Ecological and genetic consequences of pollination by sexual deception in the orchid Caladenia tentactulata.
Pollination by deceptive sexual attraction most often involves male aculeate Hymenoptera and is unique to the orchids (Nilsson 1992). Australia has the most highly developed sexually deceptive orchid flora with over 100 species, representing nine genera. Male thynnine wasps (Tiphiidae: Subfamily Thynninae) are the pollinators of some 70 species in the genera Arthrochilus, Caladenia, Chiloglottis, Drakaea, Spiculaea, and Paracaleana whereas other parasitic wasps, sawflies, and ants pollinate Cryptostylis and Calochilus, Paracaleana, and Leporella, respectively (Stoutamire 1983; Peakall et al. 1987; Peakall 1990).
In Europe and the Mediterranean region the genus Ophrys, with many more than the 30 or so described species, is pollinated by male bees and predatory and parasitic wasps (Kullenberg and Bergstrom 1976; Paulus and Gack 1990). Recently sexual attraction of predatory wasps, but not attempted copulation, has been reported for two species of South African Disa (Steiner et al. 1994). Male Diptera are also believed to be involved in deceit systems, probably pollinating Trichoceros in South American (van der Pijl and Dodson 1966), and Pterostylis in Australia (Jones 1988).
It is well established that pollinator behavior controls the pattern and extent of pollen dispersal and, in the absence other mechanisms, may regulate the mating system. The optimal foraging strategies of food seeking bees and bumblebees can lead to high levels of pollen transfer within plants (geitonogamy) and mean pollen dispersal distances are typically less than several meters (e.g., Handel 1983; Nicholls 1985; Richards 1986; Pleasants 1991). In contrast, butterfly behavior at flowers may reflect a combination of optimal foraging, mate search, and oviposition, and consequently both short and long distance pollen transfers may occur (Schmitt 1980). Murawski and Gilbert (1986) report a mean dispersal distance of 54 m (maximum 350 m) for the sparsely distributed butterfly pollinated shrub Psiguria warscwiczii, although dispersal distances appear to be more typical of bee pollination in other butterfly pollinated species (Webb and Bawa 1983; Broyles and Wyatt 1.991). Trapline foragers such as Euglossine bees may mediate very long pollen transfers (Janzen 1971), although pollen dispersal distributions have yet to be characterized.
Pollinator behavior at deceptive flowers remains poorly documented, but it is predicted that in the absence of food, flower visits will be brief and few. This may result in a limit to geitonogamous pollen transfers and long distance pollen flow (Dressler 1981). Many food deceptive species do show low levels of fruit set compared to food rewarding species (e.g., Gill 1989; Neiland and Wilcock 1995), however, pollen dispersal distances remain to be described. In sexually deceptive orchids, pollinator behavior and movements will be controlled by optimal mate seeking, rather than optimal foraging strategies. This may result in outcrossing and long distance pollen flow in some systems (Peakall 1990), but not others (Peakall and James 1989).
In this paper, we explore the ecological and genetic consequences of pollination by sexual deceit in the orchid Caladenia tentaculata. Specifically we ask the following questions: (1) Is pollinator behavior likely to result in outcrossing and long distance pollen flow? (2) Is outcrossing and long distance pollen flow observed? (3) Are there fitness advantages to outcrossing and longer distance pollen flow? (4) What are the implications of our findings for hypotheses on the evolution of sexual deception?
MATERIALS AND METHODS
Caladenia tentaculata sensu lato is part of a widespread species complex that grows in open forest and woodland in all states of Australia except the Northern Territory (Jones 1988; pers. comm. 1995). The flower bears one, (occasionally two), red and green to yellowish green flowers with perianth parts on a thin stem up to 0.5 m in height. Lateral sepals and petals are long (7 cm) and pendulous, giving rise to the common name "spider orchid." The dorsal sepal (6 cm) is held erect above the column. Tips of the sepals are glandular and yellow in color. The hinged labellum (1-1.5 cm across) is three lobed with a deep maroon color on the apex of the recurved central lobe. Clubbed maroon calli extend from near the apex to the base in four rows. The green side lobes are fringed with teeth that point upward and serve to align the pollinator with the column. Pollinia are enclosed within the anther as four lobes in two pairs without stipe or viscidia.
Caladenia tentaculata is pollinated by sexually attracted male thynnine wasps. Female thynnine wasps attract patrolling males with pheromones, perching in a premating posture on vegetation usually less than 0.5 m high (Ridsdill-Smith 1970a,b; Alcock 1981). The males grasp the females and transport them in copula to a nectar source where the females are fed by regurgitation. Females are then returned to the ground where they burrow in search of host Scarabaeid beetle larvae, which are paralyzed before oviposition. Wasps of both sexes are capable of mating several times, but females mate only once between each oviposition (Ridsdill-Smith 1970a).
The orchid exploits the reproductive behavior of the wasps both by the production of pheromone-like substances and by presenting labella structures that mimic the female shape and color. Males circle the flower and, at close range, locate the [TABULAR DATA FOR TABLE 1 OMITTED] flower in a zigzag upwind flight suggesting an odor gradient. Pollination occurs when a male lands on the labellum, aligns with the column, and then attempts to copulate with the apex of the central labellum lobe while grasping the maroon calli - the female mimic. Stigmatic secretions are smeared onto the insect when copulatory movements bring the thorax into contact with the stigma. Subsequently, one or both pairs of pollinia lobes are attached by the stigmatic glue as the insect leaves the flower.
A single wasp species, Thynnoides pugionatus, was the primary pollinator at our study sites, whereas a second less common species, Thynnoides rufithorax, was occasionally observed visiting and pollinating the flowers. No other insects capable of pollination were observed at the flowers. Dissection of the floral parts, followed by presentation to the wasps, revealed that both the glandular sepal tips and calli from the central labellum lobe are a source of attractant. Further, both calli and sepal tips from a single flower will attract wasps when presented on their own, although attempted copulation was never observed when sepals were presented without the calli (data not shown). This is consistent with previous studies that suggest olfactory cues are the primary source of attraction to the flower, but at close range visual and tactile mimicry are essential for successful pollination (Peakall 1989a, 1990).
Most of the study was centered on the three largest known populations of the orchid growing at the base of Black Spring Gully on the south side of the Weddin Mountains National Park (Table 1:W1-W3) approximately 300 km west of Sydney. The sites were located at the base of the hills within the transition zone between woodland, growing on the rocky hill slopes, and the Callitris forest on the plain. The orchid was typically found among low herbs and grasses in an open woodland of Eucalyptus, Acacia, and Callitris species. The main sites were the subject of study during the months of September and October each year from 1989 to 1992. Table 1 gives the general location for these and six other populations sampled for the genetic analysis.
