Introduction
In a paper published in Science in 1979 (Mulcahy, Reference Mulcahy1979), David L. Mulcahy hypothesized that pollen (microgametophyte) competition to fertilize ovules, in combination with the closed carpel (syncarpy) (Carr and Carr, Reference Carr and Carr1961; Armbruster et al., Reference Armbruster, Debevec and Willson2002) and insect pollination, in which many pollen grains are deposited simultaneously on the stigma (see Crepet and Niklas, Reference Crepet and Niklas2009), played a major role in the ‘rise of the angiosperms’ to dominance in the world's flora and vegetation. Mulcahy's hypothesis stimulated quite a bit of research to test the importance of pollen competition as an evolutionary force in plants. Seed germination was included in some of the studies in which the effect of pollen competition was tested on sporophyte (progeny) vigour. The purpose of this review is threefold: (1) to review briefly the concept of pollen competition; (2) to pull together the results of its effects on seed germination; and (3) to evaluate the evidence for the four criteria that must be met in order to validate Mulcahy's hypothesis.
Pollen (microgametophyte) competition: the concept
The idea here is that pollen (microgametophyte) selection (via intense competition induced by insect pollination and closed carpels) can influence the character of the sporophyte generation. Thus, (a) high pollen loads deposited on the stigma result in pollen tubes of the more fit pollen grains growing faster and thus fertilizing more ovules than the slower-growing pollen tubes of the less fit pollen grains; (b) conversely, deleterious mutations that cause slower pollen-tube growth are eliminated from the gene pool; and (c) genes of the fast-growing microgametophytes are passed on to, and increase the vigour of, the sporophyte, according to the following model (Mulcahy, Reference Mulcahy1971; Mulcahy and Mulcahy, Reference Mulcahy and Mulcahy1975, Reference Mulcahy and Mulcahy1987; Mulcahy et al., Reference Mulcahy, Mulcahy, Ottaviano and Mulcahy1975; Ottaviano et al., Reference Ottaviano, Sari-Gorla and Mulcahy1980, 1988; Winsor et al., Reference Winsor, Davis and Stephenson1987; Charlesworth, Reference Charlesworth1988; Schlichting et al., Reference Schlichting, Stephenson, Small and Winsor1990; Walsh and Charlesworth, Reference Walsh and Charlesworth1992; Quesada et al., Reference Quesada, Winsor and Stephenson1993, Reference Quesada, Winsor and Stephenson1996):

For example, genes of the superior (fast-growing) pollen that influence seed germination would be expressed in the sporophyte, i.e. seeds resulting from superior pollen would have higher germination percentages and rates than those resulting from pollination by inferior pollen.
Another aspect of the hypothesis is that the style is a selective arena that filters maladaptive pollen phenotypes. Thus, pollen vigour is just one of the microgametophyte characteristics that can be selected for (e.g. Searcy and Mulcahy, Reference Searcy and Mulcahy1985). Further, working with Hibiscus moscheutos (Malvaceae) Snow and Spira Reference Snow and Spira(1991a) put Mulcahy's hypothesis into the context of sexual selection in plants. They stated that ‘Differences between pairs of individuals [pollen donors] in pollen-tube growth rates were consistent across maternal plants [pollen recipients], suggesting that sexual selection could occur’. That is, sexual selection would occur via ‘super males’ siring a disproportionately high number of seeds across a wide range of females because of superior pollen-tube competitive ability (Snow and Spira, Reference Snow and Spira1991a, Reference Snow and Spirab, Reference Snow and Spira1996; see below for more about these three studies).
A second method used to study microgametophyte selection is to apply the same (excess) amount of pollen to the tip and to the base of the extended adaxial stigma surface of Caryophyllaceae species such as Silene latifolia. The idea here is that the degree of gametophyte selection is proportional to the distance travelled by the growing pollen tubes, i.e. pollination at the tip provides greater opportunity for competition than pollination at the base (Mulcahy and Mulcahy, Reference Mulcahy and Mulcahy1975, Reference Mulcahy and Mulcahy1987; Purrington, Reference Purrington1993; Lassere et al., Reference Lassere, Carroll and Mulcahy1996; Delph et al., Reference Delph, Weinig and Sullivan1998). This pollen-travel distance method has also been used to test for the effect of pollen-tube growth rate on sporophyte fitness with a heterostylous (distylous) species in which intramorph crosses are compatible (McKenna, Reference McKenna, Mulcahy, Mulcahy and Ottaviano1986), and in corn (Zea mays) in which the styles (silks) differ in length (Mulcahy, Reference Mulcahy1971; Ottaviano et al., Reference Ottaviano, Sari-Gorla and Mulcahy1980.
