nectar itself, they represent condition-informative cues that facilitate remote sensing of floral profitability by discriminating foragers. In a xeric environment, this level of honest communication should be adaptive when plant reproductive success is pollinator limited, due to intense competition for the attention of a specialized pollinator.

Plant–pollinator relationships—like all mutualisms—strike a tenuous balance between the conflicting interests of foraging animals and flowering plants (1, 2). For flowering plants, pollinator attraction through floral density, form, color, pattern, and odor is necessary but insufficient to ensure effective pollination. Ideally, pollinators should transfer pollen between conspecific plants in ways that result in favorable levels of outcrossing (3) and reduced pollen loss (4). Flowers commonly manipulate pollinator movement by varying quality or quantity of floral nectar. The offer of variable amounts of nectar (5, 6), toxic nectar (7, 8), or an empty promise of nectar elicit similar results: pollinators leave plants sooner, reducing for example geitonogamous inbreeding (if self-compatible) or pollen discounting (if self-incompatible).

The availability of floral rewards often coincides with floral display and pollinator activity on a gross temporal scale (i.e., hours) (9, 10). Nevertheless, most pollinators encounter rewardless (“empty”) flowers on a finer temporal scale (i.e., minutes), often due to recent visitation by another animal. From the pollinator’s standpoint, visiting empty flowers is energetically costly and reduces the profitability of a patch, with attendant fitness-related consequences (11, 12). How should pollinators respond to conventional floral stimuli (color, odor, and pattern) when their association with profitability is unpredictable? Selection should favor pollinators that are able to remotely sense the presence of nectar without landing and handling the flower, which may occur, for example, when nectar is scented (13, 14). However, in the many cases where nectar is unscented, pollinators should be selected to respond to “condition-dependent” floral traits—those whose production is mechanistically linked to nectar or the physiological processes that produce it—just as discriminating female animals are thought to choose mates on the basis of ornamental displays that most directly inform them of the male’s condition and quality (15). For example, above-ambient floral emissions of CO₂ peak at dusk, when nectar-rich flowers of Datura wrightii (Solanaceae) open, but floral CO₂ diminishes rapidly after visitation by Manduca sexta moths, whereas floral scent and visual display persist for hours after nectar removal (16). When present, flowers with elevated CO₂ should be preferred by nectar foraging Manduca as the most accurate indicator of profitability in Datura flowers, and they are (17, 18). Whatever the physiological causes, elevated floral CO₂ in D. wrightii flowers provides moths with a more temporally accurate indicator of nectar than the visual and olfactory signals that guide hawkmoths to these flowers (19).

In this context, a good candidate for a condition-informative cue revealing the presence of floral nectar would be a transient humidity gradient produced directly from the evaporation of nectar into the floral headspace. Floral humidity gradients were first measured by Corbet (20) and colleagues (21), who suggested that nectar-foraging animals might profit by attending to such cues. In this study we measured the spatial and temporal dynamics of humidity gradients in a well-studied plant–pollinator system and tested the behavioral responses of pollinators to their presence using behavioral assays in a controlled laboratory environment.

Results

Experiment 1: Temporal and Spatial Characterization of Floral Humidity Gradients During Anthesis.

Humidity measurements 25 mm above the opening of the nectar tube revealed temporal variation in relative humidity (RH) levels in the floral headspace during the first 30 min after initiation of anthesis. Oenothera cespitosa flowers produce humidity gradients increasing the RH at Hyles lineata hovering distances of about 4% above ambient levels (n = 8; Fig. 1A). In the horizontal plane, elevated RH levels were limited radially to a circle with a 50-mm radius and its center directly above the nectar tube (n = 8; Fig. 1B). Vertically, floral humidity gradients approached ambient levels 60 mm above the nectar tube (n = 6; Fig. 1C). An interpolation of the horizontal data showed a concentric pattern of elevated RH levels restricted to the area of the petals (Fig. 1D).

Fig. 1.

Temporal and spatial characterization of humidity gradients in single Oenothera cespitosa flowers [mean (black line) ± SEM (gray area)]. (A) Temporal evolution of humidity gradients over single flowers during anthesis (n = 8). (B) Horizontal dimensions of floral humidity gradients (n = 8). (C) Vertical dimensions of floral humidity gradients (n = 6). (D) Interpolation of the humidity plume in the horizontal plane 10 mm above the flower surface.

