How do floral display size and the density of surrounding flowers influence the likelihood of bumble bee revisitation to a plant?

Authors


†Author to whom correspondence should be addressed. E-mail: makinott@pe.ies.life.tsukuba.ac.jp

Summary

  • 1Most pollination biologists have used the collective pollinator visits to a plant as the measure of its pollinator attraction. However, we know very little about how many returns by the same individuals compose these visits, and how far each visitor travels after leaving the plant. Such behavioural aspects of individual pollinators are essential to understand the patterns of pollen flow among plants.
  • 2We observed plant visits by tagged bumble bees Bombus diversus in a field population of Cirsium purpuratum. By dissecting the collective visitation data into visits made by individual foragers, we addressed how ‘visitor density’ (number of individuals that visited a plant per 2 h) and ‘individual visitation rate’ (number of visits made by each individual per 2 h) are related to floral display size (number of flowering heads on a plant) and local flower density (number of flowering heads on neighbouring plants). We also tracked individual bees to determine how display size and local flower density of a plant influences its relative position in a bee's foraging area.
  • 3Plants attracted both regular visitors (bees that visited a plant more than three times per 2 h) and occasional visitors (bees that visited a plant fewer than four times per 2 h). Densities of both types of visitors increased with floral display size, whereas only occasional visitor's density increased with local flower density.
  • 4Individual bees preferred to visit central plants within their own foraging areas, plants with larger displays, and plants with lower local flower density. However, these preferences were independent from one another. Plants with large displays were not necessarily chosen by a bee as the centre of its own foraging area. On the other hand, plants with high local flower density were often located near the centre of a bee's foraging area.
  • 5The observed pollinator movements have implications for pollen flow in the plant population. Plants with larger displays probably experience greater mate diversity by attracting more occasional visitors, but they also assure matings with particular plants by increasing returns from regular visitors.

Introduction

Plants within a population often vary in the numbers of open flowers (floral display size) and also in the numbers of open flowers on neighbouring plants (local flower density). Variation in these plant characters makes alterations in pollinator responses, which can influence pollen dispersal. For example, larger floral displays attract more pollinators per unit of time (Klinkhamer and de Jong 1990; Ohara and Higashi 1994; Robertson and Macnair 1995; Goulson et al. 1998; Ohashi and Yahara 1998, 2002; Vrieling et al. 1999; Makino and Sakai 2004; Mitchell et al. 2004; Grindeland, Sletvold and Ims 2005; Miyake and Sakai 2005). Similar effects have been reported for plants with higher local flower density (Klinkhamer and de Jong 1990; Dreisig 1995; Grindeland et al. 2005). This increased attractiveness promotes increased pollen receipt or removal, or potential mate diversity (Galen and Stanton 1989; Harder and Thomson 1989; Young and Stanton 1990; Wilson and Thomson 1991; Harder and Barrett 1996; Jones and Reithel 2001; Engel and Irwin 2003). Thus floral display size and local flower density may have similar effects on plant fitness in terms of pollinator attraction.

Most previous studies have measured such pollinator attraction as total visitation rate per plant, i.e. collective visits to a plant made by unidentified individual pollinators. However, total visitation rate could be composed of frequent return visits made by one or a few individuals, or it could also be composed of occasional visits made by many different individuals. Indeed, Williams and Thomson (1998) showed substantial variation among individual bumble bees in their frequency of visits to a Penstemon plant, by dissecting the collective data into visits made by marked individuals. In other words, total visitation rate observed on a plant can be divided into two components: ‘visitor density’ (number of individuals that visit the plant per unit of time) and ‘individual visitation rate’ (number of visits made by an individual per unit of time).

These two components of pollinator visitation may have different influences on pollen dispersal from plants. Individual pollinators, such as bumble bees and hummingbirds, often confine their foraging to small areas within a larger plant population, and they partially overlap their foraging areas (Heinrich 1976; Thomson, Maddison and Plowright 1982; Thomson, Peterson and Harder 1987; Gill 1988; Thomson, Slatkin and Thomson 1997; Comba 1999; Makino and Sakai 2004). This means that pollen grains removed by different pollinator individuals are likely to be delivered to stigmas of different sets of plants. Therefore, an increase in visitor density is likely to enhance the potential mate diversity. In contrast, an increase in individual visitation rate will reduce potential mate diversity of a plant, while ensuring pollen exchange with a certain set of conspecific mates. To understand how potential mate diversity and opportunities for pollen exchange with certain mates can be related to floral display size and local flower density, therefore, we have to know the relationships between these plant characters and the two components of total visitation rate, i.e. visitor density and individual visitation rate.

In addition to its effects on pollinator attraction, variation in floral display size or local flower density may have an influence on the distance of pollen dispersal from a plant, by affecting the probability of the plant being chosen as the centre of pollinators’ foraging areas. From observations of bumble bees foraging on Aralia plants, Thomson et al. (1982, 1987) suggested that bees repeatedly visited a core set of plants while sampling other plants only occasionally, and that they gradually moved their foraging areas into regions where floral rewards were higher. If plants with larger displays or higher local flower densities provide higher rewards and lower travel costs, these plants may be chosen more frequently by pollinators as the centre of their foraging areas. If this is the case, plants with larger displays or higher local flower densities may have shorter distances of pollen dispersal; distances between central and peripheral plants are generally shorter than those between peripheral plants located at diametrically opposite positions within a foraging area.

