POLLINATION AND REPRODUCTIVE BIOLOGY OF THIRTEEN SPECIES OF BEGONIA

The reproductive biology of 13 monoecious species of Begonia L. (Begoniaceae) that occur in the Serra do Mar State Park, São Paulo, Brazil, was investigated. These species flower annually and present flowers with mostly white tepals, light and sweet odour, pollen as a reward but no nectar, numerous yellow stamens, and coiled yellow styles. Anthesis is diurnal and floral duration is long (6 15 days). The unusual appearance of pistillate flowers of these species supports the view that they are intersexual mimics of the staminate flowers. Despite consistently high levels of fruit-set, none of the 11 species tested proved to be apomictic. In contrast to earlier reports of self-compatibility in Begonia, self-pollinations of B. integerrima and B. itatinensis produced no fruits or seeds, and the complete absence of pollen tubes in the styles of self-pollinated flowers of B. integerrima suggests that the species is genetically self-incompatible. Flowers pollinated under natural conditions showed many pollen tubes that reached ovules, suggesting that adequate numbers of compatible pollen grains had been transported by pollinators. The principal pollinators were small bees of the Apidae and Halictidae. Ten species of bees were observed to visit eight species of Begonia, and pollen collection occurred by means of vibration, except for Trigona spinipes. Visits to rewarding staminate flowers were significantly more frequent than visits to unrewarding pistillate flowers. Duration of visits to pistillate flowers also was significantly shorter than the duration of visits to staminate flowers. We conclude that visits to pistillate flowers occur by mistake but with sufficient frequency to allow for successful reproduction in natural populations of these species.


INTRODUCTION
Plant-pollinator interactions involve energy investment in flowers, which offer rewards to facilitate removal and deposition of pollen by visitors (Faegri & Pijl 1979).These interactions, however, are not always mutualistic, as in deceit pollination (Baker 1976).Studies of the interactions between plants and their pollinators and the evolution of morphological characteristics that mediate these interactions can provide useful information to those charged with conserving rare species in threatened habitats (Kearns et al. 1998), such as the Atlantic Rainforest of Brazil (Alves dos Santos 2003).
There are two genera within the family Begoniaceae: Begonia and Hillebrandia (Forrest & Hollingsworth 2003).Begonia is among the largest angiosperm genera, with roughly 1 500 species distributed principally in tropical and subtropical regions (Goodall-Copestake et al. 2009), whereas Hillebrandia is monotypic and endemic to the Hawaiian Islands (Clement et al. 2004).Recent estimates put the number of species of Begonia in Brazil between 200 to 240, with highest diversity in the Atlantic Rainforest (Silva & Mamede 2001;Jacques 2002).In the Serra do Mar State Park, 24 species have been recorded (Silva & Mamede 2001).
Most species of Begonia are monoecious herbs with cymose inflorescences (Givnish 1980;Burt-Utley 1985;Ågren & Schemske 1991).They present staminate and pistillate flowers in different phases of anthesis (Ågren & Schemske 1993), with staminate flowers usually open before pistillate flowers (Givnish 1980;Jacques 2002).The tepals and the style/stigma of the pistillate flowers typically resemble the tepals and androecium of the staminate flowers, both in size and colour (Schemske & Ågren 1995;Schemske et al. 1995).There exist variations in floral morphology among species, however, that may provide answers to questions regarding their pollination biology and reproduction.For example, the wide variation in tepal length among species of section Pritzelia suggests that some species are pollinated by small bees, such as halictids, whereas other species are pollinated by much larger bees, such as euglossines.
Pollination of Begonia by bees was suggested by Seitner (1976), Wiens (1978), Faegri and Pijl (1979), and Givnish (1980), and bees are probably the main pollinator group of Begoniaceae.But pollination by birds occurs in some species, such as Begonia ferruginea (section Casparya) of Colombia, whose red, tubular pistillate flowers produce nectar and have been observed to be visited by hummingbirds (Vogel 1998).Species of section Symbegonia also have tubular flowers and are visited by nectarivorous sunbirds (Forrest et al. 2005).
We suggest that several factors play a role in the rate of deceit (i.e., the proportion of visits to unrewarding pistillate flowers).First is the ability of different pollinator species to discriminate between staminate and pistillate flowers.There may also be differences within a given species of pollinator between naïve and experienced bees, depending on their ability to learn.The rate of deceit may also be related to the species of Begonia, especially with respect to the timing and presentation of staminate and pistillate flowers.Species of Begonia in which staminate and pistillate flowers are presented simultaneously and which maintain a high ratio of staminate to unrewarding pistillate flowers are expected to show higher rates of deceit.
Little information exists regarding the reproductive biology of Begonia.It has been suggested that the genus is selfcompatible (East 1940;Burt-Utley 1985), and Ågren and Schemske (1993) confirmed self-compatibility (geitonogamy) in five species of section Doratometra.Burt-Utley (1985) and Jacques (2002) have reported natural hybrids between some species of Begonia.
We studied pollination and reproductive biology of 13 species of Begonia in Brazil that occur in the Atlantic Rainforest on the coast of the state of São Paulo.The specific objectives were to determine reproductive phenology, breeding system, and reproductive success under natural conditions and to record information on floral biology and floral visitors.The species were selected based on population size in the area of study, length of flowering period, and accessibility of flowers.This approach allowed us to compare and contrast a range of species with respect to a set of basic questions.What are the ratios of staminate to pistillate flowers and how do these change over time?Do these ratios change the behaviour of pollinators and lead to lower or higher levels of pollination and fruit-set?Do pollinators differ in their ability to discriminate between staminate and pistillate flowers?