Experiment 2: Contribution of Petals and Nectar Tube to the Floral Humidity Gradients.

Four sequential measurements in a horizontal transect over O. cespitosa flowers allowed us to decouple the relative contributions of petals and nectar tube to total water vapor emissions. First, humidity gradients of whole flowers were characterized horizontally (Fig. 2A) followed by measuring humidity gradients originating only from petals (Fig. 2B) or from the nectar tube (Fig. 2C) and ending with measurements over whole flowers (Fig. 2D; each n = 5). Control treatments (intact start and intact end) did not differ significantly (Fig. 2E) and humidity profiles were nearly identical, indicating that they were not altered over time or through experimental manipulations. Repeated measures (RM) ANOVA revealed statistical differences for the treatment (F_(3/16) = 5.25, P = 0.01), distance (F_(1.8/29) = 83.3, P < 0.0001) and treatment × distance interaction (F_(5.4/29) = 5.43, P = 0.0033; Fig. 2E). One-way ANOVA with post hoc Tukey pairwise comparisons indicated that humidity profiles created solely through water emissions originating from the nectar tube were significantly smaller than humidity profiles of intact flowers at transect sections 20 through 80 mm (P < 0.05; Fig. 2E). Humidity profiles of flowers with a reduced number of floral organs had lower amplitudes and/or radial limitation than did humidity profiles obtained from whole flowers. The humidity profiles obtained over flowers with reduced floral organs (Fig. 2 B and C), when combined, appear to reconstitute the total profile measured above whole flowers (Fig. 2 A and D).

Fig. 2.

Contributions of the nectar tube and the petals to the floral humidity gradient measured in four consecutive recordings [mean (colored line) ± SEM (gray area); n = 5]. Water emissions from plant parts were either present (black) or absent (white) during recordings. (A) Horizontal dimensions of the humidity gradient over the whole flower at the start of the sequence. (B) Horizontal dimensions of the humidity gradient over the flower with the nectar tube (NT) plugged. (C) Horizontal dimensions of the humidity gradient over the flower with the petals (P) covered. (D) Horizontal dimensions of the humidity gradient over the whole flower at the end of the sequence. (E) Comparison of humidity gradients at 20-mm increments. Treatments with different letters within one column are significantly different (Tukey HSD, P < 0.05).

Experiment 3: Effects of Preanthesis Nectar Removal or Equilibration Inhibition on Temporal Dynamics of Floral Humidity Gradients.

Floral humidity gradients of O. cespitosa were influenced by the availability of nectar and were disrupted when excisions prevented the saturation of the air space within the flower bud. RM ANOVA revealed statistical differences for the treatment (F_(3/27) = 11.87, P < 0.0001), time (F_(2.3/63) = 16.84, P < 0.0001) and treatment × time interaction (F_(7/63) = 3.45, P = 0.0034). The RH 25 mm above intact flowers increased to about 4% above ambient during the first 30 min of anthesis (n = 8; Fig. 3A). A similar humidity profile was obtained when the nectar tube was punctured at the base of the nectar tube without altering nectar volume (n = 7, sham control; Fig. 3B). If the nectar tube was punctured between the meniscus of the nectar and the sepals, enabling a mixing of the air inside the nectar tube with ambient air, humidity profiles in the floral headspace were altered (n = 8; Fig. 3C). More dramatic effects were obtained when the nectar was removed before the onset of anthesis, resulting only in trivial elevations of humidity levels during anthesis (n = 8; Fig. 3D). One-way ANOVA with post hoc Tukey pairwise comparisons indicated that nectar removal and the disrupted equilibration of air inside the nectar tube resulted in significantly lower humidity at several time points during the first 25 min of anthesis (P < 0.05; Fig. 3E).

Fig. 3.

Effects of preanthesis nectar removal or equilibration inhibition on temporal dynamics of floral humidity gradients in Oenothera cespitosa [mean (colored line) ± SEM (gray area)]. (A) Floral humidity gradients during anthesis of intact flowers (n = 8). (B) Floral humidity gradients during anthesis of flowers that were punctured at the base of the nectar tube (sham control, n = 7). (C) Floral humidity gradients of flowers whose nectar tube was punctured before onset of anthesis between meniscus of the nectar and the top of the nectar tube (n = 8). (D) Floral humidity gradients of flowers whose nectar was removed before anthesis (n = 8). (E) Comparison of temporal humidity patterns at 5-min intervals starting at the onset of anthesis. Treatments with different letters within one column are significantly different (Tukey HSD, P < 0.05).

