The Campsis-Icterus association as a model system for avian nectar-robbery studies

Avian nectar-robbing is common in some floras but its impact on plant-pollinator mutualisms, flowering phenology, and the evolution of floral traits remains largely unexplored. Surprisingly, there have been no quantitative studies of the topography and extent of floral damage inflicted on any flowering species by nectar-robbing birds. I studied nectar-robbing of orchard oriole (Icteridae: Icterus spurius) on the large reddish-orange flowers of trumpet creeper (Bignoniaceae: Campsis radicans), an ornithophilous liana of eastern North America. Floral traits that inhibit nectar-robbery by hummingbirds and bees, such as the thickened calyx and sympetalous corolla, are ineffective in deterring orioles. Orioles target the zygomorphic trumpet-shaped corollas at the 11:00 h or 01:00 h positions with a closed-bill puncture and then enlarge the incision with bill-gaping to reach the nectary. More than 92% of flowers were robbed when orioles were present. Fruit set was nil until orioles departed on fall migration in late July-early August. The timing suggests oriole nectary-robbery may be a potent selection agent for an extended flowering season or delay in the onset of flowering. The biological and geographic attributes of the Campsis-Icterus association make it a promising model system for studying the consequences of avian nectar-robbery on pollination biology and floral trait evolution.

system for avian nectar-robbery. These factors and uncertainty about the taxonomic diversity and geographic distribution of nectar-robbing birds likely explain the absence of comprehensive reviews on the subject.
Mature Campsis bear hundreds of flowers from June through September. Nectar yields of individual flowers may exceed 115 µl/day, the highest average volume reported for temperate latitude plants in North America [54][55][56] . Prodigious flower production and high nectar volume are powerful attractants for insect and hummingbird pollinators 42 . There are a few anecdotes of avian nectar-robbery 42,[57][58][59] . In the most detailed note, Wunderle 57 reported orchard orioles robbing flowers in North Carolina. Orioles punctured corollas with their bills, enlarged the punctures with bill-gaping, and lapped nectar through the incisions. Most open flowers in the oriole territory were pieced along with a smaller percentage of unopened buds. These observations raise important questions about the local and geographic scope of oriole nectar-robbery and its potential impact on the pollination biology and floral evolution of Campsis.
I conducted an exploratory study of oriole nectar-robbery for two flowering seasons in the Ozark Mountains. The data reported here were compiled with four objectives: (i) describe how orioles rob nectar from Campsis flowers; (ii) collect quantitative data on floral damage; (iii) map the topography of piercings with regard to floral symmetry; and (iv) document the effect of nectar-robbery on fruit set. Finally, I discuss the biological and geographic attributes of the Campsis-Icterus association that make it a candidate model system for the study of avian nectar-robbery.   (Fig. 1). The primary diet of insects, spiders, and fruit 61,62 is supplemented by floral nectar during migration and on the wintering grounds 62,63 . In common with several other genera of New World blackbirds (Icteridae), the genus Icterus possesses a pronounced retroarticular process (processus retroarticularis 64 ) at the posterior end of the mandible 65 (Fig. 1). The process acts as a powerful lever that depresses the lower jaw when the well-developed depressor mandibulae muscle is contracted. This muscle originates on the cranium and inserts on the retroarticular process 66 . Protraction in concert with elevation of the maxilla by the protractor pterygoidei et quadrati muscle enables a forceful opening of the bill 65 , a foraging maneuver termed bill-gaping 65 or gaping 67 . Orioles thrust their closed bill into fruit and then open the bill against the resistance of the skin and pulp. Bill-gaping is also used during nectar-robbery to widen incisions in flowers 57,65,68 . Orchard orioles use similar behaviors to legitimately pollinate the large flowers of Erythrina fusca