MATERIALS AND METHODS
This study was conducted in areas of dense, humid forest in the Núcleo Picinguaba (23° 31´ to 23° 34´ S and 45° 02´ to 45° 05´ W), Serra do Mar State Park.The elevation varies between sea-level and 1 340 m.The climate of the region is hot and extremely humid, with average temperatures above 18°C and with a drier period in winter (Nimer 1977).According to Köppen (1948) the local climate is classified as Rainy Tropical type Af, having high precipitation during all months of the year.Part of this study was conducted in an area of dense, humid montane forest in the Núcleo Santa Virgínia (23° 17´ to 23° 24´ S and 45° 03´ to 45° 11´ W), Serra do Mar State Park.The elevation varies between 850 and 1 100 m above sea-level, and the climate is Temperate Tropical (Cwa following the classification of Köppen 1948), with a mean annual precipitation in excess of 2 000 mm.In the driest months, the precipitation never falls below 60 mm (Setzer 1966).
The following species of Begonia were studied: B. caraguatatubensis Brade, B. cucullata Willd.var.cucullata (Fig.  1a, b) To determine the flowering phenology of the species studied, flowering periods were recorded for approximately 50 individuals between March 2007 and July 2009.The pattern of flowering of these species was determined according to the phenological patterns defined by Newstrom et al. (1994).
Their classification provides a logical system to describe the full range of patterns quantitatively (e.g., annual, subannual, continuous).For B. caraguatatubensis, B. dentatiloba, B. fernandocostae, B. aff.fluminensis, and B. integerrima, the number of flowers opening on a determined day was measured.Floral measurements of both staminate and pistillate flowers were taken from Silva and Mamede (2001) for all species studied.Biological characteristics of the flowers of all species, such as time of anthesis, sequence of staminate/pistillate anthesis and duration, presence of odour, tepal colour, and receptivity of the stigma (tested with 10 M hydrogen peroxide) were determined directly in the field (Dafni 1992;Kearns & Inouye 1993).Pollen viability was tested in B. cucullata and B. aff.fluminensis by observing colouration of the cytoplasm with a 1.2% solution of carminic acid (Radford et al. 1974).Other characteristics such as floral symmetry, number of flowers per inflorescence (mean ± standard deviation), form and size of tepals, stamens, stigma, and capsules were noted.
To gauge the effectiveness of pollinators, pollen tube analysis was done for 12 species of Begonia, using 7 to 20 pistillate flowers collected from natural populations.The analysis of pollen tubes in flowers of B. luxurians was not possible due to the rigidity of the ovary wall tissue.The opened pistillate flowers were collected under natural conditions (i.e., available to pollinators) before tepal senescence and preserved in a solution of 50% ethanol.The gynoecia were washed with distilled water and placed in a 2 % solution of sodium hydroxide, where they remained within an oven at 49°C for 5 to 20 minutes until the stigma, style, and ovary softened and cleared sufficiently (Iara Bressan, pers. com.).Slides were stained using an aniline blue solution (Martin 1959), and an Olympus BX51 fluorescence microscope was used to observe pollen tubes in the stigmas and ovaries.Images were captured using an Evolution®MP digital microscope camera from Media Cybernetics.Data from each period of observation were analyzed using BioStat 3. A Mann-Whitney U-Test was used to test statistical differences between duration of visits to staminate flowers and duration of visits to pistillate flowers.Histograms were created using Systat 11.For each species of visitor, the rate of deceit was calculated as the proportion of all visits that are made by mistake (i.e.visits to pistillate flowers rather than rewarding staminate flowers).The importance of each visitor as a pollinator to each species of Begonia was determined based on frequency of visits to pistillate flowers and behaviour during these visits.

