Dalechampia is a pantropical genus of climbers and shrubs comprising over 130 species (Armbruster et al. Reference ARMBRUSTER, EDWARDS, HINES, MAHUNNAH and MUNYENYEMBE1993). The plants are monoecious, but with male and female flowers closely adjoined in blossom-like pseudanthia, inflorescences functionally equivalent to a hermaphroditic flower. Two large, showy bracts (modified leaves) subtend the flowers and act like petals to attract pollinators (Armbruster Reference ARMBRUSTER1993; Pérez-Barrales et al. Reference PÉREZ-BARRALES, BOLSTAD, PELABON, HANSEN and ARMBRUSTER2013). All 10–12 known species of Dalechampia in Madagascar are endemic to the island; they can be divided into two groups based on the openness of their male flowers. Open-flowered species exhibit the structure of male flowers found in nearly all non-Madagascan species of Dalechampia, in which the pollen is openly accessible to pollinators. In contrast, closed-flowered species have male flowers in which the sepals remain fused except at their tips (Armbruster et al. Reference ARMBRUSTER, LEE, EDWARDS and BALDWIN2013), making the flowers functionally equivalent to poricidal anthers found in other plant groups. These structures restrict access to pollen to insects capable of buzz-pollination, the collection of pollen by vibrating (‘buzzing’) the flowers at high frequencies.

In addition to floral features, extra-floral structures can be important in determining plant-pollinator interactions in Dalechampia. Here we focus on the potential role of bract motility, i.e. circadian changes in the angle between bract pairs, in the pollination ecology of Dalechampia aff. bernieri, endemic of Madagascar. Bracts of most Dalechampia species show some degree of bract motility, opening during the day and closing partially or totally at night (Armbruster & Steiner Reference ARMBRUSTER and STEINER1992). Bract motility is generally associated with herbivory avoidance (Armbruster & Mziray Reference ARMBRUSTER and MZIRAY1987), but it can also play a role in partitioning pollinators temporally, making the flowers accessible to pollinators only during part of the day (Armbruster & Herzig Reference ARMBRUSTER and HERZIG1984). In contrast to Dalechampia species elsewhere, the relationship between bract motility, floral structural specialization and pollination ecology of those found in Madagascar has received little attention.

Dalechampia aff. bernieri (Dalechampia sp. nov. morondava, sensu Armbruster et al. Reference ARMBRUSTER, EDWARDS, HINES, MAHUNNAH and MUNYENYEMBE1993) is a twining climber with white pseudanthial bracts, a weak floral fragrance and ‘closed’ male flowers (Armbruster et al. Reference ARMBRUSTER, LEE, EDWARDS and BALDWIN2013; Figure 1a). Because its male flowers do not open up fully at anthesis, D. aff. bernieri is expected to be buzz-pollinated, although buzz-pollination has not yet been observed in D. aff. bernieri (Armbruster et al. Reference ARMBRUSTER, EDWARDS, HINES, MAHUNNAH and MUNYENYEMBE1993).

Figure 1. An inflorescence of Dalechampia aff. bernieri being visited by the buzz-pollinating bee Nomia viridilimbata (a). Daily changes in bract aperture observed for Dalechampia aff. bernieri (b). Solid line: segmented regression fit.

We expected bract angles of D. aff. bernieri to change during the day, being wider (bracts more open) when the (predicted) buzz-pollinating bees are more active and bracts being more closed when these pollinators have lower activity, to protect the inflorescence's flowers from herbivores and/or less efficient pollinators.

The study was conducted at Kirindy Forest (30–60 m asl), a seasonal dry forest approximately 60 km north-east of Morondava, in central-western Madagascar (20.06706°S, 44.65732°E), 5–19 November 2012. Eleven plants of D. aff. bernieri were surveyed and marked. On each of the plants, 1–12 inflorescences were individually marked, for a total of 47 inflorescences. All plants were within 5 km of each other, no further than 30 m from the side of Conoco road, the main road running through Kirindy forest. For each inflorescence, the angle between the bracts was recorded one to four times each day between 05h00 and midnight, for up to 6 d. The angles between bracts were assessed visually, estimated to the nearest 5°, by the same person (MP) during the whole study. The identity and abundance of insects visiting the inflorescences were assessed by direct observation. Pollinator observation was opportunistically carried out before, during and after measuring the bract angles, for up to 3 h d⁻¹ per plant. Specimens of the observed flower visitors were captured, photographed and identified to the lowest possible taxonomic level using Pauly et al. (Reference PAULY, BROOKS, NILSSON, PESENKO, EARDLEY, TERZO, GRISWOLD, SCHWARZ, PATINY, MUNZINGER and BARBIER2001).