in Panama by splitting the commissure of the floral standard to gain access to the nectary 62 . The skillfully woven pendant or semi-pensile nests of orioles provide additional evidence of fine motor skill of bill manipulation [68][69][70] . Orchard oriole breeds in eastern North America from the Gulf coast west to the Great Plains and north to the Great Lakes 61 . Breeding orioles display a preference for semi-isolated deciduous trees in riparian corridors, pastures, orchards, suburbia, and farmland. Breeding pairs exhibit weak or no territoriality in the traditional sense. Nests are usually spatially dispersed in suitable habitat but in exceptional cases several pairs may nest in the same tree 61,71 . Orchard oriole is notable among the breeding songbirds of eastern North America in initiating fall migration as early as mid-July. Relatively few individuals have been reported in the Ozark Mountains after mid-August. www.nature.com/scientificreports/ Description of oriole foraging behavior. I monitored oriole foraging for 1-3 h daily. Observations were made with Zeiss 10 × 54 binoculars from distances of 10-20 m. These observations were augmented with video recorded with a GoPro Hero 9 Black camera (GoPro, San Mateo, CA). Frame rate ranged from 24-240 fps. The camera was boom-mounted on a stanchion and placed 0.5-1.0 m from Campsis inflorescences. Still photography was conducted with a Canon Eos 80D equipped with a 400 mm lens. Orioles were unmarked but four sex/age classes could be distinguished in the field: ASY (after second year) males, SY (second year) males, AHY (after hatching year) females, and HY (hatching year) immatures. Piercing behavior and handling time were difficult to quantify because foraging movements were rapid and frequently screened by flowers and foliage. Orioles were observed to revisit corolla piercings but I was unable to determine if they favored pristine corollas over previously pierced corollas, as has been reported for bananaquits 72 . Owing to insufficient videographic data, behavioral characterizations presented in the Results are descriptive rather than quantitative. Ruby-throated hummingbirds (4 g) foraged daily at Campis flowers but observations were not systematically compiled. Campsis is predominately outcrossing but cryptic self-fertility has been reported 44 . Nectar production in unopened buds ramps up 12-24 h before anthesis 73 . Newly opened flowers produce copious nectar for 24 h and residual amounts to 30 h 42 . Flowers do not produce additional nectar unless nectar is removed 73 . Sugar concentrations of nectar (in sucrose equivalents) range from 24-35% 38,53,54 . Floral nectar sugars consist of sucrose (5.8%), fructose (32.6%), and glucose (61.5%) 74 . Hummingbirds are the primary pollinators but bumblebees, halictid bees, and honey bees are also capable of pollinating 42,49 . Extrafloral nectaries occur on the petiole, calyx, and external surface of the unopened corolla 75,76 . The slightly curved, spindle-shaped fruit capsules (10-28 cm long, 1.5-2.5 cm wide) contain several hundred bialate wind-dispersed seeds. Fruit capsules dehisce in winter but may remain attached until the following flowering season.
Floral data. Undisturbed corollas loosen and abscise a few days after opening 42 , leaving the calyx and reproductive parts attached to the ovary. Wind hastens corolla abscission. Corolla piercing and the jostling of flowers by foraging orioles also detaches corollas. The ground under the Campsis arbor was usually littered with freshly detached corollas after oriole foraging bouts. I collected detached corollas from the ground below the Campsis arbor, and those lodged in foliage, 2-4 times daily.
The length (mm) of detached corollas was measured from the tip of the narrow base to the level of the reflexed dorsal petal lobes (see Fig. 2b in Bertin 42 ). Oriole piercings usually parallel the long axis of the corolla, often in the form of a slit or slash (Fig. 3). "Piercing" in this paper refers to the initial bill puncture as well as the larger contiguous slash caused by bill-gaping. I measured piercing length (mm) and the distance from the proximal end of the piercing to the base of the corolla. When piercings extended through the base of the corolla the latter measurement was zero.
