1.1 Historical prolog

The ruby-throat is the most intensively researched hummingbird species (Schuchmann, 1999; Weidensaul et al., 2020), with a historical literature dating to the 17th century (Clayton, 1693; Grew, 1693; Josselyn, 1672). The earliest known illustration of an ornithophilous plant with its primary pollinator (Figure 1) is a hand-colored etching of a ruby-throat feeding at a trumpet creeper flower (Catesby, 1731). Early natural history accounts mentioned several nectar plants (Bartram, 1792; Kalm, 1770), with trumpet creeper receiving the most attention (Audubon, 1831, 1835; Catesby, 1731; Nuttall, 1832; Wilson, 1810). The mutualistic relationship between trumpet creeper and ruby-throats was well known several decades before the formal conceptualization of ornithophily as a pollination syndrome in flowering plants (Delpino, 1867).

Catesby (1731: 65) appears to be the only observer to note the considerable physical exertion required for these birds to obtain nectar legitimately from trumpet creeper flowers, “The Humming Birds delight to feed on these Flowers; and, by thrusting themselves too far into the Flower, are sometimes caught.” This observation was presumably made in the Carolinas or Virginia.

Wilson's account made no mention of ruby-throats inserting themselves in flowers or becoming trapped. It is plausible that he either overlooked Catesby's observations or did not fully appreciate their significance.

Audubon (1831), who closely studied Wilson's work, did not describe the mode of ruby-throat foraging at trumpet creeper flowers. However, his iconic painting of ruby-throats (1835) showed a swarm feeding at a cluster of trumpet creeper inflorescences. Two birds are depicted with their bills inserted in flowers, although not deeply enough to obtain nectar. Both are hovering rather than perched, giving the impression that ruby-throats are capable of legitimately obtaining nectar from trumpet creeper flowers without fully inserting their heads inside the corolla. This portrayal became the prototypical representation of ruby-throat foraging behavior and likely shaped the perception that further study was unnecessary. As a result, neither historical nor modern accounts accurately describe their feeding behavior at trumpet creeper flowers.

This paper unravels the mystery of Catesby's “caught” hummingbirds and documents an immersive foraging mode in ruby-throats that is, at the very least, rare among hummingbirds. An analysis of floral traits further clarifies the ornithophilous adaptations of trumpet creeper for hummingbird pollination.

2.1 Trumpet creeper populations

2.2 Overview of trumpet creeper floral traits

The campanulate corolla widens from a narrow elongated cylindrical base and terminates in five rounded petal lobes that unfold at anthesis. The ventral petal lobe (Figure 2) functions as a landing platform for hummingbirds and bees. The pentagonal corolla opening is vertically compressed, wider horizontally than it is tall. The base of the corolla, which encloses the basal nectary and superior ovary, is protected by a leathery cup-shaped calyx. Most corollas curve subtly downward above the calyx. The circular nectary averages 4.0 mm in diameter (Elias & Gelband, 1976). The corollary stricture marks the point where the narrow basal tube flares abruptly. The four staminal filaments and a short curled staminode emerge from the interior wall at the corollary stricture and curve along the inner walls toward the opening. The anthers are arranged in a didynamous configuration, with a proximal and distal pair. The dorsifixed anthers are positioned along the dorsal wall inside the corolla opening. The style protrudes through and partially blocks the aperture of the stricture (Figure 2). The flattened, approximately diamond-shaped stigma usually lies between the proximal and distal pairs of anthers along the dorsal wall of the floral tube.

Trumpet creeper flowers contain an anthocyanin pigment, cyanidin 3-rutinoside (Harborne, 1967), resulting in rich orange and red hues on the internal and external surfaces of the corollas (Figure 2), typical of hummingbird-pollinated flowers (Grant & Grant, 1968). Contrasting red to magenta nectar guides extend from the reflexed petal lobes at the mouth of the corolla inward past the anthers and stigma toward the corollary stricture (Figure 2).

Individual flowers can yield nectar volumes exceeding 115 μL/day (Bertin, 1982a; Edge, 2010; Van Nest et al., 2021), the highest average volume reported for ornithophilous plants in North America at temperate latitudes (see McDade & Weeks, 2004; Stiles, 1975; Stiles & Freeman, 1993). Flowers typically open between 06:00 and 14:00 h (Bertin, 1982a). Newly opened flowers have an average initial nectar volume of 40 μL. Nectar production continues at a rate of 5 μL/h throughout the day, slows abruptly at sunset, and ceases the following day (Bertin, 1982a). Trumpet creeper flowers, like those of most ornithophilous species, have little or no scent, at least to human olfaction. The sugar concentration of floral nectar, in sucrose equivalents, ranges from 24% to 35%, averaging approximately 30% (Bertin, 1982a; Edge, 2010; Van Nest et al., 2021). This concentration is slightly higher than the average 25.4% sugar observed in 202 species of hummingbird-pollinated plants but significantly lower than the average 41.6% observed in 156 bee-pollinated plants (Pyke & Waser, 1981). Floral nectar sugar of trumpet creeper is predominately sucrose (73%), with lesser percentages of fructose (18%) and glucose (9%) (Baker et al., 1998). In a sample of 137 hummingbird-pollinated species, Baker et al. (1998) found a range of 18–97% sucrose, with an average of 58%. Extrafloral nectaries have been identified on the petiole, calyx, and the outer surface of developing corollas (Elias & Gelband, 1975, 1976). These are visited by ants. There is no evidence that hummingbirds exploit extrafloral nectar, which must be miniscule in calorie equivalents compared to the copious floral nectar.