Assessment of Pollination Success in Natural Populations
A lattice of permanent 20 x 20 m quadrats across each orchid population was established at W1 and W3. Each year all flowering plants were tagged and their reproductive status monitored throughout the flowering period. Given the size of the pollinia and stigma, pollen removals and deposition were easily monitored with the naked eye. These data enabled an evaluation of flower density, the total minimum number of wasp visits (as indicated by pollen removal or deposition or both), the time of pollinator visits ([+ or -] 2 d), ratio of pollen removal to pollination, total pollination, and predation of seed capsules by insects and mammalian herbivores.
Pollinator Behavior within Experimental Populations
The study of sexually deceptive pollinator behavior is difficult because of the brevity of each wasp visit and the low frequency of pollinators. However, when "bait" flowers are artificially presented within wasp populations, males can locate the flowers quickly, and efficient experimental work is possible (e.g., Peakall 1990; Handel and Peakall 1993; Peakall and Handel 1993). Using this approach with the sexually deceptive orchid Drakaea glyptodon, Peakall (1990) identified two features critical to experimental design. First, following an initial rapid response at a single location, the number of visits declines steeply and few are observed after 5 min. Relocation of the same flowers even a few meters renews wasp response. It is therefore necessary to move bait flowers regularly and to avoid trials too near natural populations of the orchid, otherwise few or no visits occur. Second, individual flowers can vary significantly in attractiveness and it is advantageous to arrange pretrial visits to determine those most attractive to the wasps. As in the preceding studies, to avoid removing whole plants from the wild, bait flowers of C. tentaculata were cut at ground level and maintained in water. The flowers remained attractive for up to 3 wk, but only fresh flowers were used in the experiments described below.
Previous pollinator behavior experiments have presented one to several flowers as a single "small patch" and focused on pollinator behavior at the scale of several centimeters. Typically, a given pollinator will visit only one of the flowers in these small patches before departing (Peakall 1990). To determine whether this behavior persists on a larger scale, a "large patch" experiment was implemented on a single afternoon when weather conditions were ideal and wasp responses were frequent. A total of eight trials were performed, successive trials being relocated at least 40 m from the location of the last trial. For convenience, and to maximize the number of wasps responding, the experiment was performed along a track between sites W1 and W3, where orchids were absent but wasps were common. At each trial, eight flowers in four pairs, 15 cm apart, were placed on the corners of a 1 x 1 m quadrat and wasp behavior monitored for 15 min by two observers. At the same time, the experiment was recorded on video, enabling the fine details to be subsequently checked. To avoid any bias from unequally attractive pairs, at each new trial, the flowers were rotated so that different flowers were paired together.
Pollinator Movement and Pollen Flow
To determine the home range of the wasp pollinators, a mark-and-recapture study was established at six baiting stations 20 m apart along a linear transect near W1, where the wasps were abundant. On two successive days, three to four flowers were exposed for 5 min at each station. Wasps visiting the flowers were captured with a net and tagged with colored liquid paper on their thorax, as in Peakall (1990), using different color codes for each station. During the next week, baiting at all stations was undertaken again and the details of marked recaptures noted. The order of baiting at the different stations was randomized each day with the constraint that successive baiting experiments were at least 40 m apart to minimize any bias. An alternative estimate of pollinator movements was made by capturing wasps within the orchid population and attaching pollinia with clear silicon to the thorax of the wasp in much the same way as stigmatic secretions glue the pollinia to the insect. The pollinia were labelled one of three colors with micronized fluorescent powders mixed with water.
To evaluate the extent of pollen flow, pollinia were labeled with different colored histochemical stains following the technique of Peakall (1989b). Pollinia were color coded by carefully injecting 3-5 [[micro]liter] of stain inside the anther flap using a capilettor (Boehringer Mannheim, GmbH) loaded with a capillary tube with a very fine tip. The following stains were used in this study: Brilliant Green, Methylene Blue, Orange G, Rhodamine, and Trypan Red (see Peakall 1989b for details of relevant concentrations). Preliminary investigations showed that this technique did not interfere with subsequent pollinia removal and deposition. Patches of five to 10 flowers in close proximity (usually less than 2 m apart) were labeled the one color. Subsequent regular monitoring of the population enabled pollen flow to be tracked. For about 60% of all transfers, it was possible to assign paternity unambiguously and therefore measure exact pollen flow distances. When this was not possible, because pollen removal had occurred from two or more labeled flowers during the intervening period, a minimum and maximum measure of the distance between donor and recipient was made. Careful examination of recipient stigmas with a hand lens enabled determination of the minimum number of paternal types contributing to pollination.
Allozyme analysis was performed to evaluate the patterns of genetic variation within and among populations to determine whether the genetic data were consistent with expectations from the behavioral data. Floral parts proved to be the only reliable source of allozymes in this species, as the leaf had often begun to wither by flowering. Sample storage, tissue homogenation, and Titan III cellulose acetate electrophoresis (Helena Laboratories, Beaumont, Texas) was performed according to protocols published previously (Peakall and Beattie 1991, 1995). A total of 16 enzymes were resolved yielding 22 putative genetic loci. Of these loci, 12 were polymorphic in one or more populations with the remainder being monomorphic. The loci Got-3, Mdh-1, Mr, and Pgm were variable in all populations while Gdh, G3pd, Got-1, Got-2, Mdh-2, Pgi-2, Pk-1, and Tpi-1 were only variable in a subset of the populations and typically consisted of a predominant allele and one or more alleles at very low frequency. Full details of the enzyme names, codes, and allele frequencies observed are summarized in the Appendix.
For the loci Got-3, Mdh-1, Pgm, and Mr the genetic interpretation of the allozyme banding patterns was confirmed for the common alleles by evaluating progeny arrays of asymbiotically germinated seedlings from selfs and crosses of known genotype (see below). The genetic interpretation for the remaining less variable loci was inferred from the banding patterns and knowledge of the typical subunit structure of the enzyme.
Sampling for Allozyme Analysis
To facilitate the analysis of genetic variation within and among populations, plants were sampled randomly, but with an attempt to cover the entire contiguous population. Thus the sampling area varied from approximately 20 x 20 m in the smaller populations to 100 x 100 m in the largest populations with sample sizes per population ranging from 18 to 50 (see Appendix).
For the analysis of local spatial genetic structure a 20 x 40 m quadrat was established at the W3 site. All the flowering plants within the quadrat were mapped and subsequently sampled. To enable pollination and subsequent seed set to proceed, only the petals were sampled, as it is known these are not a source of the attractant and that pollination can occur in their absence (pers. obs.). Alternatively, perianth parts were removed from pollinated flowers as soon as they closed around the column. This did not interfere with subsequent seed set.