The topography of oriole piercings was quantified by visualizing the tubular corolla, in cross section, as a radial clock face with 12 h positions, each corresponding to 30° of the 360° circular cross section. The narrow corolla base was viewed with the trumpet-shaped opening facing outward from the observer. The location of piercings was recorded to the nearest clock position with 12:00 h and 06:00 h representing the dorsal (0°) and ventral midlines (180°), respectively. Clock positions were converted to radians for statistical analysis. The central question was whether piercings are uniformly distributed around the tubular corolla or clustered in one or more regions. Inspection of the raw data revealed that the piercings were bimodally clustered. To test the hypothesis of uniform distribution, I employed the Hermans-Rasson test 77 (hereafter H-R test), which significantly outperforms the popular Rayleigh test in multi-modal situations 78 . The H-R test was perfomed in R (version 4.2.0) with RStudio interface (version 2022.02.2) 79 using the package, CircMLE: Maximum Likelihood Analysis of Circular Data. Statistical analyses of corolla morphology (two sample t-tests and simple linear regression) and corolla counts (chi-square and binomial tests) were performed with the R Stats Package.

Results
Campsis-Icterus association phenology. Flowering at the study site commenced in mid-June, peaked from mid-to late July, tapered off in August, and ended in mid to late September (Fig. 2). Orioles arrived on breeding territories two months before Campsis flowering began and departed five weeks before flowering tapered off in mid-September. Orioles were last observed at the Campsis arbor on 1 August 2020 and 20 July 2021. The timing suggests that oriole nectar-robbery is facultative rather than obligatory.
Oriole foraging behavior. Orioles  www.nature.com/scientificreports/ briskly among the terminal inflorescences, were usually silent, and departed quickly after feeding. Foraging orioles perched on the stems or flower pedicels of inflorescences and inspected open flowers and buds. Once in position, an oriole tilted its bill downward at an acute angle and punctured the corolla with closed bill (Fig. 1). Some bill punctures appeared to provide direct access to the pooled nectar without any additional enlargement of the opening. However, most punctures were followed by bill-gaping that split or slit the corolla along its long axis. Orioles were never observed to insert their heads into open corollas, and it may be impossible for them to feed legitimately in this fashion because their heads are too large to insert deeply into the trumpet-shaped corollas and their bill is too short to reach the nectary. Legitimate foraging through the corolla opening, if it were possible, may increase predation risk because the oriole's vision would be blocked when its head is inserted in the flower.
Piercing metrics. Orioles pierced 93% of open corollas in 2020 with a daily average of 38 and a high count of 86 on 21 July (Fig. 2). The entire crop of abscised flowers was pierced on 5 of 24 days. Flower production was substantially lower in 2021, when orioles pierced 92% of open corollas with a daily average of 24 and a high count of 45 on 1 July (Fig. 2). The entire daily crop was pierced on 7 of 21 days. The length of corolla piercings varied from 2-67 mm (Fig. 3) with sharply peaked distributions in 2020 and 2021. Piercings were significantly longer in 2020 (x̅ = 16.7 ± 8.2 mm) than in 2021 (x̅ = 11.3 ± 8.8 mm) (t = 11.8, 2-tailed P < 0.0001). More than half (61.2%) of all piercings were 12-20 mm in length. Piercings extended through the base of corollas in 90.5% of robbed flowers. Unusually long piercings (> 25 mm) were rare (4.8%) and were likely created by multiple punctures and bill-gapings (Fig. 3). Piercing length was uncorrelated with corolla length in 2020 (F = 0.14, P = 0.71; n = 951) but significantly correlated in 2021 (F = 8.3, P = 0.004; n = 530).
Piercing topography. The vertical, horizontal, and transverse orientation of Campsis flowers may influence the angle and location of nectar-robbing attacks. On average, piercings should occur most frequently on the dorsal surface of the corolla nearest the perching oriole. Piercings were indeed clustered in 2020 (H-R test, T = 762.5, P = 0.0001) and 2021 (H-R test, T = 115.6, P = 0.0001). In the combined sample, more than three-quarters (77.4%) of piercings occurred on the dorsal surface from the 10:00 h to 02:00 h positions (41.7% of 360°). Piercings were bimodally clustered in 2020 and 2021 (Fig. 4) and were more than twice as frequent at 11:00 h (28.2%) and 01:00 h (23.6%) than at 12:00 h (8.9%). None of the other positions (02:00-10:00 h) individually accounted for more than 9.0% of the piercings. Piercings were more evenly distributed around the tubular corolla in 2021 than in 2020 (chi-square, 12 × 2 contingency table; χ 2 = 211.4, df = 11; P < 0.0001). There was no evidence of handedness or lateralization in piercing behavior in the combined sample of pierced corollas: left total (n = 675; 07:00-11:00 h) versus right total (n = 651; 01:00-05:00 h) (binomial test, P = 0.25). Handedness tendencies of individual orioles, however, could be canceled out in summary counts of piercings produced by multiple orioles.