Phenology, morphology, and floral biology
Three patterns of reproductive phenology were identified among the 13 species of Begonia.Begonia cucullata is the only species with a continuous pattern of flowering, with a long peak of approximately six months and a reduction in the number of flowers produced during the rest of the year (Fig. 2).The species that present a sub-annual pattern of flowering All 13 species of Begonia that were studied have cymose inflorescences with pistillate and staminate flowers in the same inflorescence (Fig. 3a-c) or in separate inflorescences (Fig. 3d).
Begonia lanceolata has staminate flowers in dichasial cymes and depauperate, subsessile pistillate flowers (Fig. 3d, e For example, B. integerrima has 12 ± 6 flowers per inflorescence (n = 10 inflorescences), but some of the staminate flowers open soon after the pistillate flowers.For this species there is an overlap of the two types of flowers.
The flowers are of the open type (sensu Faegri & Pijl, 1979) and present white, white to pink, white with a red centre, or white to greenish-yellow tepals (Tab.1).There is no apparent change in colour following anthesis or prior to senescence.In B. integerrima the bases of tepals in both staminate and pistillate flowers possess a red colouration (Fig. 1i, j).As with most Begonia species, pollen acts as the sole reward to visitors for all species studied.The number and size of the tepals of staminate flowers differ little from pistillate flowers, with the exception of size in B. itatinensis (Tab.1).
Aside from minor morphological differences in the tepals, staminate and pistillate flowers (Fig. 1) are quite similar for all species.The stigmas can be localized in strips or spread over large areas covered with papillae, occurring along the branches of the styles, which are yellow in colour (with the exception of the branches of the styles of B. integerrima, which are white to pale-yellow; are six in number and have a spiral form (corkscrew; Fig. 4d).Stigmas remain receptive throughout the life of the flower.Stamens may be free (Fig. 3f, 4a) or united to form a column (Fig. 3d, arrow), and the number varies greatly among species (Tab.1).Anthers are yellow and rimose (Fig. 4a), with the size of the axial aperture varying among species.Poricidal anthers were observed in only one of the thirteen species: B. integerrima (Fig. 4b).The pollen in all species studied is diminutive, dry, powdery, and whitish-yellow.
Pollen viability is high in B. cucullata (82%) and B. aff.fluminensis (78%).We detected floral odour in nine of the 13 species (Tab.1).Odours are lightly sweet and seemed identical for staminate and pistillate flowers of a given species.We did, however, detect differences between species.Odour is strongest between 0700 to 1100 hours, corresponding to the peak hours of visitation.

