1It has recently been suggested that not only nutrients but also presence of a competitor can influence competitive ability in plants, for example, by strategic changes in allocation to roots (a ‘tragedy of the commons’ effect). Such strategic changes might occur also in pollen, because this has potential to increase siring success during pollen competition in the pistil.
2I tested the new hypothesis that pollen competitive ability – a trait often strongly affected by resources of the pollen parent – was influenced by presence of a root competitor using glasshouse-grown Viola tricolor in 2 years. Plants of two maternal families were combined in pairs with their roots either separated or intermingled in the same amount of resources.
3Maternal families varied in response to root competition, with pollen performance increasing in some families, decreasing in others and unchanged in others. An increase did not mirror a decrease in the competing family, suggesting an explanation beyond differential ability to gain resources. The responses to competitive environment were often consistent across three independent competitors.
4There was a positive correlation between family responses in pollen performance and family responses in plant size to root competitors. Larger plants did not produce better pollen per se, indicating that the change in pollen performance was not a pure side effect of altered plant size. There was no support for a ‘tragedy of the commons’ effect on root production. With the experimental design used, an effect of rooting volume could not be completely ruled out.
5The currently unknown mechanism of the below-ground interactions between plants of V. tricolor was strong enough to change relative pollen competitive ability. These interactions might thus promote variation in the outcome of pollen competition.
6Synthesis. The results of this study indicate the presence of a competitor on pollen competitive ability beyond the effect of nutrients. Even though the underlying mechanism needs to be explored further, the detected link between pollen competition and soil competition suggests a phenomenon well worth investigating. Further studies may lead to increased understanding of the evolutionary consequences of selection operating in response to interactions with neighbours.
Strategic changes in the presence of a competitor might also be selected for in other competitive situations, such as during pollen competition in the pistil. When more pollen is deposited on the stigma than there are ovules to fertilize, a high pollen competitive ability should be favoured by selection (Snow & Spira 1991; see also review by Skogsmyr & Lankinen 2002). It is well known that pollen competitive ability is strongly influenced by environmental factors both during development and when deposited on the stigma (e.g. Delph et al. 1997; Lankinen 2001; Haileselassie et al. 2005; Hedhly et al. 2005). When a plant is growing under stressful conditions (e.g. due to herbivory, temperature or soil-nutrient stress) plants will provision their pollen less well, leading to reduced quality and quantity of storage products important for pollen competitive ability (Delph et al. 1997; Stephenson et al. 2003). Furthermore, the response on pollen competitive ability to growing condition is not only strictly environmental, because a large proportion of the genes expressed in the sporophytic life stage are also expressed in the gametophytic life stage (reviewed in Mulcahy et al. 1992; Walsh & Charlesworth 1992; Hormaza & Herrero 1994). Many studies have provided detailed evidence for how available resources influence pollen competitive ability. For example, soil nutrients, such as nitrogen and phosphorus, have been shown to affect size and nutrient concentration of pollen grains, germinability, speed of germination and/or pollen tube growth rate and siring success (e.g. Young & Stanton 1990; Lau & Stephenson 1993, 1994; Travers 1999; Lankinen 2000, but see Snow and Spira 1996). In most cases lower levels of nutrients had a negative effect on pollen performance, indicating a cost of producing highly competitive pollen. These studies focused on the effect of the resources per se and have not tried to evaluate the effect of the actual presence of a competitor.
