• Barrett, Spencer C. H.
  • Glover, Deborah
  • University of Toronto


Darwin proposed that the function of the stamen‐style polymorphism in heterostylous plants is to increase the probability of legitimate (compatible) pollinations among the floral morphs. Conspicuous pollen trimorphism in tristylous Pontederia cordata enables a test of the hypothesis. Comparison of the composition of pollen loads in naturally pollinated stigmas of intact and emasculated flowers were made at a population in Paugh Lake, Ontario, which was visited primarily by bumblebees. The magnitude of legitimate pollination was analyzed by ANOVA. In intact flowers, significant legitimate pollination was detected in the long‐styled morph only. Following emasculation legitimate pollination was evident in the long‐ and short‐styled morphs, with the mid‐styled morph just short of displaying significant legitimate pollination. Similar results were obtained by chi‐square analysis.

It has been suggested that heterostyly may reduce mutual interference between maternal and paternal reproductive function. Two aspects of pollen‐stigma interference were investigated in P. cordata. The potential importance of stigmatic or stylar clogging by incompatible pollen was examined by controlled field pollinations and measurements of seed set. The results indicate that prior application of large amounts of incompatible pollen has no significant effect on the seed set of open‐pollinated inflorescences. Comparison of legitimate pollen capture in intact and emasculated flowers provided no evidence that the presence of stamens within flowers of the floral morphs interferes with the receipt of legitimate pollen. Pollen‐stigma interference remains to be demonstrated in heterostylous plants.


The test of Darwin's hypothesis was conducted at a population of P. cordata bordering Paugh Lake, near Barry's Bay, Ontario, (N45°35', W77°45') between August 6-24, 1982. This corresponds to the middle of the seven-week flowering period, although peak flowering occurs about one week earlier. Further information on the floral biology, flowering phenology, and details often temporal variation in pollen flow at the site during the 1979 season, can be found in Price and Barrett (1982, 1984) and Barrett et al. (1983). At peak flowering, the stand consists of approximately 3,000 reproductive shoots and extends from the lake shore outwards along a gently sloping incline for 10-15 m, and laterally for 200 m. Water depth throughout the area where plants occur ranges from 0-1.5 m. During 1982, the floral morph frequencies in the population were L:0.324, M: 0.276, and S:0.400, N = 377.

Examination of Pollen Loads. Stigmatic pollen loads were collected from naturally pollinated stigmas of P. cordata following procedures detailed in Price and Barrett (1984). For the emasculated treatment, the anthers from all open flowers on two or three inflorescences per morph were removed prior to flower opening (0730-1100 hr, depending on night temperature). The degree of contamination resulting from this procedure was estimated by collecting and preserving stigmas from 40 flowers of each floral morph immediately after emasculation. Both emasculated and intact inflorescences were marked with tape, and, just before the flowers closed, 40 stigmas were removed and preserved in FAA separately by morph and treatment. This was repeated for nine days for a total of 360 flowers per treatment per morpho Different groups of 2-3 inflorescences per treatment were used on each day. Preserved stigmas were acetolyzed in groups of 40 corresponding to the daily treatments (for method see Ganders, 1974; Price and Barrett, 1984), and the resultant solution of pollen grains was suspended in lactophenol-glycerin with cotton blue, subsampled, and scored by size. Standards were obtained by acetolyzing undehisced anthers from each stamen level of five individuals per floral morpho The length of 100 pollen grains of each level was measured under a compound microscope equipped with an ocular micrometer. Fresh, dry pollen from glasshouse-grown plants transplanted from the Paugh Lake population was also measured for comparative purposes. The number and proportion of the stigmatic pollen load composed of pollen grains originating from long-, mid-, and short level anthers was then determined for intact and emasculated flowers of each floral morpho Statistical analysis of pollen load data employed a four-way fixed effects model analysis of variance, conducted using the GLM procedure of SAS (SASInst., 1982), on log-transformed data of the number of pollen grains deposited on stigmas of the floral morphs. Main effects were: 1) pollination treatment: intact flowers versus emasculated flowers; 2) floral morph: long-, mid-, and short-styled morphs (hereafter L, M, S,respectively); 3) pollen type: large-, medium-, and small-sized pollen (hereafter 1, m, s, respectively); 4) day of treatment (1-9). Following this analysis, separate two-way ANOVAs for intact and emasculated treatments were performed with floral morph and pollen type as main effects. Residuals from these analyses, containing the floral morph x pollen type interaction + error, reflect the relative amounts of each pollen type deposited on stigmas of the floral morphs. Mean residuals (R) were then tested for significant departure from 0 by single sample one tailed t tests. Darwin's hypothesis predicts significant (R) values for the legitimate floral morph-pollen type combinations (e.g., L/l, M/m, S/s).

Pollinator Visitation. If pollinators of P. cordata distinguish between intact and emasculated inflorescences, they may not visit them with the same frequency and, thus, may influence pollen grain deposition. To determine whether differences in visitation were evident, observations were conducted (Aug. 18-19) of the foraging behavior of bees on pairs of inflorescences of each floral morpho These were trimmed to the same flower number and were chosen for similarity in size, height above the water, and exposure. One was emasculated prior to flower opening and the other left intact. The three pairs were each observed for 10 minute time intervals for a total of 140 minutes during the two day period. The number of bees visiting each inflorescence as well as the number of flowers probed at each visit was recorded. Data were compared using G tests of independence.

Pollen-Stigma Interference.To evaluate whether the presence of stamens in a flower interferes with the capture of compatible pollen, comparison of the legitimate pollen loads of stigmas from intact and emasculated flowers were made. Mean pollen loads were compared by t tests. To investigate the potential importance of pollen-clogging to seed set in P. cordata, a field experiment involving hand pollinations and measurements of seed set was undertaken in a population at Lake Opinicon, Ontario, during July 1982. The experiment compared the seed set of inflorescences that received one of two treatments: 1) open-pollination; and 2) self-pollination by hand followed by open-pollination. The experiment commenced July 28 and involved a cohort of 60 inflorescences (10 inflorescences per treatment per morph) which commenced flowering on July 28-29. All flowers on an inflorescence received the same treatment. In self-pollinations the contents of a single anther were applied with fine forceps to stigmas, and an attempt was made to cover the entire stigmatic surface with pollen. The pollen of P. cordata is yellow and visible on the white stigmas. A single anther contains several thousand pollen grains. Hand pollinations were completed each day before insect activity had commenced. After 21 days, all inflorescences were bagged to facilitate the collection and counting of seeds approximately six weeks later. Comparisons of percentage seed set were made, following arcsine transformation, with pollination treatment and morph as main effects.