Visitors: Hymenoptera
Ten species of bees were observed, collected, and identified visiting flowers in various species of Begonia.These were all Halictidae (Augochlorodes sp., Augochloropsis sp. 1 and sp. 2, Neocorynura sp. 1 and sp.2) or Apidae (Paratetrapedia sp., Trigona spinipes, Euglossa sp., Melipona bicolor, Melipona quadrifasciata), and small to medium in size (0.3 to 1 cm).In addition, bees of large size, belonging to the genera Bombus, Xylocopa, Centris and Epicharis were observed visiting flowers of B. fernandocostae and B. integerrima, but these visits were rare and it was not possible to make collections.
The most common bee visitor to flowers across all species was Trigona spinipes, which was recorded on six species (Tab. 3).On staminate flowers, this species was observed using legs and mandibles to remove pollen from stamens.On pistillate flowers Trigona spinipes made short visits, consisting of a brief landing and occasionally longer visits in which a similar behaviour to staminate flower visits was observed: gathering movements elicited by the similar floral display which resulted in pollen deposition and/or removal of pollen deposited from earlier visits.Visit durations to staminate flowers varied from 1 to 107 seconds, whereas visits to pistillate flowers varied from 1 to 36 seconds.Trigona spinipes made visits in irregular intervals, with a flower-to-flower movement slower than the  7b).The rate of deceit of Paratetrapedia sp. in flowers of B. caraguatatubensis was 16.9% (Tab. 3).In this case, the staminate flowers were visited 4.35 times more often than pistillate flowers, even when there were fewer staminate than pistillate flowers (448:647).One month later, when there were almost four times as many pistillate flowers as staminate flowers (1:3.95), the rate of deceit diminished to 4.17%.
Similarly, the total number of visits also diminished over this period (Fig. 7c).Statistical analysis revealed a significant difference between visit duration to staminate versus pistillate flowers, when the ratio of staminate to pistillate flowers was 1:1.44 (U = 1216, Z = 5.62, P < 0.001; Fig. 7c).Moreover, considering all visits during the entire flowering period, there is a significant difference in visit duration of Paratetrapedia sp. to staminate versus pistillate flowers (U = 1216, Z = 6.11,P < 0.001).
Bees of the genus Melipona were only observed on flowers of B. integerrima, and represent the most abundant and regular visitor to this species (Tab.3; Fig. 6a, b), principally on individuals in the canopy.These bees collected pollen by vibration.The two species of Melipona made visits to 180 staminate flowers and 11 pistillate flowers during four hours of observation.Melipona bicolor was more discriminating than M. quadrifasciata, with a percentage of deceit of 2.07%, whereas M. quadrifasciata had a percentage of deceit of 16.7% (Tab.3).Duration of visits to staminate flowers was significantly longer than on pistillate flowers (U = 286, Z = 3.97, P < 0.001; Fig. 7d).
Although Euglossa sp. was not a common visitor to these species of Begonia, it was the third most frequent visitor to B. integerrima and the fourth most frequent to B. aff.fluminensis.
Pollen was collected by vibration, and this species visited only the largest-flowered species of Begonia.This pattern also was observed among bees of larger size, such as Bombus, Xylocopa, Centris, and Epicharis, which visited only the larger-flowered B. fernandocostae and B. integerrima.The other species of bees observed in this study showed low visitation frequencies, but all made visits to both staminate and pistillate flowers.
Although we were not able to control for all factors that may play a role in rates of deceit, our data show that rate of deceit tends to decrease with increasing frequency of visits made by a species of pollinator.Moreover, the rate at which a species of pollinator is deceived seems to be consistent across species of Begonia visited (Tab.3).The number of prior visits by an individual pollinator may also play a role in its ability to discriminate between pistillate and staminate flowers, thereby also affecting frequency of visits to unrewarding pistillate flowers.

Pollination by vibration
In the species of Begonia studied, staminate flowers open before pistillate flowers, and pollen serves as the only reward.
Bees therefore become habituated to visiting staminate flowers, in which vibration is used to collect pollen (except for Trigona spinipes).At pistillate flower anthesis, the bees make "mistake" visits, being deceived by the similarity between the branches of the style/stigma of the pistillate flowers and the stamens of the staminate flowers.Deceit is evidenced by the attempt to collect pollen during visits to pistillate flowers: vibration and manipulation of the style by the legs and mandible of the visitor (Fig. 6).When the bees vibrate on the pistillate flowers, grains of pollen on the body are released in a visible cloud and adhere to the stigma.Trigona spinipes X X 10.2% (59) X X TABLE 3. Percentage of visits to pistillate flowers (rate of deceit) for pollinators of eight species of Begonia in the Serra do Mar State Park, São Paulo, Brazil.The total number of visits observed is in parentheses."X" indicates observed visits by that pollinator to that species of Begonia, but no quantitative data were collected.Species of Begonia are designated by the first three letters of the specific epithet.