The presence of roots of another plant could influence pollen tube growth rate for several reasons. First, if two competing plants differ in competitive ability, presence of a competitor could influence resource availability so that one competitor would gain an advantage at the expense of the other (influencing both pollen and sporophytic traits). Second, if plants can discriminate between self and non-self roots (Holzapfel & Alpert 2003; Gruntman & Novoplansky 2004; Falik et al. 2005), this ability could be used as a cue to elicit phenotypic plasticity in response to neighbours. Presence of neighbours may indicate a larger population and thus higher expected intensities of pollen competition, which should make it advantageous to increase pollen competitive ability. It might also be beneficial to invest more in pollen competitive ability relative to plants growing nearby if pollen from neighbours will end up on the same stigma. This is indeed probable as pollinators often visit flowers on plants growing next to each other and pollen carryover is common (e.g. Schaal 1980; Thomson & Plowright 1980; Rademaker et al. 1997). Third, if plants make costly adaptive changes to the presence of root neighbours, such as increased allocation to root production (e.g. Gersani et al. 2001), presence of roots might lead to reduced pollen tube growth rate due to a trade-off between this trait and root production. It would be informative to evaluate whether the presence of root neighbours can influence pollen performance beyond that of differential resource uptake, because such effects may have important consequences for variability of pollen competitive ability. Investigating how root competition affects pollen traits in relation to other traits is important also for a more complete understanding of plant plasticity to neighbours (cf. Cahill 2002; Murphy & Dudley 2007).
In this study my aim was to investigate if pollen competitive ability is affected by the presence of a competitor in the soil using glasshouse-grown Viola tricolor L. over 2 years. I planted individuals of two different maternal families in paired pots where the roots were either kept separate in each pot or intermingled in both pots. I asked (i) whether pollen competitive ability was influenced by root treatment, (ii) whether changes in pollen performance in response to root treatment was consistent when competing with different genotypes, and (iii) whether changes in pollen performance were resource-based. To answer the first question I measured pollen tube growth rate in vitro in treatments with and without a competitor, and to answer the second question I allowed maternal families to compete with three other genotypes and with siblings (and assessed pollen performance). To answer the final question I first determined whether the change in pollen tube growth rate of both maternal families in a combination showed an average change. Lack of an average change would result if a resource-based increase in one family is counteracted by a loss in the other. Second, I investigated how the change in pollen performance was related to effects on plant size and root production. It should be noted that the effect of rooting volume was not controlled for in the experimental design used here (see Schenk 2006; Hess & de Kroon 2007; Semchenko et al. 2007a).
Viola tricolor is a hermaphroditic annual present on dry hillsides, flat rocks, sand-dunes and cultivated lime-deficient soil in nearly all of Europe and Asia (Lagerberg 1948; Mossberg et al. 1992). Plants often grow close together in natural populations, but plants growing more alone are also found. Plants can have extensive root systems spreading over some distance. This makes it reasonable to assume that root competition occurs among plants growing next to each other. The species is mostly out-crossing and insect pollinated (Lagerberg 1948; Elfving 1968). Some self-pollination can occur (Lagerberg 1948), though in the glasshouse flowers prevented from pollination produced fewer seeds (Skogsmyr & Lankinen 1999).
Seeds were collected from a large population (several hundred individuals) in south Sweden during the summer of 2001. In the experiment, plants derived from seeds produced by the same mother plant are referred to as ‘maternal family’. To increase the probability that the maternal families used in the experiment were of different genetic origin, I collected seeds from five different parts of the population (at least a few hundred meters apart). The experiment was conducted in the autumn of 2001 and in the summer of 2004.
To determine whether sharing soil influenced pollen competitive ability, I transplanted seedlings of two competing maternal families in two-pot treatments (square pots taped together, 8 cm width × 6.5 cm height) with roots either kept separate or allowed to intermingle in both pots (Fig. 1, cf. experimental design of Gersani et al. 2001). For practical reasons (e.g. data analyses, see below) I randomly denoted one maternal family ‘focal’ and the other one ‘competitor’, but both families were treated in the same way (e.g. identical measurements were conducted). To be able to unite roots of two individuals and still keep conditions similar to the treatments with separate roots, seedlings were planted over a shallow hollow cut in the upper part of the two joining pot sides (4.5 cm width × 1.5 cm height). Under both treatments, seedlings were placed close to the adjoining pot side. The type of soil used was a mixture of sand and ordinary compost for pot plants (peat with clay and silica added) (1 : 4).