In this study, we examined the visitation patterns of individual bumble bees Bombus diversus to plants with different floral display sizes and local flower densities, in a field population of Cirsium purpuratum (Maxim.) Matsum. First, we marked individual foragers and plants with numbered tags, and then quantified the visitor density and the individual visitation rates for each plant. Next, we recorded the interplant movements of individual foragers to examine how floral display size, local flower density, and the relative position within a foraging area influence the number of return visits made by each forager. Finally, we addressed how the relative position of a plant within the foraging area of a visitor relates to its floral display size and local flower density.

Materials and methods

organisms and study site

Cirsium purpuratum is a herbaceous perennial that inhabits flood plains or volcanic barrens in the Kanto and Central Districts of Honshu, Japan. It produces large, purple, nodding flower heads (400–700 florets per head) on several erect flowering stalks that elongate from a basal rosette. All florets within a flower head are hermaphroditic and protandrous. The number of flowering heads provides a practical measure of floral display size for a C. purpuratum plant (Ohashi and Yahara 1998, 2002). We studied C. purpuratum on a floodplain along the Kinu River (c. 800 m altitude), Tochigi Prefecture, Japan. The most frequent pollinator of C. purpuratum was the bumble bee Bombus diversus Smith. Other rare visitors to C. purpuratum have been described elsewhere (Ohashi and Yahara 1998). Cirsium flowers offer nectar and pollen, but most foragers do not actively accumulate pollen in their corbicular loads. Thus, we considered nectar to be the primary reward.

patterns of plant visits by individually marked bees

In September 1999, we established a 30 m × 60 m quadrat in our Cirsium population and numbered all flowering plants within the plot with plastic tape. We recorded their locations on a map and counted the number of flowering heads on each plant on 10 October 1999.

We caught foraging bees within the plot and anaesthetized them in CO2-filled plastic vials. We then glued numbered tags on to their thoraxes. Once released from CO2, the bees recovered within a few minutes. For several days before the observations, we marked all bumble bees that were caught in our plot, and marked any untagged bees found during the observations.

We measured the visitor density (number of individual bees which visited) and the individual visitation rates (number of visits per bee per unit of time) for each focal plant as described below. We selected the focal plants following Ohashi and Yahara (1998). First, we divided our plot into three subareas from east to west. Second, we divided the observation time within each day into three periods, separated by 15-min intervals (07.30–09.30; 09.45–11.45; 12.00–14.00). Third, we distinguished three categories of floral display size based on the number of flowering heads per plant (large, ≥ 7; middle-sized, 3–6; and small, ≤ 2 flowering heads). For data collection, we selected three plants that differed from each other in display size category and in subarea, and three observers directly monitored visits to the three plants (one observer each) for 2 h, identifying individual bees and counting their visits. The same procedure was repeated for each time period within a day. The three levels of display sizes were assigned to each subarea and time at random, except that each level was represented exactly once in each subarea and in each period of time within a day. We conducted the above procedure during the period of 5–7 October 1999, taking care not to select the same individual plant more than once. In total, we observed 27 individual plants.

Although we had marked as many bees as possible beforehand, we recorded 28 visits by unmarked bees on 14 plants. In particular, we observed multiple visits by unmarked bees on eight plants. For these plants, we estimated the visitor density from the total visitation rate (the total number of visits observed on a plant), using a regression equation obtained from the other 19 plants; visitor density per plant = 5·87 + 0·398 × total visitation rate per plant (R2 = 0·83, n = 19, P < 0·0001). If the estimated visitor density from the equation exceeded the sum of the visitor density of marked bees and the number of visits by unmarked bees, we assumed that all unidentified visits were made by different individuals to minimize the deviation from the regression line. On the other hand, if the estimate from the equation was smaller than the observed density of marked bees, we considered that visits by unmarked bees were made by a single individual.

We then performed multiple regression analyses to test whether variation among plants in patterns of total visits by individual bees can be explained by the two plant characteristics, floral display size (number of flowering heads on a plant) and local flower density (number of flowering heads on the neighbouring plants within a certain radius surrounding a plant). As the dependent variables, we first considered five types of visitor densities (number of individual visitors), because the total visitor density consisted of both regular visitors and occasional visitors, both of which showed large variations in their densities (Figs 1 and 2; see also Williams and Thomson 1998): (1) total visitor density (the number of individual bees that visited the plant per 2 h); (2) 1-v visitor density (the number of individual bees that visited the plant once per 2 h); (3) 2-v visitor density; (4) 3-v visitor density; and (5) regular (≥ 4-v) visitor density. Second, we considered three types of visitation rate (number of visits) as the other dependent variables: (6) total visitation rate; (7) the sum of individual visitation rates made by regular (≥ 4-v) visitors; and (8) the maximum individual visitation rate (the number of visits by the visitor that made the most frequent visits to the plant). We did not consider the sum of individual visitation rates for 1-v, 2-v, or 3-v visitors because these variables were highly correlated with the visitor densities (2)–(4). As the independent variables, we considered the floral display size of a plant and the local flower density for the radius considered, where 40 different radii (0·5, 1·0, ... , 20·0 m) were applied (i.e. we obtained 40 multiple regression equations for each dependent variable). Nonsignificant interaction term (floral display size × local flower density) was removed from each model and is not shown.