Experiment 4: Behavioral Response of H. lineata to Elevated Humidity Levels in the Headspace of Artificial Flowers.

Moths preferred artificial flowers with elevated humidity levels over artificial flowers with ambient RH. The flowers with increased RH received significantly more visits (Wilcoxon signed-rank test, Z = −2.802, P = 0.003; Fig. 4A) and H. lineata males were more likely to extend their proboscis while hovering in front of them (Wilcoxon signed-rank test, Z = −2.517, P = 0.008; Fig. 4B). In the control treatment, both artificial flowers had RH gradients similar to the ambient level and moths visited both equally (Wilcoxon signed-rank test, Z = −0.773, P = 0.52; Fig. 4A) and showed no preference of proboscis extensions (Wilcoxon signed-rank test, Z = −0.212, P = 0.91; Fig. 4B).

Fig. 4.

Effects of humidity gradients on Hyles lineata foraging behavior. (A) Number of visits of H. lineata males to artificial flowers in a binary choice assay with either one flower emitting ambient air and the other humidified air (test, n = 14) or both flowers emitting ambient air (control, n = 9). (B) Number of proboscis extensions of H. lineata males to artificial flowers in a binary choice assay with either one flower emitting ambient air and the other humidified air (test, n = 14) or both flowers emitting ambient air (control, n = 9). Different letters indicate significant differences (Wilcoxon signed-rank test, P < 0.05). The limits of the box depict the first and fourth interquartile; the middle line is the median. Error bars depict the variance.

Discussion

Here we provide evidence for transient humidity gradients associated with nectar-rich flowers of O. cespitosa and artificial flowers modeled after them. Fine-scale spatial and temporal measurements, combined with floral and bud manipulations, revealed that elevated humidity levels are produced by evaporation and transpiration, two fundamental physical and physiological processes occurring naturally during anthesis. Nectar evaporation is a known source of postsecretory variation in nectar concentration and energetic content (22, 23). Here we provide a proof of principle that elevated humidity levels produced by nectar evaporation could provide a condition-dependent indicator of profitability (in the presence of nectar, rather than its quality), to those flower-visiting animals capable of perceiving it. A similar approach was taken by Thom et al. (17), who used artificial flowers to demonstrate that floral CO₂ emissions are informative to foraging male hawkmoths. Whether elevated floral humidity functions as a mutually beneficial honest signal (synomone) or simply allows visitors to remotely sense floral resources at the plant’s expense (kairomone) is contingent upon other aspects of plant–pollinator dynamics, including plant breeding system, floral density, and competition for pollinators (13, 24). Below we discuss proximate and ultimate aspects of this phenomenon from both plant and pollinator points of view.

Proximate Mechanisms I: Ontogeny of Floral Humidity Gradients.

Generally, the sepals of Oenothera flowers are connected through intertwined epidermal cells, resulting in a rapid flower opening with the sudden release of turgor-related tension on the folded petals when they overcome resistance. Fine-scale humidity monitoring showed that the onset of anthesis is coupled with a puff of humid air leading to elevated humidity levels in the floral headspace. This ephemeral cue was present for the first 30 min after intact flowers opened (Fig. 1). Humidity gradients on the same order of magnitude were measured over intact plants with exposed vegetation, indicating that such gradients are not an artifact of our optimized experimental setup (Fig. S1). Our floral dissections revealed that humidity gradients are dependent on the presence of nectar and arise as an additive consequence of transpiration from the petals and evaporation from the nectar tube (Figs. 2 and 3). This floral behavior, in which the interplay of evapotranspirational water loss through newly expanded petals combines with water-saturated air from the nectar tube, leads to elevated humidity levels in the floral headspace of O. cespitosa.