The concentration of piercings at the 11:00 h and 01:00 h positions appears related to the morphology of the thickened cuplike calyx (length = 20-28 mm [x̅ = 25.6 ± 2.1 mm]; width = 10-12 mm [x̅ = 10.6 ± 0.6 mm]; n = 28) that protects the nectary and corolla base. The acuminate point of the dorsal lobe of the 5-lobed calyx extends distally along the midline of the corolla at 12:00 h (Figs. 1, 3). The 11:00 h and 01:00 h positions approximate the locations of notches between the dorsal and upper lateral calyx lobes. About half (51.8%) of the corolla piercings were located in the two uppermost calyx notches. The calyx is considerably tougher than the underlying corolla, but the fusion lines between the dorsal and lateral calyx lobes split easily when pressure is applied at the apex of the notch. Orioles take advantage of the weak suture by puncturing the corolla above the notch with closed bill and then bill-gaping to split the corolla and calyx along the same line. The deep split opens the corolla to its base and permits the oriole to lap up the pooled nectar with its brushy-tipped tongue 65,80 . Temporal patterns of piercing and fruit set. Orioles frequently pierced unopened buds 57  I used the criterion of the ovary overtopping the calyx as an indicator of effective pollination 42 . Daily inspections from 9 July-8 August 2020 revealed no fruit (orioles last observed on 1 August). When daily inspections resumed on 24 August, several immature fruit capsules were present. Eighteen maturing fruiting capsules were www.nature.com/scientificreports/ present on 10 October. In 2021, the first fruiting capsules emerged on 2 August (~ 12 days following the oriole departure). The number of fruiting capsules increased from five on 5 August to a maximum of 10 on 22 August.

Discussion
The concentration of oriole piercings at the 11:00 h and 1:00 h positions on the dorsal surface of zygomorphic Campsis corollas was perhaps the most unanticipated finding. Little is known about the ontogeny of complex foraging maneuvers 62 in the Icteridae but naïve orioles likely learn to target the weakly fused sutures between calyx lobes through trial and error. Precision piercing presumably enhances nectar extraction efficiency 81,82 .
Year to year variation in piercing statistics likely reflect differences among orioles in piercing technique and serves as a reminder that inferences drawn from single-season field studies may be skewed by dominant individual performances.
The intensity and constancy of nectar-robbery were also unexpected. Orioles monitor developing buds and begin piercing unopened corollas before hummingbirds and bees can access nectar through the corolla mouth. Orioles pierced 92% of corollas that opened during their residency. The total daily crop of flowers was pierced on 27% of days (12 of 45) when orioles were present. It remains to be determined if the level of nectar-robbery observed in this study is the norm when Campsis occurs on oriole territories.
The mid-summer departure of orioles from the study area constituted a natural experiment on the impact of nectar-robbery on Campsis reproduction. Fruit set was nil when orioles were present in 2020 and 2021 despite daily visitation of hummingbirds and bees at the Campsis arbor. Although the bulk of flowering occurred in July (Fig. 2), the earliest evidence of fruit capsule growth was observed in August, ~ 12 days after oriole departure. Flower production naturally tails off in August after the oriole exodus but pollination success increased sufficiently to result in a moderate number of fruiting capsules by the end of the flowering season in 2020 (n = 18) and 2021 (n = 10). The timing suggests that a few nectar-robbing orioles may be capable of disrupting local Campsis reproduction. These observations add an unanticipated twist to the interpretation of fruit set frequency observed in other North American populations of Campsis. The low rate of fruit production (1.2-8.9%) at four sites in Illinois was attributed to insufficient pollen delivery by hummingbirds and bees 42 . Oriole nectary-robbery was noted at several sites but its potential impact on pollination was not addressed.