2.3 Floral data

I stored newly opened flowers in plastic bags and transported them in a refrigerated cooler for later measurement and dissection. Insect-damaged flowers were culled in the field, while undersized flowers were retained in the population samples. Most floral measurements were made with digital calipers to the nearest 0.1 mm. The width of the corolla opening (horizontal plane) was measured first. The corolla was then split distally along the ventral midline far enough to measure the position of the stigma in relation to the proximal and distal pairs of anthers. The calyx, nectary, ovary, and stigma were then removed as a single unit. Corolla length (minus calyx) was measured from the base, which approximates the position of the nectary, to the notches flanking the ventral petal lobe at the mouth of the floral tube. The corolla was then split from end to end along the ventral midline, pressed flat under a sheet of glass, and photographed alongside a scale bar (Figure 2). The distance from the corolla base to the corollary stricture, approximated by the base of the distal anther filaments, was measured from scaled photographs (to the nearest mm). I also measured the distance from the corolla base to the filament tips of the longest proximal and distal pairs of anthers. The stigma-nectary distance was derived from the position of the stigma relative to the distal pairs of anthers.

2.4 Hummingbird foraging behavior

I observed ruby-throats foraging at the White River site during three flowering seasons: July 9, 2020, to August 8, 2020, June 30, 2021, to August 11, 2021, and July 29, 2023, to September 20, 2023. Hummingbirds at this site were habituated to humans and commonly fed at trumpet creepers within 2 m of seated observers. None of the hummingbirds was marked or tagged. Video recordings of foraging were made with a Nikon Coolpix P1000™ camera. Still photography was also conducted with a Canon Eos 80D™ equipped with a 400 mm lens.

2.5 Statistical analysis

The pollination biology literature provides limited guidance on the expected statistical distribution of floral trait measurements in ornithophilous plants. To address this gap, I calculated summary statistics for six key floral measurements that bear directly on hummingbird pollination and nectar extraction in trumpet creeper (Appendix 1). The D'Agostino-Pearson K² test (hereafter K² test), which combines the skewness ($$$ \sqrt{b_1} $$$) and kurtosis (b₂) statistics, was used to evaluate the goodness-of-fit of floral trait measurements to a normal distribution (D'Agostino et al., 1990). The K² test has robust power over a wide range of sample sizes and is preferred over the Shapiro–Wilk W test (Shapiro & Wilk, 1965) when data sets include repeated values. I report φ coefficients as a measure of the degree of similarity between observed distributions and a normal distribution. The values of φ range from −1 to 1, where zero indicates perfect conformity to a normal distribution.

None of the six floral traits displayed significant skewness (Appendix 1). Kurtosis was observed in a single trait in the White River population. However, the omnibus K² test indicated that the distribution of several traits in both populations deviated significantly from normality. Corolla length (Figure 3) was notable in exhibiting high χ² values and correspondingly low p-values in both populations, resulting from peaked and left-skewed distributions. Overall, floral traits in both populations exhibited low to moderate φ coefficients (observed values, 0.10–0.55).

Rather than employing data transformations to normalize trait measurements, I used the nonparametric Spearman rank correlation coefficients (r_s) to evaluate the strength of correlations between traits. Population differences in traits were evaluated with the nonparametric Mann–Whitney U-test (Hollander & Wolfe, 1973). This test combines measurements from the White River and Norfork populations into a single sample and then assesses the range and location of the lowest group's distribution within the overall sample range against a ranked distribution that approaches normality. All p-values are two-tailed. To reduce the risk of Type 1 errors resulting from multiple comparisons, a Bonferroni correction was applied to similar analyses within each population. The significance level was adjusted by the number of tests, α = 0.05/6 = 0.0083.

3.1 Floral trait variation

Corolla length in the two populations ranged from 58 to 75 mm (Appendix 1). The minimum length represents the shortest distance that must be bridged by hummingbirds or bees to reach the nectary through the corolla opening. Median corolla length and opening width were 7.6% longer and 4.2% wider, respectively, in the White River population than at Norfork. Corolla length and width were marginally correlated at White River (r_s = 0.23, p = .01, n = 115) and less so at Norfork (r_s = 0.15, p = .09, n = 123). Differences in flower size between the two study sites may be partly attributed to site favorability, Norfork being higher and drier than White River.

The positions of the stigma, proximal anthers, and distal anthers, with respect to the nectary, were relatively similar in both populations (Appendix 1). Anthers and stigma were inserted well inside the corolla opening. The stigma was usually located between the proximal and distant pairs of anthers (Figure 3). The stricture-nectary distance was similar in the two population samples (Mann–Whitney, p-value = .83).