Transpiration from petals could be linked to (i) cuticular permeability to water and (ii) water loss through stomatal gas exchange. Cuticular permeability to water is influenced by physicochemical properties. For example, the cuticle of snapdragon flowers (Antirrhinum majus; Plantaginaceae) is dominated by branched alkanes and hydroxy esters (25), which provide a less effective diffusion barrier than long-chain wax hydrocarbons lacking branches, double bonds, and functional groups (26–28). Furthermore, increased transpiration rates have been shown to be associated with reduced cuticular thickness (29). In addition to permeable petal cuticle, water also could be given off through petal stomata. In night-blooming Mirabilis jalapa (Nyctaginaceae), epidermal stomata are open at the beginning of anthesis and gradually close over time, supporting the prediction of enhanced gas exchange during flower opening (30). In O. cespitosa stomatal density on the adaxial petal surface is three- to fourfold lower than stomatal density on leaves (Figs. S2 and S3). Nevertheless, water vapor emissions from petals do not differ significantly from those of leaves in O. cespitosa (Fig. S4). Thus, either the gas exchange from petal stomata is higher than from leaf stomata or additional water emissions through the petal epidermis supplement water loss through petal stomata, supporting the concept of a “leaky” petal cuticle described above.

We provide emiprical evidence for the causal relationship of transient humidity gradients during anthesis and the availability of floral nectar. Water evaporation and its effects on variance in nectar volume, concentration, and foraging behavior by pollinators is well documented (31, 32) and is especially potent in flowers with open, unprotected nectaries (21). Humidity gradients might therefore not accurately indicate nectar quality, but still present an honest signal in the regard of nectar availability.

Proximate Mechanisms II: Hygroreception and Behavior.

We present behavioral evidence that hawkmoth pollinators respond to subtle variation in RH in a foraging context when other floral stimuli are held constant. Hygroreceptive sensilla have been identified from the antennae of a related hawkmoth (M. sexta) (33), and we expect to find similar sensilla on the antennae of H. lineata. Single sensillum recordings in the cockroach Periplaneta americana (Blattidae) demonstrated an exceptional sensitivity to humidity differences at the receptor level. In both moist and dry hygroreceptor cells an increase of one impulse per second can be elicited by a change of just 0.16% in instantaneous humidity (34). Similarly, an average difference of only 2–3% in RH is required to elicit increased action potential spiking from both moist and dry cells in Pterostichus oblolongopunctatus (Carabidae) beetles (35). Nishino et al. (36) investigated central nervous system processing of hygrosensory signals in the antennal lobe of P. americana. Firing rates of projection neurons innervating glomeruli that receive signals from moist, dry, and cold receptors, were affected by moist (+5% RH) and dry (−3% RH) stimuli (the authors only tested these two humidity levels) (36). In sum, insects are equipped to perceive differences in RH on the order of those measured in our flowers in terms of both neural inputs (receptors and interneurons) and behavioral outputs.

Most examples of insect behavioral responses to RH have been studied in the context of microhabitat selection and survival (reviewed in ref. 37). When insects such as H. lineata use local humidity differences in a nectar-foraging context, the humidity gradients are more likely to convey information about floral profitability than microhabitat suitability. Although previous studies document insects seeking high humidity during foraging [e.g., bees balancing osmotic needs in desert environments (38, 39) or male butterflies seeking salts in moist earth (40)], our experiments demonstrate that insects can discriminate between subtle differences in RH in a nectar-foraging context. Our results suggest that local humidity gradients could not only affect pollinator behavior but also impact foraging behavior of other flower visitors. For example, secondary nectar robbers (41) may use such capabilities to find and feed from punctured floral corollas. In this light, local humidity gradients could facilitate the location of extrafloral nectar present in small droplets on the plant surface by carnivorous arthropods (42).

Ultimate Mechanisms: Honest Signaling Effects on Plant–Pollinator Interactions.

Here we present proof of concept that floral water vapor emissions can function as an informative sensory cue with potential to mediate plant–pollinator interactions. Elevated humidity levels in the floral headspace are linked to nectar production, a fundamental physiological process in hawkmoth pollinated flowers, and therefore reliably indicate resource availability to foraging pollinators. The presence of condition-dependent indicators of nectar profitability would offer clear fitness benefits to such animals, by reducing search and handling times (43, 44). However, flowering plants often benefit by decoupling floral advertisements and rewards in ways that both reduce floral investment by the plant and, through pollinator movement, enhance outcrossing (45). Under what conditions would plants also benefit by honestly revealing the presence of floral nectar?