Statistical Analysis of Allozyme Data
For the population analysis, the percentage of polymorphic loci (P), gene diversity ([H.sub.e]), gene correlation statistics, the fixation index F, chi-squared tests for Hardy-Weinberg equilibrium, and differences in allele frequency were calculated using the computer program BIOSYS-1 (Swofford and Selander 1981). Since it is costly and time consuming to germinate the minute seeds of C. tentaculata, which normally require mycorrhizal infection in the wild, the genetic analysis of progeny arrays was restricted to hand pollinated fruits. Thus, it was not feasible to estimate outcrossing rates by the standard genetic analysis of progeny arrays (e.g., Clegg 1980). However, relative outcrossing estimates under the assumption of no selection between fertilization and the adult stage were calculated as t = (1 - F)/(1 + F) (Jain 1979).
The analysis of local spatial structure was performed using a new approach recently introduced by Nason et al. (in prep.). With this method the analysis of spatial structure is based on the estimation of Wright's relatedness coefficient p. This coefficient is linearly related to Moran's I - a coefficient commonly used in spatial autocorrelation analysis. The following is based on this manuscript, but a brief explanation is also provided in Loiselle et al. (1995). Spatial relationships between the frequencies of homologous alleles, [p.sub.i] and [p.sub.j]k, in pairs of mapped individuals, i and j, [[Rho].sub.ij] can be estimated as:
[Mathematical Expression Omitted], (1)
where p is the frequency of the allele with mean [Mathematical Expression Omitted], n is the number of individuals in the sample, k = n(n - 1)/2 is the total number of possible pairwise connections between n individuals. A multilocus measure of spatial genetic structure is obtained by weighting the result for each locus by its polymorphic index, [summation of][p.sub.i](1 - [p.sub.i]). Values of [r.sub.ij] range from -2 to + 2 with values of zero indicating there is no spatial structure. When there is inbreeding, [r.sub.ij] will overestimate [[Rho].sub.ij], but it nevertheless remains a useful statistic for describing spatial variation in population genetic structure (Nason et al. in prep.). Significant spatial genetic structure is detected when the observed value for [r.sub.ij] is outside the 95% confidence interval for the mean value generated by the random permutations of the data.
In this study, estimates of [r.sub.ij] within the quadrat were obtained for the loci polymorphic at the 95% level using a series of programs provided by J. Nason (Department of Botany, University of Georgia, Athens). In practice, the conclusions drawn from spatial autocorrelation and [r.sub.ij] analyses appear to be similar (Peakall and Beattie 1995). However, the existing commercially available spatial autocorrelation program SAAP was unable to accommodate the large data set in this study. Furthermore, Nason's programs enable tests of significance by random permutation, which is not provided by SAAP, although it is theoretically possible to test spatial autocorrelation statistics in the same way.
Tests of Self-Compatibility
To examine whether the species is self-compatible, artificial self- and cross-pollination treatments were performed on plants collected in the field but maintained in pots in an insect-free greenhouse. To ensure maximum seed set, each pollination used a complete pollinium (all four lobes). Just prior to dehiscence the fruits were collected and the seed stored dry until required. Embryo length and width were measured for 60 seeds per fruit whereas the percentage of seeds with embryos was determined by examining 200 seeds per fruit for four selfs and four crosses.
To determine whether self- and cross-pollinated seed germinate and develop equally, seed was germinated by asymbiotic sterile seed culture at the Kings Park Botanic Gardens (Perth, Western Australia) according to the methods of Dixon (1989). Caladenia orchid seeds develop into the photosynthetic seedling stage via a nonphotosynthetic protocorm phase. Several months after germination it was apparent that developmental phases were asynchronous among seed from different fruits. To determine whether this was related to the pollination treatment, the proportion of seedlings to protocorms was determined for four self- and four cross-progeny. This was repeated a second time 3 mo later. For each survey, two out of several plates per plant were randomly chosen and the proportion of each phase determined for a 2 cm wide strip across the diameter of the plate. For both this experiment and all seed measurements described above, the experiment was performed blind such that the observer had no knowledge of the treatment.
TABLE 2. Patterns of pollinator visitation, pollination, pollinia removal, and predation of flowers and fruits across sites and years. NA = not assessed. W1 W1 W2 W3 W3 Site/Year 1991 1992 1992 1991 1992 Pollinations 82% 12% 33% 29% 26% Visits 86% 28% 49% 38% 49% Pollinia removals 66% 22% 42% 33% 40% Pollination and pollinia removals 63% 7% 27% 24% 18% Pollination, no pollinia removals 19% 5% 6% 5% 9% Pollinia removal, no pol- lination 3% 15% 15% 9% 23% Predated flowers and fruits NA 16% 10% NA 20% Total no. flowers sur- veyed 131 207 78 541 257
Pollinator Behavior in Natural Populations
Despite working more than 300 h within the orchid populations, only 17 wasp visits to the flowers in situ were observed. Of these, two visits resulted in attempted copulation. In each case, pollination and pollen removal were observed. The remaining visits were merely brief pauses on the flower. Five of the wasps carried pollinia, showing that they had attempted copulation with other flowers.
Pollinator Behavior in Experimental Populations
In contrast to the paucity of observations within natural populations, 287 male wasps were attracted to artificially presented flowers over the 120 min of the large patch experiment. Sixty of 287 wasps approached within centimeters of the flower but did not land, the remaining 227 wasps made a total of 280 floral contacts. Most of the contacts (92.5%) represented a brief pause on the flower while only 7.5% resulted in attempted copulation - the behavior critical for pollination.
Fifty-two of 280 (19%) floral contacts were followed by movement from one flower to another within the patch. Forty-four (84%) of these intrapatch movements merely involved a wasp landing on one flower then briefly moving to another before leaving the patch. Four (8%) movements within the patch involved landing on one flower then moving to another where attempted copulation occurred, whereas an additional [TABULAR DATA FOR TABLE 3 OMITTED] four (8%) movements consisted of contact with a second flower following pseudocopulation with the first. No intrapatch movements involving successive pseudocopulations at two flowers within the patch were observed, nor were multiple movements between flowers observed. The proportion of intrapatch moves within pairs versus among pairs of flowers at each corner of the patch were approximately equal (28 versus 24, respectively). Wasps responded very rapidly when flowers were presented at the trials, but declined quickly with time. Over all trials, 60 wasps (20%) approached in the first minute, 162 (56%) within the first 5 min, whereas the remaining visits were spread more or less evenly over the remaining 10 min.
The number of wasps responding at each of the eight trials varied significantly ranging from a low of 15 to a high of 71 visits in the 15 min trials conducted 40 m apart on a 320 m transect ([[Chi].sup.2] = 72.5, df = 7, P [less than] 0.0001). Thus the wasp distribution was heterogeneous. This was also apparent across a 120 m long transect used in the mark-and-recapture study with total responses at the 6 baiting stations varying from six to 34 ([[Chi].sup.2] = 28.9, df = 5, P [less than] 0.0001).
Flower Visitation and Pollination Success in Natural Populations
Patterns of pollinator visitation, pollination, pollinia removal, and predation are summarized in Table 2. The percentage of pollinated flowers varied from 12% to 82%, with both these values observed at W1 in successive years. Subtraction of the number of pollinia removals from the number of pollinations indicated minimum pollinium losses ranged from 4% to 16%. Based on surveys at the end of the flowering season, predation of the flowers and capsules varied from 10% to 20% of the total flowers, and was largely attributed to grazing by kangaroos and the introduced rabbit.