Methods and results
Altogether, we examined about 120 published papers, 70 of which are cited in the References, dealing with various aspects of Mulcahy's hypothesis. Thirty of these papers contained information on the effects of pollen load/pollen-tube growth rate on seed germination (including seedling emergence). The 30 case studies include 15 families, 19 genera and 22 species (Table 1). Fourteen (46.7%) of the case studies found that a high pollen load (HPL) or long-travel distance (L-DT) ( = high pollen competition, HPC) of pollen tubes in the style significantly increased percentage and/or rate of germination compared to a low pollen load (LPL) or to short-distance travel (S-DT) of pollen tubes in the style. In the other 16 cases, germination did not differ between seeds produced by HPL/L-DT vs. LPL/S-DT.
Table 1 Effect of pollen competition on percentage and/or rate of seed germination/seedling emergence

a yes = germination of seeds produced from a high pollen load, or from pollen that was required to grow a long distance in the style (i.e. style tip in Dianthus and Silene and the pin style of Anchusa), was significantly better than that of seeds produced from a low pollen load or from pollen that was required to grow over a short distance in the style (i.e. style base of Dianthus and Silene and short style of thrum morph of Anchusa). It could be argued that the ‘yes’ we have assigned to Silene vulgaris (Delph et al., Reference Delph, Weinig and Sullivan1998) should be a ‘no’.
b no = germination of seeds produced from a high pollen load or from a long-travel distance of pollen in style was not significantly better than that produced from a low pollen load or from a short-travel distance by pollen in style. The ‘no’ for Brassica campestris (Palmer and Zimmerman, Reference Palmer and Zimmerman1994) would have been a ‘yes’ if the study had not included the second sporophyte generation, which showed that the increase in percentage germination via a high pollen load was non-genetic.
In the 14 cases in which seeds produced by HPL germinated better than those from LPL, seed mass also was used as a sporophyte trait in 11 of them. In two of the 11, seed mass of HPL>LPL, in eight HPL = LPL and in one HPL < LPL. In the 16 cases in which germination of HPL was not better than LPL, seed mass was also determined in 12 of them: in one, HPL>LPL, in 10 HPL = LPL and in one HPC < LPC. Seed mass was not reported in seven of the studies.
Criteria for validation of Mulcahy's hypothesis
Several criteria must be met in order for pollen competition to be of significance in angiosperm evolution (Walsh and Charlesworth, Reference Walsh and Charlesworth1992), and here we will evaluate each of them.
More pollen grains are deposited on the stigma than are necessary to fertilize the ovules
Some degree of pollen limitation for fruit/seed set is widespread in angiosperms in natural habitats, and its degree of occurrence can vary among individuals (e.g. size, isolation and floral density); species; taxonomic families; years; times within the growing season; sites/fragments/populations (size and floral density); biomes (e.g. temperate vs. tropical); life histories (e.g. monocarpic vs. polycarpic); life forms (e.g. herbaceous vs. woody); breeding systems (self-compatible vs. self-incompatible); habitat/ecological conditions, e.g. changes caused by anthropogenic disturbance that lead to invasiveness and loss of native pollinators (Burd, Reference Burd1994; Larson and Barrett, Reference Larson and Barrett2000 and numerous references cited in these two reviews).
Pollen-tube growth rate (PTGR) is heritable via microgametophyte gene expression, and sperm from faster-growing pollen tubes must fertilize more ovules (i.e. sire more offspring) than those from slower-growing pollen tubes
Using pollen donors from two lineages of Raphanus raphanistrum that differed in ability to sire seeds, Snow and Mazer (Reference Snow and Mazer1988) showed that competitive ability of pollen derived from two previous generations of multiple-donor intense selection (i.e. intense pollen competition) was not more successful in siring seeds (i.e. number of seeds) in any of six female pollen recipients than pollen derived from two previous generations of single-donor weak selection (i.e. weak pollen competition). In fact, in two of the six female pollen recipients, pollen derived from previous weak competition was superior in siring seeds (i.e. sired significantly more seeds) than that derived from intense competition. Thus, a two-generation experiment on pollen competition (pollen load on stigma) did not detect heritability of this trait (Snow and Mazer, Reference Snow and Mazer1988).