Nectar-derived water vapor gradients are likely to occur in other long-tubed or -throated flowers with an interstitial volume of air within the flower bud and copious nectar production. Large night-blooming flowers pollinated by hawkmoths and/or bats often provide an overabundance of nectar, consistent with pollinator limitation in many nocturnal systems (46, 47). When competition for pollinators is high, as is common for obligately hawkmoth-pollinated, self-incompatible plants (48, 49), honest advertising of nectar rewards might be the best strategy to effect successful outcrossing. Restricted flowering times, short flower longevity, and high pollinator efficiency (e.g., male or female floral fitness is satisfied by one pollinator visit) also are likely to increase the benefits of honest signaling. In plants with multiple flowers, pollinators would be guided to newly opened ones—maximizing geitonogamy and the chance of a copious nectar meal, maintaining the interaction for longer periods (50). Humidity gradients represent one of several mechanisms, including dissolved volatile compounds, by which nectar can advertise its presence (13, 14). Another such mechanism is colored (visually apparent) nectar. Plants growing in insular habitats, such as islands and mountain ranges, may be extremely pollinator limited if they are self-incompatible or need pollinators to increase pollination efficiency (51, 52). In two recently documented cases, the coupling of signal and reward through colored nectar allowed the pollinator to assess nectar availability before flower visitation, increasing their efficiency as foragers and pollinators (53, 54).

Night-blooming flowers often remain turgid and scented, maintaining visual and olfactory displays the morning after anthesis [e.g., Lonicera japonica (Caprifoliaceae) (55)], whereas nectar may be depleted soon after anthesis when hawkmoths are abundant (56, 57). Our previous studies of floral CO₂ (16, 18) were motivated in part by the hypothesis that M. sexta and other foraging hawkmoths should attend to transient CO₂ plumes emitted by newly opened flowers of D. wrightii as more temporally accurate indicators of nectar availability. When elevated CO₂ is liberated as profitable flower buds open, it represents the same kind of condition-dependent cue defined by Hill (15) in the context of sexually selected ornaments in animal courtship and proposed by us for floral humidity gradients. Nevertheless, humidity gradients are likely not to be functional in all plant–pollinator systems. Flowers with large petal surface area, permeable cuticles, and low nectar volumes would have an increased “dishonest” contribution of water vapor emission. In theory, plants could cheat pollinators by diluting their nectar, because humidity gradients convey information about nectar presence rather than quality. Nevertheless, the value of water in comparison with other nectar components should not be underestimated in xeric environments. In Mediterranean habitats nectar concentration is generally higher than in temperate communities (reviewed in ref. 58), which is seemingly related to water limitations. Humidity gradients would lose their functionality in flowers that offer unpalatable nectar and would not form in flowers that bloom in very humid environments or when the nectar tube is physically obstructed (e.g., by hairs). Finally, animals that cannot perceive differences in RH on a floral scale could not use humidity gradients to remotely sense floral profitability. Considering these points, transient humidity gradients should be most reliable in nectar-rich flowers with waxy cuticles, in plants that bloom in xeric environments, and for which reproductive success is limited by effective but rare or unpredictable pollinators.

Materials and Methods

Insects.

A laboratory colony of white-lined sphinx moths, H. lineata (Sphingidae), originating from eggs and larvae collected in 2009 (Colorado Springs, Eldorado County, CO) was maintained in a growth chamber [16 h light (L):8 h dark (D), 24 ± 1 °C, 60 ± 5% RH]. Larvae were reared on a corn-meal–based artificial diet (59). Pupae were sorted by sex into separate incubators (16 L:8 D, 23 ± 1 °C, 54 ± 4% RH; Precision 818), where they eclosed within 45 × 45 × 45 cm screen cages.

Host Plants.

Evening primrose plants, O. cespitosa ssp. marginata (Onagraceae) seeds were collected in the field in 2008 (Logan Canyon, Cache County, UT; accession RAR01-55) and were grown in a greenhouse (average temperature 28.8 °C, 58.8% RH). The photoperiod was based on the natural light cycle with additional lighting (0600–2100 hours) from November through March (approximately 15 L:9 D).

Experimental Setup of the Measurement Chamber for Monitoring Local Humidity Gradients.

Changes of humidity levels in floral headspace during anthesis were measured in a dark room (19 ± 1.2 °C, 40 ± 10% RH). Single flower buds were placed in a 46 × 46 × 46 cm chamber (BioQuip) with plastic walls. The front panel was equipped with a sleeve opening (22 cm diameter) and the floor with a guillotine-like access (0–7 mm diameter). Floral headspace and the ambient RH was recorded simultaneously every 5 s with two humidity and temperature transmitters (Humitter 50Y) and stored on a datalogger (Micrologger CR23×).

Spatial characterization of humidity gradients.