All statistical analyses were performed using R version 3.0.2. The temporal change in bract angle was analysed with a linear mixed-effect model, with plant identity as a random effect, using function lmer from the package lme4; a second-degree polynomial model was used to assess the non-linear relationship between the bract angle and the time of the day. In case a statistically significant non-linearity was detected, we fitted a segmented regression model to estimate the time of the day at which bract aperture was maximum; this was done using function segmented from the package segmented (Muggeo Reference MUGGEO2003).

Overall, the bracts of D. aff. bernieri showed statistically significant non-linear variation in angle with time of day according to the second-degree polynomial model: bract aperture (degrees) = 49.15° + 4.96° h − 0.23° h² (SE equals 7.50°, 0.99° and 0.04°, respectively), whereas inflorescences from different individuals showed differences in degree of bract movement. According to the one-break-point segmented regression model, the angle between bracts of D. aff. bernieri inflorescences was 50.2° (SE = 5.16°) at 00h00 and increased by 3.34° h⁻¹ (SE = 0.66° h⁻¹) toward a peak of ~90° at 10h45, followed by a decrease in bract aperture by 3.75° h⁻¹ (SE = 0.58 ° h⁻¹) (Figure 1b).

Dalechampia aff. bernieri was visited by Nomia viridilimbata (Halictidae: Nomiinae), which collected pollen by buzzing the flowers, transferring pollen to the stigmas in the process (Figure 1a). This is the first observation of Nomia bees pollinating Dalechampia flowers and the first documentation of buzz-pollination of D. aff. bernieri, which confirms the expectation for the species’ pollination strategy based on the structure of its male flowers. Dalechampia aff. bernieri was also visited by Liotrigona sp. bees (Apidae: Meliponini). These small, dark, stingless bees are incapable of buzz-pollination and, in contrast to reports from previous studies (Armbruster et al. Reference ARMBRUSTER, EDWARDS, HINES, MAHUNNAH and MUNYENYEMBE1993), we never observed them contacting the female flowers; these bees were thus acting as pollen thieves rather than pollinators.

Our study is the first attempt at measuring circadian bract movement quantitatively in Madagascan Dalechampia. The circadian patterns of bract opening of D. aff. bernieri match the activity patterns of its observed and expected pollinators, as the temporally restricted diurnal bract opening of D. aff. bernieri matches the diurnal foraging of specialized Nomia bees. Our results strengthen the suggestion that bract movement could be related to specialized floral structures that restrict the access to pollen to a few species of pollinator.

Bract motility might be an element of a pollination syndrome, such that closed-flowered species have motile bracts to favour visitation by diurnal buzz-pollinating insects, while the bracts of open-flowered species always form a wide angle to attract unspecific pollinators at any time of day and night. During our study we could also conduct some observations on D. subternata, an open-flowered species known to be pollinated by generalist insects. The bracts of this species show little motility and form a ~180° wide angle around the clock (Armbruster et al. Reference ARMBRUSTER, EDWARDS, HINES, MAHUNNAH and MUNYENYEMBE1993; the authors’ unpubl. data), thus supporting our hypothesis; yet, this anecdotal speculation will require appropriate testing based on a larger amount of observations and on a comparison among all the Madagascan species of Dalechampia.

Future studies should delve deeper into the ecological relevance of bract motility in D. aff. bernieri. Although the daily changes in bract angles were statistically significant, we did not test for the biological significance of bract motility; bracts that form wide angles may attract pollinators more effectively than bracts at narrower angles by increasing the inflorescence visibility, but the angles observed at night may still be too wide to preclude insect visits to the flowers. This should be addressed experimentally, with more extensive pollinator observations, and by assessing the correlation between individual fitness (i.e. seed production) and the extent of bract motility, which varies markedly between individual plants.