Snow (Reference Snow1990) used the weak and intense pollination competition lines of Snow and Mazer (Reference Snow and Mazer1988) to test the effects of previous competition, (current) pollen load and donor diversity on progeny fitness in both greenhouse and field. There were four treatment combinations in her experimental design based on F2 lineage (previous competition), F3 (current) pollen load and F3 pollen donor diversity: (1) weak previous pollen competition, light F3 pollen load, one pollen donor; (2) intense, light, one; (3) intense, heavy, one; and (4) intense, heavy, three. In general, the pollination treatments did not significantly enhance seed mass; seed germination, i.e. neither percentage nor rate (speed) of seedling emergence; seedling survival; dry weight (biomass production); or fecundity (flower, fruit and seed production). In fact, treatment (4) had a significant negative effect on seedling emergence. Maternal parent had a significant effect on seed production and seedling emergence. Snow stated that ‘This study of wild radish failed to conform to the hypothesis that pollen competition leads to selection for superior progeny’.
Heritability (h2) in pollen grain germinability (PGG) and pollen-tube growth rate (PTGR) were high, i.e. 0.77 and 0.71, respectively, in Zea mays (Sari-Gorla et al., Reference Sari-Gorla, Pé, Mulcahy and Ottaviano1992). Most of the high genetic variability found in these two characters was due to quantitative trait loci (QTL) related to PGG or PTGR, ‘… which suggest that they are genetically controlled by a specific set of genes’ (Sari-Gorla et al., Reference Sari-Gorla, Pé, Mulcahy and Ottaviano1992).
F2 progeny from large F1 pollen loads of Cucurbita pepo× C. texana were significantly more vigorous than those from small pollen loads. Since the effects of pistil–pollen interaction, environmental maternal effects (seed mass, number and position in ovary) and non-random seed abortion were controlled for, it seems likely that the effects of pollen load were not due to non-genetic maternal effects or non-random seed abortion, but to genetic effects of pollen performance (Quesada et al., Reference Quesada, Winsor and Stephenson1993). In another study on C. pepo× C. texana by Quesada et al. (Reference Quesada, Winsor and Stephenson1996), two generations of selection (large pollen load) for pollen performance resulted in significantly fewer days to seedling emergence and marginally significantly (P< 0.10) higher seedling leaf area (at 14 d) than two generations of non-selection (small pollen load). The effects of pistil–pollen interaction, environmental maternal effects (seed mass, number and position in ovary) and non-random seed abortion were controlled for. In the field, seedlings from seeds fertilized by the selected (high-pollen) line were more vigorous, and plant reproductive output higher, than for the non-selected (low-pollen) line. Thus, selection on pollen-tube growth of the microgametophyte leads to an increase in sporophyte vigour (see Table 1 for more details about this study).
Havens (Reference Havens1994) found substantial phenotypic variation in in vitro pollen-tube growth rate (PTGR) within clonally derived offspring of Oenothera organensis, but the variation was not explained by genotype. Clonal repeatability for PTGR was only 9.4%, indicating that heritability for this trait was quite low. Effects of flower identity and plant identity on PTGR, on the other hand, were highly significant. Thus, Havens suggested that ‘… the environment may be more important than genotype in determining pollen performance in this species’.
Pollen-tube growth rates in Mirabilis jalapa were influenced by pollen load, but not by pollen diversity (i.e. number of pollen donors), thus not by potential genetic differences in pollen (Niesenbaum, Reference Niesenbaum1999). Niesenbaum concluded that pollen-tube growth rates were caused by interactions of pollen tubes in the style, i.e. heavy pollen load → more interactions → increased pollen-tube growth rate.