Other visitors
Besides bees, species of Diptera (Syrphidae), Coleoptera, and Lepidoptera were observed visiting both staminate and pistillate flowers of Begonia, but with lower frequency of visits (Diptera and Lepidoptera), shorter visits (Lepidoptera), and a general failure to pick up and deposit pollen (personal observations).On staminate flowers, Syrphids removed pollen directly from the anthers and also collected pollen that had fallen onto the tepals.A single visit from a Syrphid fly had a duration of 16 minutes and five seconds to a staminate flower of B. fernandocostae.On pistillate flowers, Syrphid flies behaved similarly, possibly removing pollen deposited by earlier visits.Species of Coleoptera were very common on Begonia flowers, especially on staminate flowers, where they consumed pollen.The infrequent visits of Lepidopterans were always short (<1 sec) and each of the five individuals observed made only one visit.

Phenology and Floral Biology
The 13 species of Begonia show a range of phenological patterns, from annual, to subannual, to continuous, as defined by Newstrom et al. (1994).Differences in reproductive phenology are evident even between closely related species, such as B. fernandocostae and B. aff.fluminensis, both of which belong to section Pritzelia.
Large differences in flower size also occur among taxonomically closely related species: Begonia caraguatatubensis and B. hookeriana, which produce many small flowers, versus B. fernandocostae, B. aff.fluminensis, and B. valdensium, which produce few, large flowers, all belong to section Pritzelia (Doorenbos & de Wilde, 1998).These differences are probably the result of selective forces imposed by pollinators.
On the other hand, differences in size and number of flowers could be phylogenetically fixed (Schemske & Ågren 1995).Schemske and Ågren (1995) speculated that a decrease in the number of flowers per inflorescence in exchange for larger individual flowers may be disadvantageous, because number of flowers per inflorescence may be critical in attracting pollinators.Nevertheless, both large-flowered (small inflorescence) and small-flowered (large inflorescence) species studied attracted similar numbers of pollinators and frequencies of visits, resulting in high reproductive success.
The morphology of all species with respect to number and size of tepals in staminate versus pistillate flowers helps to deceive pollinators due to the high similarity between both types of flowers.The one exception is B. itatinensis, the tepals of whose pistillate flowers are twice the length of the tepals of the staminate flowers.Bees visiting Begonia have been observed to prefer larger flowers (Schemske & Ågren 1995).The number of observed visits to pistillate flowers is always less than to staminate flowers.This is possibly related to the ability of bees to learn to recognize unrewarding flowers and avoid them.This ability exists not only in honeybees, but also in other social and solitary bees as well (Menzel 2001).Yet even with a lower number of visits, pistillate flowers of each species appear to receive sufficient visits to ensure high fruit-set.Prolonged receptivity of stigmas and long lifespan of flowers likely allow the plant to weather low pollinator visitation, as in B. gracilis (Castillo et al. 2002).Additionally, the long duration and receptivity of pistillate flowers may facilitate pollination, since greater floral longevity would likely mean a greater chance of attracting and deceiving visitors if they are open and receptive for a long period (Schemske et al. 1995;Jacques 2002).Since pistillate flowers depend entirely on mistake visits by pollinators, floral odour should have an important role (Ågren & Schemske 1991;Schemske & Ågren 1995).The similarity of odour between staminate and pistillate flowers detected in this study would be expected to facilitate deceit.The spiral branches of the style increase the stigmatic area and therefore the probability of pollen deposition, similar to stigmas of wind pollinated species (Faegri & Pijl 1979).Since the studied Begonia species are not pollinated by wind the increase of the stigmatic area may reflect an adaptation regarding the unpredictability of visitors to unrewarding flowers.The stigmatic papillae also increase the receptive area and facilitate adherence of pollen grains through direct contact with the pollinators or during vibration when pollen grains are liberated from the pollinator's body.
The position of rimose anthers, grouped in large numbers, facilitate the collection of pollen by vibration.The anthers of B. integerrima and other species of section Solananthera (Buchmann 1983;Silva & Mamede 2001;Jacques 2002) are poricidal.We suggest that this type of anther may help to exclude pollen-robbing insects, which commonly remove large numbers of grains from open, longitudinally dehiscing anthers.Poricidal anthers also ensure that pollen is released more slowly over the life of the flower.Finally, pollen is effectively dispersed from pores only when insects buzz at a special frequency that causes anther contents to be released, often as a clump.High pollen viability is expected in all species of Begonia, although we were only able to confirm it for Begonia cucullata and B. aff.fluminensis.Even though production of large numbers of viable pollen grains represents a large energetic cost, it may be especially important in Begonia and similar plants in which visits to pistillate flowers depend on deceit.