Relationships between Visitation and Flower Density
Mean flower density within the populations ranged from 1.1 [+ or -] 0.6 flowers/[m.sup.2] (N quadrats = 41) at W2 in 1992 to 2.1 [+ or -] 1.6 flowers/[m.sup.2] (N quadrats = 254) at W3 in 1991. For 10 m quadrats, mean density ranged from 3.69 [+ or -] 1.9 flowers/10 [m.sup.2] to 16.9 [+ or -] 17.8 flowers/10 [m.sup.2] at the same sites. Table 3 summarizes the pattern of pollination visitation for the 1 [m.sup.2] quadrats with different flower densities. Any quadrat with one or more flowers showing pollinium removal and/or pollination was counted as a "visited" quadrat. Contingency chi-squared analysis revealed that quadrat visitation was independent of density for three of the four 1 [m.sup.2] analyses. The notable exception was the W1 site in 1992 with visitation to quadrats in the highest density class greater than expected. Only 5% (N quadrats = 170) of the 10 [m.sup.2] quadrats were unvisited, preventing a meaningful chi-squared analysis at this scale. However, 15 of the 18 quadrats containing a solitary flower (1 flower/10 [m.sup.2]) were visited, showing that visitation occurs well below mean flower densities.
Pollen Flow and Mark and Recapture
Mean pollen flow distances varied among sites (F = 2.28, P = 0.07, N = 66). The minimum mean distance of 6.5 [+ or -] 8.7 m (mean [+ or -] SD, N = 11) was observed at W3 in 1992, whereas the maximum mean distance of 26.4 [+ or -] 21.0 m (N = 13) was observed at W2 in the same year [ILLUSTRATION FOR FIGURE 1 OMITTED]. Over all populations and seasons, the mean pollen flow distance based on the minimum estimate was 16.6 [+ or -] 17.5 m (N = 66), which was not significantly different to the mean pollen flow distance of 17.7 [+ or -] 17.3 m (F = 0.11, P = 0.74, N = 66) based on the maximum estimate. The maximum pollen flow distance observed was 58 m.
Inspection of Figure 1 shows the pollen flow distributions to be more or less linear up to 20 m with subsequent sporadic points to the maximum. Vector-mediated self-pollinations have been observed to occur in sexually deceptive systems during very active attempted copulation in which pollinia removal and deposition occur in the same visit (pers. obs.). Based on the minimum estimate, seven pollen transfers may have been selfs, suggesting a maximum selfing rate of 10%. However, in five cases paternity was ambiguous (i.e., other potential donors existed), therefore self-pollination may have been as low as 3%.
Distributions of nearest neighbor distances and all possible pairwise distances in the population were calculated for the W1 population in 1992, but the number of flowers was similar in 1991. The strongly leptokurtic nearest neighbor distance distribution represents the expected pollen flow distribution if pollen flow was strictly among neighbors, whereas the multimodal distribution of all pairwise distances represents the expected shape of the pollen flow distribution if pollen flow was random. It is apparent that the mean pollen flow distance (15.2 [+ or -] 17.4 m, mean [+ or -] SD) was far greater than the mean nearest neighbor distance (1.1 [+ or -] 1.9 m, mean [+ or -] SD), with the majority of pollen transfers well beyond nearest neighbors. However, the mean pollen flow distance was shorter than the mean of all possible pairwise distances (36.6 [+ or -] 23.4, mean [+ or -] SD) and the maximum pollen flow distance of 58 m was substantially shorter than the maximum possible distance of 104 m [ILLUSTRATION FOR FIGURE 2 OMITTED].
Some 20 wasps had fluorescent colored pollinia successfully attached to their thorax. Subsequently, six flowers were found to be pollinated with pollen from these wasps. Six wasps were also captured with colored pollen labeled for the pollen flow experiments. The mean distances between pollen marking and either pollen deposition or recapture of the pollinator was 28.7 [+ or -] 15.5 m (mean [+ or -] SD, N = 12) with a range of 12.5 m to 57 m.
In the mark-and-recapture study, a total of 19 wasps were tagged and subsequently 14 recaptures were made of 123 wasp visits to the baiting stations over a seven-day period. Four recaptures were made at the marking site (zero distance), eight were at 20 m, and two were at 40 m from the marking site (mean = 17 m). Despite the potential for recaptures up to 120 m, none were made beyond 40 m.
Careful examination of the labeled pollen deposited on recipient stigmas with a hand lens, enabled estimation of the amount of pollen deposited (in pollinia lobe units) and minimum number of paternal types. The amount of labeled pollen deposited per stigma varied from a mere trace, representing less than one of the four pollinium lobes (N = 12) to essentially the entire pollinium (N = 3). Most commonly, an even spread of labeled pollen was apparent on the stigma representing approximately two pollinium lobes (N = 23). The number of stigmas identified as receiving a single labeled pollinium, or part thereof, was similar to the number of stigmas that were pollinated with both labeled and unlabeled pollen (19 versus 18, respectively). This suggests approximately [TABULAR DATA FOR TABLE 4 OMITTED] 50% of pollinations involved a single father. Multiple paternity arises in two ways: visitation by a single pollinator carrying multiple pollinia or multiple visitation by different pollinators. Some of the captured pollinators were observed to be carrying up to five pollinia. Multiple visits to flowers were also detected. In particular, flowers that had received only trace amounts of pollen remained open and were again pollinated several days later. On the other hand, flowers pollinated with two or more pollinia lobes usually closed overnight.
Allozyme Variation within and among Populations
Sample sizes, mean number of alleles per locus, percent polymorphism, heterozygosity, the fixation index F, and relative outcrossing estimates are shown in Table 4. Overall, the values among populations were similar, with percent polymorphism varying the most from 18.2% to 36.4%. Chi-square analysis of Hardy-Weinberg equilibrium showed that genotype frequencies for most loci in each population were consistent with the expectations for random mating. However, for all populations except W5, the Mr locus showed significant departures from expectation even when the rare alleles were pooled. This was always a consequence of an excess of homozygotes at the most common alleles. Despite apparent conformity to Hardy-Weinberg equilibrium for most loci, some variation in the fixation index F was apparent, with the mean population values ranging from 0.28 to 0.03 giving an overall mean of 0.14. This translates to a mean outcrossing estimate t of 0.839 (Table 4).
Table 5 summarizes the results of gene correlation analysis. While there was some variability among loci, with Tpi-2 and Mdh-2 showing the most allelic differences among the populations, very low levels of genetic differentiation were indicated by the low mean value of [Theta].