In the present study, suppression of fruit set during the oriole residency was likely related to resource and interference competition with legitimate pollinations as well as the collateral effects of nectar-robbery. This study did not focus on oriole-hummingbird interactions or on hummingbird pollination, but the presence of larger-bodied orioles (17-24 g) appeared to have at least a moderately dissuasive influence on hummingbird (4 g) visitation.
Orioles robbed most open flowers and this fact alone may have reduced the frequency of visits by pollinating hummingbirds and bees. Additionally, hummingbirds invariably probed oriole piercings when they were present rather than accessing nectar through the corolla mouth. Piercing also leads to premature corolla abscission which removes the primary visual signal for legitimate pollinators. Hummingbirds frequently probe emptied calyces 42 but they seldom contact the protruding reproductive structures. Finally it is possible that oriole bill punctures and bill-gaping damage the ovary or style. Gentry (1974) characterized eight floral types associated with specific categories of pollinators among the 78 species of native Bignoniaceae of Costa Rica and Panama. Campsis does not occur naturally in Central America but Gentry mentioned that it has Martinella-type flowers, which are usually bright red-orange or deep redviolet with thickened corolla walls, an open corolla throat, and large thin calyces that enclose the base of the corolla tube. Flowers of this floral type are typically odorless, produce abundant nectar, and are pollinated by hummingbirds. Although the penetrability and toughness of floral structures in the Bignoniaceae have yet to be quantified, it is difficult to envision any combination of floral traits present among the ~ 82 recognized genera (827 + species) [39][40][41]83,84 that could impede or prevent oriole nectar-robbery. Perhaps the most feasible evolutionary response of Campsis to chronic oriole nectar-robbery would be a delay in the onset of flowering, a prolongation of flowering season, or a shift in peak flowering from July to August or September when pollinating hummingbirds and bumblebees face little or no competition from nectar-robbing orioles. Hummingbirds were observed at the study site as late as 13 October 2020 and 15 October 2021.

Campsis-Icterus as a model system. The biological and geographic attributes of the Campis-Icterus
association make it a promising model system for the study of avian nectar-robbery and its consequences on pollination biology and floral trait evolution.
• Flower size and nectar production. Campsis bears the largest ornithophilous flowers (corolla length × diameter) in the native flora of temperate North America (see 85,86 ). Large corollas offer a range of quantifiable attack points for orioles and copious nectar production provides a favorable precondition for quantifying nectarrobbing efficiency. Both attributes facilitate hypothesis testing. • Floral damage. The size, shape, and topography of oriole piercings are easily quantified in the field without excessive magnification or special preparation. Oriole piercings are difficult to confuse with smaller punctures made by insects. • Pollination network simplicity. Campsis is pollinated by the ruby-throated hummingbird and a relatively small roster of bumblebees, halictid bees, and honeybees in its native range 42,87 . Only two avian species, both orioles, have been reported to rob nectar from Campsis 42,57,58 . Network simplicity aids hypothesis testing, experimentation, and data interpretation. • Geographic range overlap. The large geographic overlap (1.5 million km 2 ) between the breeding range of orchard oriole 88 and the natural and anthropogenic range of Campsis 48 sets the stage for local and regional www.nature.com/scientificreports/ comparisons of the mode, intensity, and reproductive impact of oriole nectar-robbery on Campsis populations. • Pollination biology. Detailed studies of the floral anatomy, stigma receptivity, and nectar and pollen production 42,73,75,76 provide an advanced starting point for investigations of oriole nectar-robbery. • Academic accessibility. Campsis and orioles co-occur near dozens of R1 doctoral granting universities in the eastern United States. This convenience confers significant logistical and financial advantages over bird-plant systems that are accessible only at remote tropical field sites.

Data availability
The datasets used and analysed during the current study are available from the corresponding author on reasonable request.