Humidity gradients were characterized over single natural flowers. Flower buds were introduced through the guillotine-like opening into the chamber. While the reference probe was stationary, the test probe was moved along horizontal and vertical 100-mm transects (11 mm/min). Horizontal transects were made 10 mm above the opening of the nectar tube. For vertical transects, the starting point of the exploring probe was directly above the nectar tube. One transect was made per flower per night.

Temporal characterization of humidity gradients over single flowers.

RH during anthesis was monitored over single flowers. The test probe was placed 25 mm above the base of the flower bud. Subsequently, experimental manipulations were applied to investigate the role of nectar amounts and the accumulation of water-saturated air inside the nectar tube on RH levels in the floral headspace. Before anthesis either (i) the nectar tube was punctured at its base and nectar removed with a syringe, (ii) the nectar tube was punctured at its base but the needle was left in place and no nectar was removed (sham control), or (iii) a square window (2 × 2 mm) was cut into the nectar tube between the meniscus of the nectar and the sepals (5 mm below the sepals) without removing nectar. The temporal evolution of RH levels was measured for one flower per night.

Contribution of nectar tube and petals to humidity gradients in floral headspace.

Four successive horizontal transects (length, 100 mm; height above the flower, 10 mm; and speed, 11 mm/min) were carried out per flower in the following order: (i) a transect over the entire uncovered flower, (ii) a transect over the flower with the nectar tube plugged by modeling clay, (iii) a transect over a flower covered with a Petri dish (97 mm diameter) with a hole (11 mm diameter) in the center, leaving the nectar tube uncovered, and (iv) one final transect over the entire uncovered flower.

Experimental Setup of the Behavioral Assay.

Bioassays were carried out in a flight cage (1.2 × 1.2 × 1.2 m; 16 L:8 D, 20 ± 0.5 °C, 40 ± 10% RH). Illumination was provided by 50 cool-white and 50 warm-white light-emitting diodes. Artificial flowers consisted of a pipette tip connected to a bottom part of a Petri dish (90 × 20 mm) covered with laminated yellow paper. Each flower was connected with silicone tubing to a system consisting of a pump (TOP FIN XP-125), a flowmeter, and a bubbler delivering humidified air at 200 mL/min, mimicking RH gradients of natural flowers (test flower, Fig. S5). In control flowers, the air passed through an empty bubbler and therefore acquired humidity levels similar to ambient air. To achieve a homogeneous diffusion of delivered air, the flower surface contained pinholes in a checkerboard style (5 × 5 mm). Either a control flower was presented in combination with a test flower (test setup) or with another ambient flower (control setup; spaced 29 cm) on a wire frame (9 × 30 × 46 cm) covered with black cotton. A cotton swab impregnated with bergamot oil (Body Shop) was positioned below the wire frame. Bergamot oil is a reliable feeding stimulant for M. sexta (60) and, like many hawkmoth-pollinated flowers, is dominated by linalool and related monoterpenoid odors (61). At the 5th hour of photophase, eight 3-d-old naïve and starved H. lineata males were introduced into the flight cage and flew freely for 24 h. Moth behavior was recorded with a webcam and scored for floral approaches and proboscis extensions. This procedure was repeated 14 times for the test setup and 9 times for the control setup, using eight new moths each time (i.e., each moth was used only once), testing a total of 112 and 72 moths, respectively.

Statistics.

To analyze the contribution of different floral organs to elevated RH, RM ANOVA (JMP; 8.0.2; Greenhouse-Geisser corrections are reported when sphericity was violated) was used to test the significance of the univariate repeated measures factor with four levels. The distance × treatment interaction term indicated whether RH profiles over the different flowers were affected similarly by the presence of different numbers of floral organs. Subsequently, separate one-way ANOVAs (post hoc Tukey HSD) were used to test whether treatments caused significant changes in RH at six different points along the transect (0, 20, 40, 60, 80, and 100 mm). Effects of nectar amounts and saturated air in the nectar tube on temporal patterns of humidity gradients were analyzed as described above. The time × treatment interaction term indicated whether RH profiles over the flowers were affected similarly by experimental manipulations. Pairwise comparisons indicated whether treatments caused significant changes in RH during anthesis (0, 5, 10, 15, 20, and 25 min). Behavioral data were analyzed using the Wilcoxon signed-rank test. A replicate consisted of the total number of moth approaches and proboscis-extended probing events observed in 24 h to the two treatments.