TABLE 5. Gene correlation statistics treating each of the sites as subpopulations. The means and standard deviation were calculated by jackknifing across loci according to the methods of Weir and Cockerham (1984). Gene correlation parameters are related to F-statistics as f = [F.sub.IS], [Theta] = [F.sub.ST], and F x [F.sub.IT]. Locus f [Theta] F G3pd -0.007 0.006 -0.001 Gdh 0.286 -0.007 0.281 Got-1 -0.011 0.007 -0.004 Got-2 -0.038 0.018 -0.019 Got-3 0.222 0.008 0.228 Mdh-1 0.111 0.038 0.145 Mdh-2 0.378 0.104 0.443 Mr 0.522 0.033 0.538 Pgi-2 0.133 0.025 0.154 Pgm 0.078 0.020 0.096 Pk-2 -0.006 0.043 0.037 Tpi-2 0.215 0.175 0.352 Mean 0.275 0.034 0.300 SD 0.136 0.008 0.131
Local Spatial Structure
The changes in the mean relatedness coefficient (rij) per locus in the 20 x 40 m quadrat at W3 and 95% confidence intervals about the mean for the null hypothesis of no spatial structure are shown in Figure 3. All loci showed positive and mostly significant values of [r.sub.ij] in the shortest distance classes, declining through zero between 2 m and 7 m. Subsequent oscillation of the curves between positive and negative values is also apparent and consistent with a pattern of local spatial structure. The points at which graphs first cross the x-axis provide estimates of patch size (Sokal and Wartenberg 1983) and indicate patch sizes between 2 m and 7 m. The most notable result was observed for Mr, which gave the largest patch size estimate and showed very clear oscillation indicative of a strong patch structure. It is of interest that this locus also showed a consistent excess of homozygotes relative to random expectations that may well reflect this pattern.
Figure 4A compares mean embryo width and length for artificial self- and cross-pollinated seed. Although the variance was greater for selfed seeds, the means were not statistically different. Similarly, the mean percent of seeds with embryos was similar between selfed and crossed fruits, although the variance was again greatest for the selfs (self: 74.0 [+ or -] 12.9, mean [+ or -] SD, N fruits = 4; cross: 73.7 [+ or -] 27.3, mean [+ or -] SD, N fruits = 4; F = 0.0003, P = 0.98).
In contrast, the proportion of seedlings to protocorms was significantly different among treatments [ILLUSTRATION FOR FIGURE 4B OMITTED]. At both sampling times, the proportion of seedlings to protocorms was greater for crossed seed. Further, even at the second sampling time (3 mo after the first), the mean proportion of seedlings was similar to that for the crosses at the first sampling. Thus, the development of self-progeny was approximately 3 mo behind the crossed progeny.
Does Pollinator Behavior Result in Outcrossing and Long Distance Pollen Flow?
The wasp pollinator behavior observed in this study was similar to that observed at other orchids including Drakaea glypotodon (Peakall 1990), Chiloglottis reflexa (Handel and Peakall 1993), and C. trilabra (Peakall and Handel 1993). The rapid peak response to artificially presented flowers, followed by a decline with time occurred in all cases. Mark-and-recapture data for D. glypotodon and the present study considered together showed that the wasps rarely revisit orchid flowers within trials. Late responses probably represent new individuals that have just moved within range of the stimulus rather than revisits. The decrease in wasp visits over time occurred despite the continuous orchid stimulus demonstrated by the ability of the same flowers to attract wasps when moved to new locations. The decline in wasp activity indicates a site-specific refractory response (Peakall 1990). Furthermore, the present study showed that within larger artificial patches that modeled realistic flower densities, most wasps visit only one orchid before leaving the patch. The few within-patch moves observed would not have led to pollen transfers within the patch.
The low frequency of attempted copulations found in this study has also been observed in D. glypotodon and C. trilabra (Peakall 1990; Peakall and Handel 1993), but not in C. reflexa, where 80% of all wasp visits resulted in attempted copulation (Handel and Peakall 1993). At long range, the primary pheromone mimicry appears to be very good but, at short range, the visual and tactile mimicry maybe less effective. The low frequency of pseudocopulation also implies that pollination proceeds only when males have a low stimulus threshold, perhaps reflecting virginity or a long period since a previous mating (Peakall 1990).
Overall, the data suggest that pollination in this system typically proceeds in the following ways. (1) During patrols for females, males encounter the orchid stimulus. (2) Driven by strong competition, most males that encounter the stimulus will approach the source. (3) If the stimulus quality is high, and the stimulus threshold of an attracted wasp is low, pollinium removal will proceed (although more often the male will leave the flower after a short pause). (4) Following either a short pause on a flower or successful pollen removal, the male leaves the orchid and the patch and will not respond to the same flower or patch for some time. (5) Subsequently, if a suitable flower is encountered, steps 2 through 5 are repeated, resulting in pollination. This behavior is expected to produce long distance pollen flow with near neighbor pollination occurring only by chance and not systematically as is typical for insects foraging for floral food rewards.
Strong competition among males may be a key feature of this pollination system. However, in contrast to the pollinators of Ophrys (Nilsson 1992), male preemergence is not required to generate it because the female spends most of her time underground. Thus, although the basic sex ratio is 1:1 (Ridsdill-Smith 1970a), the actual ratio is strongly skewed toward males, promoting strong competition. This may be the major reason for the rapid response of males to any stimulus that resembles a female. Given that most wasps pause only momentarily at the flower and then subsequently avoid them, suggesting there will be little or no fitness disadvantage to the wasps. Thus, as in most plant-animal interactions, including those typically thought to be coevolved (Beattie 1985, p. 144), the evolution of this interaction is strongly asymmetric and perhaps exclusive to the plant side.
Is Outcrossing and Long Distance Pollen Flow Observed?
The pollen labeling experiments confirmed that outcrossing predominates in C. tentaculata with pollen transfers between plants accounting for a minimum of 87% of all pollinations. Furthermore, the comparison of nearest neighbor with pollen flow distributions support the prediction that longer distance pollen flow does occur in this system. Pollen flow extended well beyond nearest neighbors with less than 10% of pollen transfers being shorter than the mean nearest neighbor distance [ILLUSTRATION FOR FIGURE 2 OMITTED]. The observed maximum pollen flow distance of 58 m was, however, substantially less than the maximum pollen transfer distance possible within the populations [ILLUSTRATION FOR FIGURE 2 OMITTED]. This finding, combined with the outcome of the mark-and-recapture experiments, suggest that wasp movement was confined to a maximum mate search distance of between 40 m and 60 m. This is smaller than that of the pollinator of D. glyptodon which had a maximum mate search distance of 130 m as determined by mark and recapture experiments (Peakall 1990).