For the wind-pollinated species Betula pendula, there was significant variation in pollen-tube growth rate, and siring success was significantly correlated with pollen-tube length (growth rate). Consistency of ranking of the six pollen donors across the 11 pollen recipients was statistically significant (Pasonen et al., Reference Pasonen, Pulkkinen, Käpylä and Blom1999). In another study on this species by Pasonen et al. (Reference Pasonen, Pulkkinen and Käpylä2001), the only positive correlation between pollen-tube growth rate and progeny performance was for seed mass. The authors concluded that ‘… pollen-tube growth rate is not a good predictor of progeny performance in Betula pendula’.
Pollen performance can be affected by various environmental factors during pollen development, including herbivory, temperature and soil nutrients and pH (Lankinen, Reference Lankinen2000 and references cited therein). The effect of pollen load in Fragaria virginiana depended on post-germination life-history stage and on progeny growth environment, i.e. high vs. low nutrient level (Kalla and Ashman, Reference Kalla and Ashman2002). However, the siring success of Hibiscus moscheutos was not affected by salinity stress or soil nutrients (Snow and Spira, Reference Snow and Spira1996).
In the chasmogamous violet Viola tricolor, offspring fitness (seeds/capsule) in generation 3 was positively related to siring ability of father in generation 2, which was positively related to pollen-tube growth rate (Skogsmyr and Lankinen, Reference Skogsmyr and Lankinen2000). Heritability of pollen-tube growth rate (‘to a certain degree’) in V. tricolor was shown in a father–offspring regression.
A father–son regression for in vitro pollen germination percentage (germinability), which, along with pollen-tube growth rate, is the main component of pollen fitness (Mascarenhas, Reference Mascarenhas1990; Sari-Gorla et al., Reference Sari-Gorla, Pé, Mulcahy and Ottaviano1992), across six populations of Silene latifolia was significantly positive, suggesting that pollen germination percentage, which is related to pollen competitive ability, is heritable. F1 males reared in a common environment were used in crosses to avoid the possibility of confounding parental effects (Jolivet and Bernasconi, Reference Jolivet and Bernasconi2007). In vitro pollen-tube growth rate in Collinsia heterophylla differed among donors and showed significant narrow-sense heritability (36%) and evolvability in a father–offspring regression. Pollen-tube growth rates in vitro and in vivo were significantly correlated. However, pollen-tube growth rate did not correlate significantly with sporophyte fitness in either parental or offspring generation (Lankinen et al., Reference Lankinen, Maad and Armbruster2009).
Snow and Spira (Reference Snow and Spira1991a, Reference Snow and Spirab) found non-random paternity in Hibiscus moscheutos via variation in pollen-tube growth rate between pairs of pollen donors, and differences between pollen donors were repeatable in different seasons and under different growing conditions. Pollen donors with fast-growing (vs. slow-growing) pollen tubes sired a disproportionate (non-random) number of seeds, i.e. up to 72% vs. expected 50% if siring success was random (Snow and Spira, Reference Snow and Spira1991b). In another study on H. moscheutos, Snow and Spira (Reference Snow and Spira1996) reported that pollen donors differed in number of seeds sired, from expected 50:50 ratio to a maximum ratio of 68:32 via differences in pollen-tube competitive ability, i.e. 68% of seeds sired by pollen donors with fast-growing pollen tubes. In V. tricolor, siring success increased with PTGR, i.e. mating success was non-random, and variability in PTGR was high enough to overcome the stochastic factors of time and place of pollen deposition on the all-around receptive spherical stigma. That is, FAST pollen deposited later and/or further away from the ovules than SLOW pollen could still sire seeds (Skogsmyr and Lankinen, Reference Skogsmyr and Lankinen1999). These authors concluded that it seemed reasonable that PTGR could be selected for.
Overlap in expression of genes that confer advantage to pollen-tube growth of the microgametophyte and those that confer fitness to the sporophyte
In a review, Mascarenhas (Reference Mascarenhas1990) summarized the requirement for a genetic connection between pollen competition and sporophyte vigour, and concluded that ‘It is thus reasonable to expect that selecting for genes in the male haploid phase could increase the success of the sporophyte’. (Presumably, loci expressed for pollen-tube growth rate are the same as those expressed for sporophyte fitness traits.) However, seeds of Silene vulgaris from the first ovules to be fertilized, independent of pollen-tube growth rate (i.e. fast vs. slow) gave rise to the most vigorous progeny, i.e. larger seeds that germinated faster and larger plants (more biomass) (Delph et al., Reference Delph, Weinig and Sullivan1998). Thus, overlap in gametophyte and sporophyte vigour was not caused by overlap in their gene expression but by maternal effects. The authors suggested that seeds from ovules fertilized first give rise to more vigorous progeny because they were better provisioned by resources from the maternal plant than those fertilized later (see Table 1 for more details about this study).