In treatments with shared soil, roots of each competitor were equally divided between two pots (50% in each pot). The hollow between pots were too shallow to allow new roots to grow between pots. Prior to potting, I carefully removed as much of the old soil as possible from the roots, that is, all roots of all plants were treated in the same way. This procedure resulted in some loss of the smallest roots, but this damage was not different from that of the normal transplanting practice. Violets do not seem to suffer at all from transplantation to larger pots and they are also easy to grow in the glasshouse. Seedlings were about 2-months-old at the time of transplantation (average plant height = 5.3 ± 3.0 (SD) cm (2001); 6.49 ± 2.6 (SD) cm (2004)). They had produced their first flower or were about to start flowering. Until transplantation seedlings of the same maternal family were grown together in a large tray, but at this time roots were too small to intermingle.
In the main experiment, 14 maternal families were paired in seven combinations in 2001, and 22 maternal families were paired in 11 combinations in 2004, that is, all maternal families only had just one competitor. Number of plants in 2001 equalled 2 competition treatments × 7 combinations × 2 families per combination × 3 replicates = 84 (Fig. 1). However, because plants stopped flowering faster than predicted this year (probably due to a lack of day light when grown in the autumn), this reduced the n-value for pollen competitive ability (to 79; see Appendix S1 in Supplementary material for details) but not for plant size. It should be noted that there was no difference between treatments in how fast plants stopped flowering (Appendix S1). Number of plants in 2004 amounted to 2 competition treatments × 11 combinations × 2 families per combination × 3 replicates = 132 (Fig. 1). To control for size differences among plants (Weiner 1990; Laird & Aarssen 2005), I combined maternal families that produced seedlings of similar size. To try to avoid combining close relatives, I picked two plants that originated from different areas of the population.
To further investigate whether response to root treatments were consistent when competing with different genotypes, I made additional combinations in the intermingled root treatment. In 2004, a subset of 12 maternal families (belonging to six of the 11 maternal family combinations) were allowed to grow with two additional competitors. Total number of additional plants equalled 6 combinations × 2 families per combination × 2 additional competitors × 3 replicates = 72. The additional combinations were constructed by pairing the six original family combinations into three groups. Within these groups each maternal family were allowed to share soil with both maternal families of the other original combination, that is, in total, all pair wise combinations were made for the four maternal families within a group (main experiment + two additional combinations). This resulted in the combination of seedlings that differed in size. In 2001, two siblings of the focal plant family were planted together in five of the seven focal plant families (Fig. 1, number of plants = 5 focal families × 2 siblings per combination × 3 replicates = 30, whereof I was able to assess pollen tube growth rate on 27, see Appendix 1 for details). All maternal family combinations of all treatments were replicated three times.
All replicates of a particular maternal family combination were placed together in the glasshouse in order to minimize potential microclimatic effects within combinations. Differences among maternal family combinations could thus be due to environmental factors. One exception to this arrangement was the additional plants involved in testing for consistency of families across unrelated competitors. These plants were placed randomly in the glasshouse, but all three replicates of an additional combination were kept together. All maternal family combinations were rotated a few times to try to minimize microclimatic effects (in connection with rearrangements made to provide more space as the plants grew).
Plants were allowed to grow in the different treatments in the glasshouse for approximately 2 months before data sampling began. To ensure the effects of root competition could be detected, I wanted the plants to be large enough so that their roots had been occupying much of the space in the pots for some time, that is, towards the end of their life cycle. Since violets are flowering constantly from the time they are large enough to produce flowers until the end of their life, it is possible to measure pollen competitive ability late in the life cycle. No effect of plant age has been detected on pollen competitive ability in normal looking violet flowers (some much smaller, more selfing flowers with slower pollen tube growth rate tend to appear at the end of the life cycle; Skogsmyr & Lankinen 1999; Lankinen 2000).
measurements of pollen competitive ability and sporophytic traits
As an indication of pollen competitive ability I measured pollen tube growth rate in vitro, since previous evidence supports that this is a good measure of siring ability following two-donor crosses in V. tricolor (Skogsmyr & Lankinen 1999, 2000; Lankinen & Skogsmyr 2002). Pollen tube growth rates are correlated in vitro and in vivo in this species, but measurement error is much smaller for the in vitro method (Lankinen 2001).