Short mean pollen flow distances and strongly leptokurtic distributions, reflecting movements among near neighbors, are typical of food rewarding species. However, the frequency distributions of pollen flow distances may have very long tails in some species. For example, using paternity analysis, Ellstrand and Marshall (1985) found that interpopulation pollen flow in Raphanus sativus can occasionally occur up to 1 km. In fact, few studies have attempted to measure interpopulation pollen flow so maximum distances may be routinely underestimated. Mean pollen flow distances in C. tentaculata are among the largest reported in the literature. However, while sexual deception may produce nonneighbor pollination, upper limits to pollen flow may be imposed because the pollinator has some maximum mate search distance.
Are the Genetic Data Consistent with the Behavioral Predictions?
The results of the population genetic analysis were consistent with the behavioral evidence that C. tentaculata is a predominantly outbreeding species. Estimates for the fixation index F were close to the average of 0.16 observed for other outcrossing species (Brown 1979). The mean number of alleles per locus was above average, while mean %P and mean He were a little below average for animal pollinated outcrossing species, but close to typical values for mixed-mating plants (Hamrick and Godt 1989). On the other hand, the mean value of 0.034 for [Theta] (equivalent to [G.sub.ST]) were well below average [G.sub.ST] values in other plants including the values of 0.099 for wind pollinated plants and 0.143 for wind dispersed plants (Hamrick and Godt 1989) indicating that genetic differentiation among populations is unusually low for this species.
An unexpected outcome of the genetic analysis of C. tentaculata was the finding of strong local spatial structure. Although theoretical models predict that plant populations will exhibit local genetic structure when gene flow is restricted(e.g., Wright 1978; Sokal and Jacquez 1991; Ohsawa et al. 1993), many empirical studies of plants have found little or only weak local spatial genetic structure (Heywood 1991; Peakall and Beattie 1995). It is generally assumed in these cases that spatial structure is weak or absent because gene flow has been sufficient to minimize its development.
In the case of C. tentaculata, it is clear that pollen flow is extensive, therefore, in the absence of microgeographic selection, local spatial genetic structure observed in C. tentaculata most probably reflects restricted dispersal of seed. Given that seeds require mycorrhizal infection for germination, seed falling close to the parent may have a greater chance of contacting the appropriate fungus that frequently remains associated with the adult plant (Warcup 1981). At a larger scale, it is apparent that the extent of seed dispersal between populations was sufficient to minimize genetic differentiation among populations.
Are There Fitness Advantages of Outcrossing and Longer Distance Pollen Flow?
Although no visible differences were apparent among self-and cross-pollinated seed, there was convincing experimental evidence that outcrossed seed developed more quickly than selfed seed, at least on artificial asymbiotic media. It remains to be determined if outcrossed seedlings develop faster than selfed seedlings in the wild but, in general, studies of inbreeding depression find much stronger effects under natural conditions (Charlesworth and Charlesworth 1987). It is also not yet known if delayed growth has fitness consequences. However, it is feasible that a shorter development time would enhance fitness. For example, seedlings that become established earlier in the season may have a greater chance of surviving the hot and dry summer.
In addition to inbreeding by selfing, biparental inbreeding among related individuals is possible in C. tentaculata given significant local spatial structure. Examination of the quantiles for the total pollen flow data, with zero values (self-pollination) excluded, indicates that approximately 75% of the labeled pollen transfers were greater than 7 m, which was the maximum patch size detected. Thus, longer distance pollen dispersal may well promote outbreeding by maximizing pollen transfers among unrelated individuals. In contrast, if pollination was largely among near neighbors, the extent of biparental inbreeding would be substantially greater. Although self-progeny showed delayed development, it has yet to be determined whether partially inbred progeny show similar effects. Nevertheless, it follows that if outbred progeny exhibit a fitness advantage over both partially inbred and selfed progeny, such as more rapid germination, development, and establishment, then longer distance pollen dispersal will be advantageous.
If some selfing and biparental inbreeding occur in C. tentaculata, a finding of little or no evidence for inbreeding in the adult populations would support the hypothesis that out-bred progeny are at a selective advantage in natural populations. However, the genetic data were equivocal. While they indicated that outbreeding predominates, small but positive fixation indexes (both F and [F.sub.IS]) were found in some populations. Indeed, assuming no selection, the mean fixation index across all loci and populations translated to a mean relative outcrossing estimate of 84%, which is close to the maximum selfing rate detected by the pollen flow experiment. Thus, it may appear there is little support for selection against inbred progeny. On the other hand, genetic sampling across subpopulations may also give rise to a positive fixation index (the so-called Wahlund effect).
Formal distinctions between inbreeding and the Wahlund effect are complex, particularly for multiallelic loci. However, if inbreeding were the basis for the positive F-values, then it would be expected to affect all loci equally. In contrast, if positive F-values were observed for some loci but not others, the Wahlund effect would be indicated (Hartl and Clark 1989). The latter was observed in this study. Of particular interest was the observation that the largest deviations from Hardy-Weinberg equilibrium were observed for the Mr locus, which also showed the most extensive spatial structure. Other loci showed little deviation or even excesses of heterozygotes. Collectively, the data suggest that apparent deviations from Hardy-Weinberg equilibrium may well reflect sampling across local subpopulations rather than inbreeding.
In other sexually deceptive orchids the benefits of long distance pollen flow may well be substantially greater than in C. tentaculata, especially if outcrossed progeny are favored. For example, longer distance pollen flow will minimize pollen transfers within clones of the many clonal species in Arthrochilus and Chiloglottis. Sexual deception in the multi flowered Arthrochilus, Spiculaea, and Calochilus may also minimize pollen transfers within plants (geitonogamy) if the pollinators leave the plant after visiting a single flower.
Are There Fitness Advantages of Pollination at Low Density?
Evaluation of the patterns of pollinator visitation within natural populations showed that pollination in C. tentaculata was functional even in low density populations (Table 3). This finding is consistent with the observation that pollinators were readily attracted to a single bait flower. Further, only the glandular sepal tips from a single flower are required to attract a pollinator (pers. obs.). Although at times, local abundances can be very high, low density populations are typical of many orchids, both terrestrial and epiphytic and this is certainly the case for many terrestrial Australian species. Pollination may be limited at low floral density in food reward and food deception systems because pollinator interest cannot be sustained. Thus, pollination by sexual deception may be advantageous under low density conditions because seed set can be maintained. Even so, seed set in these species may still be pollinator limited as for some years in C. tentaculata (Table 2).
Implications for the Evolution of Sexual Deception
Five hypotheses for the evolution of pollination by deceit have been proposed (Dressler 1981; Ackerman 1986a,b; Nilsson 1992), although few experimental studies have tested them. (1) The pollinia hypothesis suggests that a switch from reward to deceit was possible given that the single transfer of a pollinium may yield high levels of fruit set. (2) The pollinia loss hypothesis proposes that a reduction in the attrition of pollinia typically associated with inconstant generalist pollinators was reduced by the evolution of highly specialised deceit systems. (3) The resource limitation hypothesis argues that deceptive pollination is advantageous when sexual reproduction is limited by resources and, although annual seed set may be low, a few high quality offspring are produced over the lifetime of the plant. (4) The low density hypothesis proposes that deception evolved when plant densities were insufficient to sustain the services of food-seeking pollinators. (5) The outcrossing hypothesis is based on the assumption that deception promotes outcrossing and long-distance pollen flow. While specifically formulated for the evolution of food deception these hypotheses may also be applied to the evolution of sexual deception. However, there is the complication that sexual deception is generally assumed to have arisen via food deception (Ackerman 1986a; Nilsson 1992).