Breeding Systems
The large numbers of pollen tubes observed in flowers collected from natural populations suggest that pistillate flowers receive many visits and/or that the visitors are efficient pollen vectors.Because visits to pistillate flowers generally are less frequent than to staminate flowers, it is likely that in one visit to a pistillate flower, a large load of pollen will be transferred.Visits to pistillate flowers where vibration is performed by the visitor probably result in more pollen being transferred than in visits where little contact is made with stigmas, especially taking into account that Begonia pollen is small and powdery.A single brief visit may be sufficient to guarantee deposition of large pollen loads.The abundance of pollen tubes in stigmas, styles, and ovaries indicate that pollinators are effective, which is also confirmed by high fruitset observed in natural populations.Ågren and Schemske (1991) also observed high fruit-set in natural populations of B. involucrata in Costa Rica.
The absence of apomixis and geitonogamy (in species with inflorescences with flowers of both sexes open synchronously) suggests an absolute dependence on floral visitors for reproduction by these species of Begonia.Earlier researchers have considered the entire genus to be self-compatible (East 1940;Burt-Utley 1985).The only experimental studies of breeding systems in Begonia showed five species from section Doratometra in Central America to be self-compatible (Ågren and Schemske 1993).Moreover, Ågren and Schemske (1993) found extremely low (<5%) rates of outcrossing in natural populations of B. hirsuta and B. semiovata.In contrast, our results indicate that at least two species from Brazil are self- incompatible.More crossing studies of species of Begonia from different sections of the genus are needed.

Hymenoptera
The bees that we observed visiting the eight species of Begonia studied are of suitable body size and exhibit behaviour fitting to effect pollination of pistillate flowers.Moreover, their behaviour, frequency, and duration of visits reveal these bees as the principal pollinators of all eight species of Begonia investigated.
Trigona spinipes, the most abundant visitor, is known for its opportunistic behaviour, frequently acting as a robber of floral rewards in many species of plants (Roubik 1989;Sazima & Sazima 1989).In our study these bees occasionally made visits of long duration to pistillate flowers, possibly enhancing pollen deposition and/or removal of pollen that had already been deposited on the stigmas.For this species of visitor only, duration of visits to pistillate flowers did not differ significantly from the duration of visits to staminate flowers.
The slow movement of T. spinipes between flowers probably is compensated by its high frequency of visits.
Paratetrapedia sp. is one of the most important pollinators of the species of Begonia studied, due to its frequency of visits and the diversity of species of Begonia visited.Due to the varied floral morphology and reproductive phenology of the species of Begonia visited by this bee, Paratetrapedia sp.probably visits many of the other 16 species of Begonia in the study area.
Species of Melipona are probably important pollen vectors for species of Begonia that occur high in the canopy, like B. integerrima, because these bees frequently forage at high levels of the forest (Nieh 1995;Ramalho 2004).Taking into account that the percentage of mistake visits was greater in Melipona quadrifasciata than Melipona bicolor, the former is probably a more efficient pollinator of flowers of B. integerrima, despite the fact that it makes fewer visits.Euglossa spp.are probably important pollinators of B. aff.fluminensis and B. integerrima, promoting cross-pollination as these bees forage over long distances (Janzen 1971).This fact may be very important for B.