Pollen was germinated in Hoekstra medium (Hoekstra & Bruinsma 1975) for 2.5 h (2001) or 2 h (2004) in a dark chamber at a constant 21 °C. Pollen tubes reached approximately similar length in both years. Since glasshouse temperature can have a large impact on pollen tube growth rate in V. tricolor (Lankinen 2001), flowers were always collected at the same time every day. Pollen from all replicates of a particular maternal family combination was germinated at the same time. Pollen from three flowers per individual was used. As an indication of pollen tube growth rate, I measured the length of the eight first tubes encountered in the microscope view. The average length of these eight pollen tubes was used for statistical analysis. Since previous studies have shown a high repeatability of in vitro measurements (Lankinen 2001), pollen was in most cases germinated once for each plant. Collection of the data took approximately 2–3 weeks in 2001 and 4 weeks in 2004.
In both years, final plant size was recorded as the total length of all shoots. In 2004, I also estimated dried root weight. Plants grown together had to be harvested as one unit, that is, the roots of the two plants could not be separated. Roots were rinsed in water and dried at 70 °C for 24 h. Root samples were stored at 40 °C until weighing.
Sporophytic traits were recorded after pollen competitive ability. Maternal families were assessed in the same order as for pollen competitive ability. In 2001, measurements were made over a period of 1 week. In 2004, measurements were made during 3 weeks. Both plant size and root weight was recorded at the same time for each maternal family combination.
To investigate effect of separate and intermingled root treatment on pollen tube growth rate and plant size, I used a GLM with maternal family (random effect), root treatment (fixed effect) and their interaction (main experiment, analysis 1). Plant size was used as a covariate in the model for pollen tube growth rate. The interaction between the covariate and root treatment was estimated to evaluate the effect of root competition on allocation, that is, if the slope between pollen tube growth rate and the covariate differed between treatments. To obtain statistically independent samples I only included focal maternal families in the test. The overall significance of tests for both years was calculated following Fisher (p. 795, Sokal & Rohlf 1995).
To determine whether the change in pollen tube growth rate or sporophytic traits of both maternal families in a combination showed an average change (or was a result of resources increasing in one plant and decreasing in the other), I used a GLM with maternal family nested within family combination (random effect), family combination (random effect), root treatment (fixed effect), and the interaction between family combination and root treatment (main experiment, analysis 2). Because it was not possible to separate roots of the two maternal families in the intermingled root treatment, family was not included in the model for root weight. Plant size was used as a covariate in the model for pollen tube growth rate and root weight, and the interaction between the covariate and root treatment was estimated. Due to missing values in 2001 (see Experimental set-up) number of family combinations included in the test was reduced from seven to five for pollen tube growth rate. I only included combinations with at least two recorded replicates of both competing plants.
To analyse if pollen tube growth rate in the intermingled root treatment was consistent for a maternal family across three competing families, I used a GLM with the random effects maternal family and competitor, and their interaction. Plant size was included as a covariate. The effect of relatedness when sharing soil was analysed with a mixed effect anova with the factors kin nested within root treatment (fixed effect), maternal family (random effect), root treatment (fixed effect) and the interaction between family and treatment.