Our study of C. tentaculata provide data consistent with the outcrossing and low density hypotheses. However, our findings may not be uniquely associated with sexual deception. In Caladenia at least, which has both food deceptive and sexually deceptive species, pollination by food deception may also produce similar outcomes. A detailed comparative study within a phylogenetic framework will be required to rigorously test these hypotheses to minimize errors that may arise from the lack of statistical independence among related species and inappropriate comparisons and data pooling (Armbruster 1992). The evolution of pollination by oil collecting euglossi bees within Dalechampia (Armbruster 1992, 1993) and hummingbird pollination within Aphelandra (McDade 1992) has been investigated effectively by this method. These studies revealed unexpected multiple origins for some pollination syndromes leading to the refinement of hypotheses and the development of new testable predictions. At a higher taxonomic level, Chase and Hills (1992) combined phylogenetic and ecological data to examine the evolution of male euglossi bee pollination within two orchid subtribes, Catasetinae and Cyrtopodiinae.
An ecophylogenetic study within Caladenia maybe particularly useful for testing whether sexual deception has indeed evolved from food deception, since it is the only genus worldwide to include both pollination strategies. An analogous comparative ecological and genetic study of a food deceptive species that is closely related to C. tentaculata may also provide important evolutionary clues. However, a food deceptive species is likely to be more difficult to study than C. tentaculata because pollinators cannot be readily attracted to experimentally presented flowers. Natural pollination rates may also be very low, making it difficult to track pollen flow by color coding. Also, Australian terrestrial orchids may be less suitable for investigating the evolution of food deception because food reward comparisons are unavailable, that is, genera pollinated by deceit are exclusively so and there are no closely related food rewarding genera. Thus, it may be necessary to look at other orchid groups to examine the evolution of food deception in more detail.
The lack of a phylogenetic framework is but one problem associated with the existing hypotheses for the evolution of deception: only one hypothesis is specific to the orchids (or nearly so as the Asclepiadaceae also have pollinia). Furthermore, deceptive pollination is clearly only one evolutionary alternative available to solve the set of problems associated with the last three hypotheses. For example, problems of resource limitation can be overcome by various strategies, while outcrossing can be promoted by a vast array of mechanisms besides pollinator behavior. Similarly, within the orchids, hyperdispersion may have promoted the evolution of three alternative specialized pollination strategies: trap-line, euglossine, and deceptive pollination (Ackerman 1986b). Yet, trapline and euglossine pollination are common in many families, while deception is most common among the orchids. Consequently, it seems that there must be unique features of the orchids that predispose them to evolve deceptive pollination strategies. Insight into the evolution of sexual deception is further dependent on our understanding of the evolution of food deception. Given the importance of floral fragrances as the primary attractant in sexual deception, perhaps orchid-specific biochemical pathways exist. This and other orchid-specific hypotheses require much further investigation to more fully account for the evolution of pollination by both food and sexual deception.
We wish to thank C. Bower for critical information on the locations of the orchid and for identifying the wasp species, D. Jones for advice on the taxonomy of the orchid, K. Dixon for germinating the orchid seed, J. Nason for the rij programs, C. Angus and I. Oliver for much assistance in both the field and laboratory, and two anonymous reviewers for helpful comments on the manuscript. Funds for this research were provided by the American Orchid Society, the Australian Research Council (ARC) Large Grants Scheme and as a ARC Postdoctoral Research Fellowship to RP.
ACKERMAN, J. D. 1986a. Mechanisms and evolution of food deceptive pollination systems in orchids. Lindleyana 1:108-113.
-----. 1986b. Coping with the epiphytic existence: Pollination strategies, Selbyana 9:52-60.
ALCOCK, J. 1981. Notes on the reproductive behavior of some Australian thynnine wasps (Hymenoptera: Tiphiidae). J. Kansas Entomol. Soc. 54:681-693.
ARMBRUSTER, W. S. 1992. Phylogeny and the evolution of plant-animal interactions. Bioscience 42:12-20.
-----. 1993. Evolution of plant pollination systems: Hypotheses and tests with the neotropical vine Dalechampia. Evolution 47: 1480-1505.
BEATTIE, A. J. 1985. The evolutionary ecology of ant-plant mutualisms. Cambridge Univ. Press, Cambridge.
BROWN, A. H. D. 1979. Enzyme polymorphism in plant populations. Theor. Popul. Biol. 15:1-42.
BROYLES, S. B., AND R. WYATT. 1991. Effective pollen dispersal in a natural population of Asclepias exaltata: The influence of pollinator behavior, genetic similarity, and mating success. Am. Nat. 138:1239-1249.
CHARLESWORTH, D., AND B. CHARLESWORTH. 1987. Inbreeding depression and its evolutionary consequences. Annu. Rev. Ecol. Syst. 18:237-268.
CHASE, M. W., AND H. G. HILLS. 1992. Orchid phylogeny, flower sexuality and fragrance-seeking. Bioscience 42:43-49.
CLEGG, M. T. 1980. Measuring plant mating systems. Bioscience 30:814-818.
DAFNI, A. 1984. Mimicry and deception in pollination. Annu. Rev. Ecol. Syst. 15:259-278.
-----. 1986. Floral mimicry - Mutualism and unidirectional exploitation of insects by plants. Pp. 86-94 in T. R. F. Southwood and B. E. Juniper, eds. Plant surface and insects. Edward Arnold, London.
DIXON, K. W. 1989. Seed propagation of ground orchids. Pp. 1826 in K. W. Dixon, B. J. Burichell, and M. T. Collins, eds. Orchids of Western Australia: Cultivation and natural history. 2d ed. Western Australian Native Orchid and Conservation Group Inc., Perth.
DRESSLER, R. L. 1981. The orchids natural history and classification. Harvard Univ. Press, Cambridge, MA.
ELLSTRAND, N. C., AND D. L. MARSHALL. 1985. Interpopulation gene flow by pollen in wild radish, Raphanus sativus. Am. Nat. 126:606-616.
GILL, D. E. 1989. Fruiting failure, pollinator inefficiency, and speciation in orchids. Pp. 458-481 in D. Otte and J. A. Endler, eds. Speciation and its consequences. Sinauer, Sunderland, MA.
HAMRICK, J. L., AND M. J. GODT. 1989. Allozyme diversity in plant species. Pp. 43-63 in H. D. Brown, M. T. Clegg, A. L. Kahler, and B. S. Weir, eds. Plant population genetics, breeding and genetic resources. Sinauer, Sunderland, MA.