Pollination by vibration
Collection of pollen by vibration is characteristic of all the species of Begonia in our study.Pistillate flowers also benefit from vibration, as this behaviour favors deposition of more pollen on stigmas than those deposited by contact with the body of a visitor alone.It is likely that, during vibration, many pollen grains adhered to the abdomen of the bee fall on to the stylar branches.This would explain the large pollen loads on stigmas and the high levels of fruit-set under natural conditions that we observed.This study confirms Roubik's (1989) inclusion of Begonia on a list of genera with species characterized by poricidal anthers (B.integerrima), as also reported by Silva and Mamede (2001).Trigona spinipes, the only bee incapable of vibration, was notably absent from B. integerrima flowers.We suggest that pollination by vibration predominates in those species of Begonia pollinated by bees that present staminate flowers with connivent rimose or poricidal anthers.Poricidal anthers do not appear to be widespread in the genus, as this was observed in only one of the 13 species studied.

Other visitors
Non-hymenopteran visitors are not considered principal pollinators of Begonia.The fact that these insects visit both staminate and pistillate flowers, however, suggests that they may occasionally effect pollination.Butterfly visits probably were made by immatures that had not yet learned that Begonia flowers lack nectar.

Diversity of pollinators
The diversity of pollinators observed and recorded in this study differs from other studies, which reported low diversity of visitors to flowers of Begonia (Ågren & Schemske 1991;Schemske & Ågren 1995;Corff et al. 1998).Ågren and Schemske (1991) reported that a single species of Trigona made 98% of all observed visits to B. involucrata.Similarly, Schemske et al. (1995) reported that one species of Bombus made 97% of all visits to B. oaxacana.Corff et al. (1998) observed that Trigona fulviventris made 95% of visits to B. tonduzii.The high diversity of visitors to the species of Begonia studied in Brazil is similar to that recorded on the flowers of B. urophylla, a member of section Gireoudia (Corff et al. 1998).That species was visited by seven species of bees from four genera (Augochloropsis, Melipona, Neocorynura, Trigona), all represented by visitors to the species of Begonia in this study.The high diversity of bee visitors to the eight species of Begonia studied appears to be related to their presentation of open flowers with readily accessible pollen.The diversity of visitors, including other functional types, such as Diptera and Coleoptera, may indicate a tendency to generalization of floral type in these species of Begonia.

Future directions
This study confirms for a number of previously unexamined species the role of intersexual mimicry in the deceit pollination system of Begonia.It also shows that there is wide variation in ratios of staminate to pistillate flowers and in the timing of flowering.It remains to be shown how phenology and flower ratios influence pollinator behaviour and the rate of deceit.In any case, the rate does appear to differ depending on the species of pollinator, yet overall levels of pollination appear to be adequate to maintain high levels of fruit-set in natural populations.More detailed studies of individual species of Begonia may reveal greater differences in the behaviour of J Poll Ecol 6(14) individual pollinators, including the possibility of learning on the part of individual pollinators.
This study also reports that two species of Begonia from Brazil are genetically self-incompatible.Previous work has found Central American species of Begonia section Doratometra to be self-compatible, with very high levels of selfing in natural populations.Apparently, breeding systems in Begonia are more diverse than previously thought, and it will prove interesting to map these transitions on a phylogenetic tree of the genus.Moreover, examination of other floral characteristics in a phylogenetic context should prove informative regarding the evolution of Begonia in relation to its pollinators.
).All species are monoecious and present staminate and pistillate flowers in different phases of anthesis, with the exception of B. cucullata, B. integerrima, and B. itatinensis, in which staminate and pistillate phases overlap.In nine of the 13 species studied, staminate flowers open before pistillate flowers.Typically, pistillate flowers open and become receptive only after staminate flowers have senesced.These species may produce more than one inflorescence per individual, but inflorescences always occur in different phases (staminate or pistillate).The species that present staminate and pistillate flowers opensynchronously on the same inflorescence are B. integerrima and B. cucullata.Begonia itatinensis also presents synchronously open staminate and pistillate flowers, but the inflorescences are reduced to a single flower (Fig. 3f).The number of flowers per inflorescence and flower size vary widely among the species studied.Begonia caraguatatubensis (Fig. 3a, b), B. hookeriana, and B. luxurians possess many small (2 -4 mm; Tab. 1) staminate and pistillate flowers.Begonia caraguatatubensis presents 24 ± 28 flowers (n = 10 inflorescences) open per inflorescence in the staminate phase and 57 ± 66 open flowers (n = 10 inflorescences) per inflorescence during the pistillate phase.These species show a temporal separation of open staminate and pistillate flowers, with anthesis of pistillate flowers following the senescence of staminate flowers.On the other hand, B. fernandocostae, B. aff.fluminensis, B. integerrima, and B. valdensium possess larger staminate and pistillate flowers (9 -15 mm; Tab. 1) and relatively fewer flowers per inflorescence than other species.