To be able to compare the effect of root treatment on pollen performance with the effect on sporophytic traits, I correlated response to root treatment for these traits. I estimated the response to root treatment as a (arcsine-transformed) quotient of the performance in intermingled root treatments divided with the sum of both separate and intermingled root treatments. Untransformed values ranged between 0 and 1, where 0.5 indicated no response to the treatment. Values below 0.5 indicated a relatively better performance when grown alone, while values above 0.5 indicated that plants were relatively superior when grown together. I used the average of maternal families (pollen performance vs. plant size) or the average of maternal family combinations (pollen performance vs. root weight).
Most statistics were conducted using spss 11.0 (1999). Type III SS was used in all GLMs. Root weight was log-transformed. The distribution of the other variables did not deviate significantly from normal.
pollen performance in separate vs. intermingled root treatment
In the experimental year with more data (2004), a significant maternal family by root treatment interaction indicated that the effect on pollen tube growth rate when sharing soil varied among focal maternal families (Fig. 2, Table 1). In 2001, there was a non-significant trend in the same direction (Fig. 2, Table 1). Combining the P-values of both years showed an overall significant interaction (Fisher: χ2 = 13.55, d.f. = 4, P < 0.01). No significant main effect of root treatment was found (Table 1). Considering how pollen tube growth rate of all maternal families (focal and competing) responded to growing together rather than alone, there was a significant difference between treatments in 21.4% (3 out of 14) of all maternal families in 2001 and in 22.7% (5 out of 22) of all maternal families in 2004 (Appendix S1). In 2001, all three significant maternal families showed decreasing pollen tube growth rate when growing with intermingled roots (by 31.3%). In 2004, three maternal families showed increasing (by 23.5%) pollen tube growth rate and two showed a decrease (by 24.0%; Appendix S1).
Table 1. GLM (mixed model) for in vitro pollen tube growth rate and final plant size when plants of a focal maternal family were grown separately or with their roots intermingled with plants of a competing seed family in years 2001 and 2004 (main experiment, analysis 1). Final plant size was used as a covariate for pollen tube growth rate, and plant size by root treatment interaction was estimated
Source of variation
In 2001: 2.5 h, in 2004: 2 h.
Denominator MS = MS (Error).
Denominator MS = MS (Maternal family × treat).
Denominator MS = 0.034 MS (Maternal family × treat) + 0.966MS (Error).
Denominator MS = 0.033 MS (Maternal family × treat) + 0.967MS (Error).
If the detected differences for focal maternal families were only a result of differential resource uptake of the two competitors, an interaction effect between root treatment and maternal family combination (focal and competing) should not be expected. Testing how pollen tube growth rate of maternal family combination responded to root treatment showed a significant family combination by root treatment interaction in both experimental years (Table 2). This result indicates that the effect of growing with a competitor varied not only among maternal families but also among combinations. Furthermore, there was a significant difference between maternal families within combinations in both years, but no difference among combinations (Table 2).
Table 2. GLM (mixed model) for in vitro pollen tube growth rate and sporophytic traits when plants of two maternal families (focal and competitor) were combined either with their roots separated or intermingled in years 2001 and 2004 (main experiment, analysis 2). Maternal family was nested within family combination (comb) for pollen tube growth rate and final plant size. Final plant size was used as a covariate for pollen tube growth rate and root weight, and plant size by root treatment interaction was estimated
Distribution of pollen-response to root treatment of two families within combinations (i.e. both showing increasing or decreasing response, or one showing increasing and one showing decreasing response) could either be expected to be random, or one type could be more common than the other. If it is more common with combinations where both families respond in the same way, this might for example, be an effect of a similar microclimate. The number of family combinations where both maternal families showed, on average, either an increase when growing in contact (n = 6) or a decrease (n = 4) did not differ from the number of maternal family combinations where one maternal family showed an increase and the other a decrease (n = 6) (expected random distribution: 8 : 8; two-tailed binomial test: P > 0.2, Appendix S1). This suggests that the distribution of the pollen-response to root treatment of two maternal families within combinations could not be separated from a random distribution.