The proximal anther-nectary distance exhibited a marginal correlation with corolla length at White River (r_s = 0.28, p = .002, n = 113), but a stronger correlation at Norfork (r_s = 0.48, p < .0001, n = 123). In both populations, there were strong correlations between corolla length and distal anther-nectary distance (White River, r_s = 0.45, p < .0001, n = 113; Norfork, r_s = 0.53, p < .0001, n = 123). The stigma-nectary distance was also significantly correlated with corolla length at White River (r_s = 0.37, p < .0001, n = 113) and Norfork (r_s = 0.26, p = .003, n = 123). All floral traits, external and internal, exhibited low coefficients of variation within populations (0.04–0.09).

3.2 Hummingbird foraging behavior

Ruby-throats were observed at trumpet creeper flowers at the White River site from mid-June to mid-September. The frequency of visits increased in August coincident with the departure of orchard orioles (Icterus spurius), which compete for nectar and pierce flowers, and the augmentation of local hummingbird populations with hatching year birds. Ruby-throats typically explore multiple inflorescences during foraging visits, making hovering passes around flower clusters before approaching closely. This surveillance likely serves multiple functions, including assessing bud development and flower maturity, locating oriole piercings as shortcuts to nectar, and checking for potential predators.

After selecting a flower, a ruby-throat hovers forward and grasps the ventral petal lobe at the corolla opening with both feet (Figure 4). The curved claws penetrate the soft spongy petal surface, providing leverage to thrust its head and upper body into the trumpet-shaped corolla. When the fit is snug, individuals may become wedged or “caught” as observed by Catesby (1731). During these instances, wing movements typically pause, with the wings either folded or held in a posterior-facing delta position (see video URL in Data Availability Statement). Ruby-throats appear to employ leg thrusts and wing beats to back out of the corolla. Although scores of hummingbird species are known to insert the full length of their bills into flowers (see eBird photo archives, Media Search–Macaulay Library and eBird), until now, none have been documented to consistently thrust their entire head and upper body into snug tubular corollas while clinging. I propose the term “floral-diving” to characterize this distinctive feeding technique, which was the only method ruby-throats used to obtain nectar through the corolla opening at the Ozark study site.

As ruby-throats enter and exit the corolla, their heads brush the stigma and both pairs of anthers, which are positioned well inside the flower opening. The snug fit and the position of the anthers near the dorsal midline ensure precise deposition of pollen on the hummingbird's crown. Individuals feeding at trumpet creeper flowers are dusted and often caked with pale yellow pollen (Bertin, 1982a). This foraging mode may require horizontal or upward-facing flowers, as grasping the petal might be difficult from an inverted position (but see Reid et al., 2023). Hummingbirds may perform the floral-diving maneuver 2–3 times in rapid succession at the same flower. These quick entrances and exits might reduce the risk of ambush by predators such as mantids (Nyffeler et al., 2017), vespine wasps (Carr & Golinski, 2020), and birds (Graves, 2022b), though this remains speculative.

Prolific flowering and the abundance of floral nectar and pollen make trumpet creeper highly appealing to hummingbirds and bees (Bertin, 1982a). However, the copious nectar is also irresistibly attractive to the nectar-robbing orchard oriole (Graves, 2022a). Orioles use their bills to penetrate the corolla and the leathery calyx to access nectar. The slit-like piercings typically measure 12–20 mm in length and extend through the base of the corolla parallel to the long axis of the flower. During the peak blooming season, orioles robbed more than 92% of flowers at the White River site before they departed in late July (Graves, 2022a). Hovering ruby-throats readily visit oriole piercings and empty calyces after the corollas have been detached by orioles. While ruby-throats are known secondary nectar-robbers, there have been no documented observations of ruby-throats piercing trumpet creeper corollas themselves, although this remains a possibility (see Inouye (1980) for discussion of primary and secondary nectar-robbery).

3.3 Bill length and nectar accessibility

The conventional wisdom that corolla length of hummingbird-pollinated plants matches the bill length of its primary pollinator (Dalsgaard et al., 2009; Maglianesi et al., 2014; Rico-Guevara et al., 2021; Snow & Snow, 1980; Temeles et al., 2009) does not apply in this case. Trumpet creeper corollas are roughly three times the length of the longest bills of female (18.3 ± 0.7 mm) and male (16.4 ± 0.6 mm) ruby-throats (bill measurements from Weidensaul et al., 2020; Temeles, 1996) used feeder experiments to demonstrate that the maximum depth of nectar extraction for female ruby-throats, with fully extended tongues, was approximately 40 mm when feeding from 3 mm diameter tubes. This indicates that ruby-throats must insert their bill tips about 36–53 mm past the corolla opening for their tongues to reach the circular nectary at the base of the tubular trumpet creeper flower. This distance may be a few millimeters less if nectar is pooled above the nectary, but the hummingbird's head must be fully inserted in either case. The eyes of ruby-throats, located 22–24 mm behind the bill tip, are positioned 12–29 mm inside the tubular corolla during floral-diving. While it is unclear whether their eyes are closed while feeding, the birds are certainly blinded and highly vulnerable to predators during this time.