The effects of pollen competition on the sporophyte are heritable, i.e. any effects on increased vigour in F1 should persist in F2, etc
Even though the effects of a high pollen load (pollen competition) are correlated with increased performance in the F1 progeny of some species, the evidence that these effects are passed on to progeny of future generations is equivocal. For example, Snow and Mazer (Reference Snow and Mazer1988) and Snow (Reference Snow1990) did not detect heritability of previous enhanced effects of pollen-tube competitive ability on offspring vigour of F3 of two generations of weak and intense selection for pollen competition in Raphanus raphanistrum (discussed above under heritability of pollen-tube growth rate). Positive effects of F1 progeny vigour in Cucurbita pepo of high (vs. low) pollen load applied to stigma of F0 plants were not transmitted to F2 progeny (Schlichting et al., Reference Schlichting, Stephenson, Small and Winsor1990). Evidence for heritability of pollen competitive ability across generations in C. pepo× C. texana was discussed above under heritability of PTGR (also see Table 1) (Quesada et al., Reference Quesada, Winsor and Stephenson1993, Reference Snow and Spira1996).
For Brassica campestris, Palmer and Zimmerman (Reference Palmer and Zimmerman1994) found second-generation effects of average number of seeds produced by crosses of F1 progeny derived from high and low pollen loads. Thus, pollen from plants sired under intense competition (HF1) sired twice as many seeds per application as did plants sired under relaxed (LF1) competition. These results suggest strong selection for pollen performance. However, there were no second-generation effects on seed number per fruit, total seed mass per fruit or percent germination per fruit (but P< 0.10 for the latter characteristic). Pollen selection (large pollen load) in C. pepo× C. texana over two generations led to production of pollen that gave rise to more vigorous progeny than pollen non-selected (small pollen load) for two generations (Quesada et al., Reference Quesada, Winsor and Stephenson1996; see Table 1 and text above on heritability of PTGR for more details about this study).
Is there support for pollen competition being a significant force in the diversification of angiosperms?
In short, the results of studies related to attempting to verify the four criteria for pollen competition of significance in angiosperm evolution are equivocal. Even David Mulcahy (Lassere et al., Reference Lassere, Carroll and Mulcahy1996) admitted that ‘Until more complete information is available for a variety of species, the importance of pollen competition remains open to interpretation’. Since publication of this statement in 1996, it seems that more studies have failed to support the evolutionary importance of pollen competition (e.g. Mitchell, Reference Mitchell1997; Pasonen et al., Reference Pasonen, Pulkkinen and Käpylä2001; Németh and Smith-Huerta, Reference Németh and Smith-Huerta2003; Armbruster and Rogers, Reference Armbruster and Rogers2004; Lankinen et al., Reference Lankinen, Maad and Armbruster2009) than have supported it (e.g. Brown and Kephart, Reference Brown and Kephart1999; Aronen et al., Reference Aronen, Nikkanen, Harju, Tiimonen and Häggman2002).
We think that a statement made by Lankinen and Madjidian (Reference Lankinen and Madjidian2011) pretty well sums up the situation regarding the significance of pollen competition in the evolution of angiosperms: ‘Even after 30 yr of research, the benefit of enhancing pollen competition is still not clear’. These authors found that size of pollen load (high vs. low) did not have a significant effect on most sporophyte traits measured in Collinsia heterophylla, including proportion of seeds that germinated (Table 1). Even more recently, Field et al. (Reference Field, Pickup and Barrett2012) showed that pollination intensity in the wind-pollinated, dioecious, annual species Rumex hastatulus increased the female sex bias ratio, but they found no evidence that pollen load influenced sporophyte vigour. Of the 30 case studies on seed germination (Table 1), 10 were published since 1996. Eight of the 10 do not support, and thus only two support, the importance of pollen competition as a possible evolutionary force in this stage of the life cycle. Further, it could be argued that the ‘yes’ we have entered for Silene vulgaris (Delph et al., Reference Delph, Weinig and Sullivan1998) in Table 1 could be considered a ‘no’. In which case, the only study on our list published since 1996 that reported a positive relationship between large pollen load and germination is the one by Brown and Kephart (Reference Brown and Kephart1999).