When pollen donors respond differently to root competition, relative pollen performance may also be affected, that is, some donors may produce relatively better pollen when sharing soil compared to when growing alone. In 2004, the effect of root competition on pollen tube growth rate was strong enough to alter ranking order of donors, as this trait was not correlated between treatments (Pearson correlation: r = 0.364, n = 22, P = 0.096). In 2001, ranking order of donors was consistent between treatments (Pearson correlation: r = 0.790, n = 14, P = 0.001).
consistency of pollen performance in root treatments across competitors
Pollen tube growth rate in the intermingled root treatment was significantly affected by maternal family but not by competitor when a subset of maternal families was combined with three other families in 2004 (Fig. 3, Table 3). In 9 out of the 12 maternal families direction of change in pollen performance from separate to intermingled treatment was consistent across competitors (Fig. 3, one-tailed binomial test: P = 0.015). Seven of the nine families showed increasing and two families showed decreasing pollen tube growth rate. For the five maternal families with at least one out of three significant differences between treatments, the direction of change was always consistent (four showed increasing and one showed decreasing pollen tube growth rate, 6F, 6C, 7C, 10C and 11C in Fig. 3, one-tailed binomial test: P < 0.031).
Table 3. anova (random effects) for pollen tube growth rate in vitro when plants of a maternal family were grown with their roots intermingled with plants of three competing families (competitor) in year 2004. Final plant size was used as a covariate
Source of variation
Final plant size
Maternal family × competitor
There was a difference between how siblings responded to growing together compared to when unrelated plants were combined together in 2001 (Fig. 4, Table 4). When the competitor was a sibling there was no detected difference in pollen tube growth rate when grown separately or with intermingled roots. In two of the five maternal families included in the analysis, pollen performance did decrease significantly when grown with an unrelated maternal family (Fig. 4, Appendix S1).
Table 4. anova (mixed model) for pollen tube growth rate in vitro when plants of a focal maternal family were grown with their roots separated or intermingled either with plants of a competing family or with a sibling in year 2001. The effect of relatedness (kin) was nested within root treatment
Source of variation
Denominator MS = MS (Error).
Denominator MS = MS (Maternal family × treat).
Denominator MS = 0.978 MS (Maternal family × treat) + 0.022MS (Error).
sporophytic performance in separate vs. intermingled root treatment
There was no significant effect on plant size of growing with separate or intermingled roots either when testing focal maternal families (Table 1) or when testing family combinations (Table 2). There was, however, a positive correlation between how plant size and pollen tube growth rate of maternal families responded to root treatment (Partial correlation (controlling for year and maternal family combination); rp = 0.412, d.f. = 33, P = 0.014). Larger plants per se did not produce more competitive pollen (Tables 1–3). In the analysis involving only focal plant families, root treatment altered the slope between plant size and pollen tube growth rate in 2004 (marginally significant) and the same non-significant trend was seen in 2001 (Table 1). The change resulted in a slight increase in allocation to pollen performance relative to plant size in the intermingled root treatment (2001: separate, y = 0.69 – 5.9 × 10−5 x, intermingled, y = 0.64 – 1.9 × 10−5 x; 2004: separate, y = 0.57 + 0.32 × 10−5 x, intermingled, y = 0.53 + 6.3 × 10−5x). Combining the result of both years indicated that the change in the slope was significant (Fisher: χ2 = 10.07, d.f. = 4, P < 0.05). In the analysis involving both maternal families in a combination (focal and competing), a similar change in the slope between pollen performance and plant size was seen (Table 2, combining both years: Fisher: χ2 = 14.50, d.f. = 4, P < 0.01; 2001: separate, y = 0.72 – 17 × 10−5 x, intermingled, y = 0.64 – 9.8 × 10−5 x; 2004: separate, y = 0.55 + 4.3 × 10−5 x, intermingled, y = 0.54 + 8.6 × 10−5 x).