FIGURE 2 .
FIGURE 2. Flowering phenology of 13 species of Begonia in the Serra do Mar State Park, São Paulo.Thick lines represent flowering peaks and thin lines represent the entire period of observed flowering.

FIGURE 4 .
FIGURE 4. Rimose stamens of Begonia cucullata flower (a) and poricidal stamens of B. integerrima (b).Branches of the style and stigmatic surfaces of B. aff.fluminensis (c) and of B. integerrima (tepal removed) (d).Note the spiral form of the branches and the papillae on the stigmatic surfaces (arrow).

FIGURE 6 .
FIGURE 6. Pollination by vibration in Begonia integerrima.Melipona quadrifasciata in vibration posture on a staminate flower (a); M. bicolor adopting vibration posture on a staminate flower (b); Pistillate flower being visited by Augochloropsis sp. 2, in vibration posture (c); Augochloropsis sp. 2 in vibration posture on staminate flower (d).Staminate flowers open and become available to pollinators before pistillate flowers in most of the species studied.This likely promotes cross-pollination (Burt-Utley 1985; Ågren & Schemske 1993; Jacques 2002).It may also promote visits to pistillate-phase inflorescences because visitors become habituated to visiting staminate flowers and when the pistillate flowers open, due to their similarity to staminate flowers, the visitors are easily deceived.It was hypothesized by Schemske and Ågren (1995) that species with staminate and pistillate flowers open at the same time would more easily deceive visitors than species with only pistillate flowers open.Our data indicate, however, that species of pollinator may play a larger role in determining visits to pistillate flowers.

FIGURE 7 .
FIGURE 7. Frequency histogram of visit duration of Trigona spinipes to flowers of Begonia aff.fluminensis (a).Frequency histogram of visit duration of Paratetrapedia sp. to flowers of B. caraguatatubensis (b).Number of visits by Paratetrapedia sp. to staminate and pistillate flowers of B. caraguatatubensis during two periods of observation (c).Frequency histogram of visit duration of species of Melipona on flowers of B. integerrima (d).

TABLE 1
. Floral characteristics of the species of Begonia studied in the Serra do Mar State Park, São Paulo (n = tepal number).Tepal number and length taken from Silva and Mamede (2001).

TABLE 2
bee species observed, visiting only 2 to 5 flowers per trip.In total during one five-hour period, 53 staminate flowers and six pistillate flowers of B. aff.fluminensis (Fig.7a) were visited.For T. spinipes there was no significant difference in duration other aff. fluminensis, which is endemic to the Ubatuba municipality and occurs in small, isolated populations.It is likely that the observed species of Euglossa visits other species of Begonia with flowers similar to B. aff.fluminensis and B. integerrima, such as B. fernandocostae, which is sympatric with B. aff.fluminensis, has a similar reproductive phenology and floral morphology, and nearly identical floral odour.Other examples would be B. solananthera and B. radicans, which are also vines with very similar floral morphology to B. integerrima.These species occur in the area of study but were not studied.Visits by Euglossa sp. were characterized by a high percentage of mistake visits to pistillate flowers.