Thus, it seems that, ‘the jury is still out’ on this possibly very important issue concerning how angiosperms came to dominate the Earth's terrestrial flora by the mid to late Cretaceous/early Tertiary (Lidgard and Crane, Reference Lidgard and Crane1988, Reference Lidgard and Crane1990; Boulter et al., Reference Boulter, Gee and Fisher1998; Lupia et al., Reference Lupia, Lidgard and Crane1999, Reference Lupia, Crane, Lidgard, Culver and Rawson2000). However, even if the hypothesis is proven to be incorrect, the size of the pollen load deposited on the stigma seems to be another of the many factors that can cause seed germination percentage/rate, and other traits of F1 progeny, to vary – i.e. pollen load, even if genes in the microgametophyte are not heritable in the sporophyte, can apparently, via such factors as (genotype-dependent) pollen–pistil interaction, non-random seed abortion, differential provisioning of resources to seeds by the mother plant and parental plant environment, affect traits in the F1 (see Schlichting et al., Reference Schlichting, Stephenson, Small and Winsor1990; Snow, Reference Snow1990; Delph et al., Reference Delph, Johannsson and Stephenson1997, Reference Delph, Weinig and Sullivan1998).
It is well known that within a species large seeds, more often than not, germinate better than small ones (Baskin and Baskin, Reference Baskin and Baskin2014). Further, sometimes differences in seed mass can obscure the real results on studies testing the effects of certain factors on seed germination. For example, in Salvia pratensis, germination percentage and size of selfed seeds were significantly lower than they were for outcrossed seeds, and mean seed mass was significantly positively correlated with germination. Thus, the effect of selfing on germination disappeared when differences in seed mass were controlled for, which suggests that differences in germination were due to seed mass and not to inbreeding depression per se (Ouborg and Van Treuren, 1994). Does seed mass rather than seed vigour per se explain why seeds (in 46.7% of the cases in our survey) produced from HPL germinate better than those produced from LPL? Probably not. In the majority of cases in which germination of HPL>LPL (and seed mass was reported), seed mass of HPL = LPL. In general, then, seed mass was not a confounding factor. Thus, it seems that whatever the cause of the (sometimes) positive effects of HPL on seed germination, it is not likely to be seed size.
From a phylogenetic/evolutionary point of view, it is clear that PTGR increased greatly during the evolution and diversification of seed plants, being much faster in extant angiosperms than in gymnosperms (Williams, Reference Williams2012). Williams (Reference Williams2012) stated that ‘… pollen tubes of angiosperms have evolved across a spectacularly broad range of PTGRs that are faster than those of other seed plants’. The maximum recorded in vivo PTGR (35,000 μm h− 1) is for the phylogenetically advanced eudicot family, Asteraceae, which is 1.75 × 103 times faster than the maximum PTGR for extant gymnosperms (see fig. 4 in Williams, Reference Williams2012).
One final point: formerly, it was thought that the coevolution of flowering plants with insects was the major driving force for the evolution of the great diversity of both insects and angiosperms (see Gorelick, Reference Gorelick2001). However, the thinking now is that although insects probably participated in the major radiation of angiosperms in the uppermost Cretaceous and lower Palaeogene, the so-called ‘biotic pollination hypothesis’ certainly cannot, by itself, explain the great ecological and evolutionary success of flowering plants (Crepet, Reference Crepet1984; Gorelick, Reference Gorelick2001; Crepet and Niklas, Reference Crepet and Niklas2009). In fact, after critically reviewing the various hypotheses for angiosperm diversity, Gorelick (Reference Gorelick2001) concluded that ‘Coevolution with animal pollinators appears to be neither a necessary nor sufficient condition for large-scale diversification of seed plants’. Thus, Darwin's ‘abominable mystery’, i.e. the reason for the abrupt origin and rapid diversification of the angiosperms in the Cretaceous (Friedman, Reference Friedman2009), is now believed to be multifaceted and thus ‘… not the product of any one functional trait or syndrome of traits’ (Crepet and Niklas, Reference Crepet and Niklas2009).
Conflicts of interest
None.