No significant effects were found on root weight in separate and intermingled root treatment apart for an effect of family combination (Table 2). Furthermore, there was no relationship between how root weight and pollen performance (average of family combination) responded to root treatment (Pearson correlation; r = −0.204, d.f. = 9, P = 0.55).
When two maternal families of V. tricolor experienced root competition in the glasshouse, pollen tube growth rate of the focal family increased, decreased or was unaffected compared to growing alone in the same amount of resources. The variation in response was not only caused by an opposite response in the two competing families. The responses to root competition were often highly genotype-specific, and siblings were not affected by sharing soil. These results suggest that pollen tube growth rate in this species is not only influenced by nutrients, but also by interactions with other plants below-ground.
presence of a neighbour and differential resource availability
Environmental factors during pollen development often influence pollen competitive ability (Delph et al. 1997; Stephenson et al. 2003). In violets, for example, both temperature and soil phosphorus have previously been found to affect pollen tube growth rate (Lankinen 2000, 2001). In the present study, however, I asked if pollen tube growth rate was influenced by presence of a competitor by comparing a competitive and a non-competitive root-environment containing the same net amount of soil. I did detect a response to root treatment. The nature of the response varied substantially between maternal families and only about 20% of all investigated maternal families responded significantly to root treatment. To some extent this might be explained by the fact that the sample size was too low to discover small differences, or the result was only a product of chance. The magnitude of the detected changes were indeed strong (on average about 25%) compared to when altering soil phosphorus (maximum change 13%, Lankinen 2000). Furthermore, finding a consistent direction of the response to root competition in 9 out of 12 maternal families competing with three competitors is unlikely to have occurred by chance.
If the only mechanism influencing the outcome in root competition is that one of the two plants in a combination is superior at taking up nutrients or water (Callaway et al. 2003; Schenk 2006), pollen tube growth rate of one of the maternal families should increase and pollen tube growth rate of the other should decrease when grown together. Instead, I rather found that pollen tube growth rate of both competitors increased, decreased or was on average unaffected, indicating that differential resource accumulation rate was not the sole explanation for the result.
Root treatments of the two families in a combination were placed together in the glasshouse, that is, they shared a common microenvironment. This similar environment could have influenced the response to root treatment in a family combination. On the other hand, this is less likely because of the high consistency when competing with the two other competitors placed randomly in the glasshouse. Furthermore, the two families in a combination did not show a similar response more often than what should be expected by chance.
presence of a neighbour – cue or constraint for pollen performance?
In the current study, pollen response to root competition was often consistent across unrelated competitors, but there was no detected effect when siblings shared soil compared to growing alone. If root interactions function as a cue in violets, there is no particular reason why plants should respond differently depending on the genetic identity of the neighbour, unless the competitor is a sibling. The sibling competition hypothesis states that because relatives are more equal in their capacity to compete (e.g. rates of nutrient uptake), competition should be stronger between siblings (e.g. higher probability of using the same competitive strategy) than between unrelated individuals (Cheplick 1992). Studies have found both stronger (e.g. Delesalle & Mazer 2002; Cheplick & Kane 2004) and weaker (e.g. Willson et al. 1987; Donohue 2003) effects when plants are competing with their relatives rather than with unrelated plants. Weaker effects might indicate kin selection effects (Hamilton 1964), such as recognition and avoiding roots of siblings (cf. self/non-self recognition systems, e.g. Holzapfel & Alpert 2003). In a recent study on Cakile edentula, plants growing with strangers showed higher allocation to roots than when growing with their siblings, indicating that root interactions may function as a cue for kin recognition (Dudley & File 2007).