HANDEL, S. N. 1983. Pollination ecology, plant population structure, and gene flow. Pp. 163-211 in L. Real, eds. Pollination biology. Academic Press, New York.
HANDEL, S. N., AND R. PEAKALL. 1993. Thynnine wasps discriminate among heights when seeking mates: Tests with a sexually deceptive orchid. Oecologia 95:241-245.
HARTL, D. L., AND A. G. CLARK. 1989. Principles of population genetics. 2d ed. Sinauer, Sunderland, MA.
HEYWOOD, J. S. 1991. Spatial analysis of genetic variation in plant populations. Annu. Rev. Ecol. Syst. 22:335-355.
JAIN, S. K. 1979. Estimation of outcrossing rates: Some alternative procedures. Crop Sci. 19:23-24.
JANZEN, D. H. 1971. Euglossine bees as long-distance pollinators of tropical plants. Science 171:203-205.
JONES, D. L. 1988. Native orchids of Australia. Reed, Australia.
KULLENBERG, B., AND G. BERGSTROM. 1976. The pollination of Ophrys orchids. Bot. Notiser 129:11-19.
LOISELLE, B. A., V. L. SORK, J. NASON AND C. GRAHAM. 1995. Spatial genetic structure of a tropical understorey shrub, Psychotria officinalis (Rubiaceae). Am. J. Bot. 82:1390-1398.
MCDADE, L. A. 1992. Pollinator relationships, biogeography, and phylogenetics. Bioscience 42:21-26.
MURAWSKI, D. A., AND L. E. GILBERT. 1986. Pollen flow in Psiguria warscewiczii: A comparison of Heliconius butterflies and hummingbirds. Oecologia 68:161-168.
NEILAND, M. R. M., AND C. C. WILCOCK. 1995. Maximisation of reproductive success by European Orchidaceae under conditions of infrequent pollination. Protoplasma 187:39-48.
NICHOLLS, M. S. 1985. Pollen flow, population composition and the adaptive significance of distyly in Linum tenuifolium. Biol. J. Linn. Soc. 25:235-242.
NILSSON, L. A. 1992. Orchid pollination biology. TREE 7:255-259.
OHSAWA, R., N. FURUYA, AND U. YASUO. 1993. Effect of spatially restricted pollen flow on spatial genetic structure of an animal-pollinated allogamous plant population. Heredity 71:64-73.
PAULUS, H. F., AND C. GACK. 1990. Pollinators as prepollinating isolation factors: evolution and speciation in Ophrys (Orchidaceae). Israel. J. Bot. 39:43-79.
PEAKALL, R. 1989a. The unique pollination of Leporella fimbriata (Orchidaceae): Pollination by pseudocopulating male ants (Myrmecia urens, Formicidae). Plant Syst. Evol. 167:137-148.
-----. 1989b. A new technique for monitoring pollen flow in orchids. Oecologia 79:361-365.
-----. 1990. Responses of male Zaspilothynnus trilobatus Turner wasps to females and the sexually deceptive orchid it pollinates. Funct. Ecol. 4:159-167.
PEAKALL, R., AND A. J. BEATTIE. 1991. Genetic consequences of worker ant pollination in a self-compatible, clonal orchid. Evolution 45:1837-1848.
-----. 1995. Does ant dispersal of seeds in Sclerolaena diacantha (Chenopodiaceae) generate local spatial genetic structure? Heredity 75:351-361.
PEAKALL, R., AND S. N. HANDEL. 1993. Pollinators discriminate among floral heights of a sexually deceptive orchid: Implications for selection. Evolution 47:1681-1687.
PEAKALL, R., AND S. H. JAMES. 1989. Outcrossing in an ant pollinated clonal orchid. Heredity 62:161-167.
PEAKALL, R., A. J. BEATTIE, AND S. H. JAMES 1987. Pseudo-copulation of an orchid by male ants: A test of two hypotheses accounting for the rarity of ant pollination. Oecologia 73:522-524.
PLEASANTS, J. M. 1991. Evidence for short-distance dispersal of pollinia in Asclepias syriaca L. Funct. Ecol. 5:75-82.
RICHARDS, A. J. 1986. Plant breeding systems. Allen and Unwin, London.
RIDSDILL-SMITH, T. 1970a. The biology of Hemithynnus hyalinatus (Hymenoptera: Tiphiidae), a parasite of scarabaeid larvae. J. Aust. Entomol. Soc. 9:183-195.
-----. 1970b. The behaviour of Hemithynnus hyalinatus (Hymenoptera: Tiphiidae), with notes on some other Thynninae. J. Aust. Entomol. Soc. 9:196-208.
SCHMITT, J. 1980. Pollinator foraging behaviour and gene dispersal in Senecio (Compositae). Evolution 34:934-942.
SOKAL, R. R., AND G. M. JACQUEZ. 1991. Testing inferences about microevolutionary processes by means of spatial autocorrelation. Evolution 45:152-168.
SOKAL, R. R., AND D. E. WARTENBERG. 1983. A test of spatial autocorrelation analysis using an isolation-by-distance model. Genetics 105:219-237.
STEINER, K. E., V. B. WHITEHEAD, AND S. D. JOHNSON. 1994. Floral and pollinator divergence in two sexually deceptive South African orchids. Am. J. Bot. 81:185-194.
STOUTAMIRE, W. P. 1983. Wasp-pollinated species of Caladenia in South-western Australia. Aust. J. Bot. 31:383-394.
SWOFFORD, D. L., AND R. B. SELANDER. 1981. BIOSYS-1: A FORTRAN program for the comprehensive analysis of electrophoretic date in population genetics and systematics. J. Hered. 72: 281-283.
VAN DER PIJL, L., AND C. H. DODSON. 1966. Orchid flowers: Their pollination and evolution. Univ. of Miami Press, Coral Gables, FL.
WARCUP, J. H. 1981. The mycorrhizal relationships of Australian orchids. New Phytol. 87:371-381.
WEBB, C. J., AND K. S. BAWA. 1983. Pollen dispersal by hummingbirds and butterflies: A comparative study of two lowland tropical plants. Evolution 37:1258-1270.
WEIR, B. S., AND C. C. COCKERHAM. 1984. Estimating F-statistics for the analysis of population structure. Evolution 38:1358-1370.
WRIGHT, S. 1978. Evolution and the genetics of populations. Vol 4. Univ. of Chicago Press, Chicago.
[TABULAR DATA FOR APPENDIX OMITTED]
|Printer friendly Cite/link Email Feedback|
|Author:||Peakall, Rod; Beattie, Andrew J.|
|Date:||Dec 1, 1996|
|Previous Article:||Pollen-tube competition, siring success, and consistent asymmetric hybridization in Louisiana irises.|
|Next Article:||Evolutionary genetics of allorecognition in the colonial hydroid Hydractinia symbiolongicarpus.|