Even though there was no significant effect of root treatment on plant size in the present study, the response to root treatment was positively correlated for pollen tube growth rate and plant size. This correlation suggests a similar root treatment–response for plant size and pollen performance, though the former response was weaker. This result is in line with the commonly occurring genetic overlap between sporophytic and gametophytic life stages (reviewed in Mulcahy et al. 1992; Walsh & Charlesworth 1992; Hormaza & Herrero 1994), or both traits are dependent on a third factor, that is, general fitness. Because there was no support that larger plants per se produced more competitive pollen, it is unlikely that the detected response on pollen competitive ability is only a side effect of a response on plant size. Interestingly, there was even a (marginally significant) change in the allocation pattern between plant size and pollen performance, so that plants invested relatively more in pollen competitive ability when they shared soil.
I could not find any evidence that violets altered their allocation to root production in response to root competition, that is, a ‘tragedy of the commons’ effect found in other experiments where presence and absence of root competition was varied in the same amount of resources (Gersani et al. 2001; Maina et al. 2002; O’Brien et al. 2005). The design of these experiments, including the design of my experiment, does not, however, keep total rooting volume constant. This might be a problem if plants respond to a larger rooting volume, as has been shown in a number of recent studies (Schenk 2006; Hess & de Kroon 2007; Semchenko et al. 2007a). Even though I could neither detect any significant difference between treatments nor a correlated effect for root weight and pollen tube growth rate, it is indeed possible that the detected difference between growing alone or growing together was influenced by the difference in rooting volume.
It is hard to understand the mechanism behind the detected response to root competition in V. tricolor from this study. Even though I did not find any evidence for a trade-off between pollen performance and root weight when experiencing root competition, it is still possible that plants have to pay costs in the presence of neighbours. For example, if plants have a polymorphic ability to compete below-ground (cf. Cahill et al. 2005) some competitive strategies may benefit from growing together while others are doing worse when combined, and the outcome may indirectly influence pollen competitive ability. Other possible explanations for the result of this study include changes in a trade-off between number and quality of pollen, or changes in allocation to male and female function. The former explanation is probably less likely, because plants producing low quality pollen also seemed to have flowers with less pollen (personal observation). It has frequently been shown that sex allocation in hermaphrodites of gynodioecious species can differ depending on environmental factors (Ashman 2002). If violets growing together do invest relatively more in pollen than in plant size (as my results may indicate) and plant size is correlated to seed production, this might suggest that differential allocation to male and female function at least partly influenced the change in a competitive environment.
Despite the currently unknown mechanism of the below-ground interactions in this study, at least in the second year (with more data), the effect of root competition was strong enough to alter relative performance of pollen donors. Because previous studies have confirmed that pollen tube growth rate in this species is important for siring success (Skogsmyr & Lankinen 1999, 2000; Lankinen & Skogsmyr 2002), intraspecific interactions below-ground might have important consequences for variability in pollen competitive ability in natural populations, such as how it can be maintained when selection is strong (Fisher 1958; Walsh & Charlesworth 1992).
This study on V. tricolor showed a previously undetected link between root competition and pollen competition. Even though nutrient availability influences pollen competitive ability in this species (Lankinen 2000), this trait was affected by presence of a competitor in the soil. Maternal families varied substantially (increased, decreased or were unaffected) in how their pollen performance responded to root competition when grown in the same amount of resources as when grown alone. This variation was not only an effect of an opposite response in the two competing families. The responses to root competition were often highly genotype-specific. Even though it is not possible to unravel the mechanism underlying the link between root competition and pollen competition without further studies, the findings of this study suggest that there is a phenomenon worth pursuing. Increased knowledge in this area may be important for our understanding of how selection operates on plants in response to interactions with neighbours and for our understanding of the evolutionary consequences of such selection.
The first year of the experiment was performed at IBLS, Glasgow University during a research visit in 2001. I wish to thank L. Delph, A.-M. Fransson, T. D’Hertefeldt, R. Härdling, J. Järemo, I. Skogsmyr and anonymous referees for helpful discussions and comments on previous versions of the text, and R. Cowan for taking care of the plants in 2001. The work was supported by the Swedish Research Council and Knut and Alice Wallenberg Foundation.