Figure 1.

A schema of frequency distributions of visits by individual bees to explain the eight variables analysed in multiple regressions. Numbers in parentheses correspond to the text. Abbreviations (1-v, 2-v and 3-v) indicate the individual visitation rates, e.g. 2-v visitors were the bees that visited a plant twice. ‘Regular visitor’ represents bees that visited a plant more than three times.

Figure 2.

Examples of the frequency distributions of visits to plants by marked bumble bees on 5 October 1999. Visits by unmarked bees are shown in the filled bars. The letters in each segment represent the number of flowering heads on the plant (‘H’), the number of individual bees (i.e. total visitor density, ‘D’), and the sum of individual visitation rates of all bees (i.e. total visitation rate, ‘R’). Each distribution is laid out according to the display size. Note that plants often shared the same individual bees as common visitors. Similar patterns were observed for the other 18 plants.

Interplant movements of individual bees

To determine the spatial extent of bumble bee movements, four observers followed individually marked bees and recorded their plant visit sequences. During the observations, each observer chose several individual bees to follow. When an observer found one of the chosen bees visiting our plot, he trailed the bee and recorded the sequence of individual plants the bee visited until he lost sight of the bee, the bee left the plot, or the bee stopped moving between flowering plants (probably due to a drop of its body temperature). These observations were conducted on 8 and 9 October in 1999 from 07.30 to 14.30.

For each bee, we counted the total number of visits to individual plants during the 2-day observation period and obtained the distribution of visits on plants. We quantified the shape of the distribution among bees, using Simpson's measure of evenness (E) calculated as follows.

image

where ni is the number of visits to plant i by a bee, N is the total number of plant visits made by the bee (equal to ∑ ni), and s is the total number of individual plants visited by the bee (Krebs 1999). E equals 1 when the bee's visits are distributed randomly among plants, and approaches 0 as the bee concentrates their visits to a certain set of plants within its foraging area.

We used a chi-square goodness of fit to determine whether the bees differed from one another in the shape of distribution of visits among plants. We cast the observed data into a table of frequencies, where the rows are the observed bees and the columns are the number of visits on ‘1st most frequently visited’, ‘2nd most frequently visited’, … , and ‘least frequently visited’ plants. Then we compared it with the table of expected frequencies when the rows are homogeneous. If the expected frequencies were fewer than five in more than 20% of the cells, we summed the adjacent columns to correct for bias in chi-square calculations (Zar 1984).

Next, we tested whether each bee makes frequent visits to plants that form the central part of its foraging area while making fewer visits to peripheral plants. For each plant that was visited by each bee, we calculated the distance from the plant to the centroid of the bee's foraging area. The positional coordinate of the centroid was calculated by averaging the coordinates of the plants within the foraging area of each bee. We then performed a Kendall's rank correlation analysis between the distance from a plant to the centroid of the foraging area of a bee and the number of visits to the plant by the bee, considering individual bees as the blocking variable (Korn 1984). To control the effects of floral display size and local flower density on visits, we also performed an ancova with the number of visits as dependent variable, floral display size, local flower density, the distance from the centroid as covariates, and individual bees as a factor (Grafen and Hails 2002).

Finally, we examined the effects of floral display size and local flower density on the centre location of a bee's foraging area. We performed an ancova with distance from the centroid as dependent variable, floral display size and local flower density as covariates, and individual bee as a factor.

Results

visitor densities and individual visitation rates to plants

Total visitor density varied greatly among plants: some plants received visits by numerous visitors, while the others attracted only a few visitors (Fig. 2, Table 1). In addition, these visitors differed in their individual visitation rates, namely, while a few visitors returned to a plant frequently, most visitors visited the plant only one to three times during the 2-h observation period (Fig. 2). Accordingly, we observed various shapes of frequency distribution of individual visitation rates, as shown in Fig. 2. In other words, there were large variations among plants in the densities of 1-v, 2-v, 3-v and regular visitors, the total visitation rate, the sum of individual visitation rates of regular visitors, and the maximum individual visitation rate (Table 1).

Table 1.  The minimum and maximum values, 25% and 75% quantiles, median, mean and SD of each variable used in multiple regression analysis (n = 27 for each row)
VariableMin.25% QuantileMedianMean75% QuantileMax.SD
Total visitor density (bees per plant) 515·027·033·2 48·5 7720·9
1-v visitor density (bees per plant) 011·521·020·2 26·0 4611·8
2-v visitor density (bees per plant) 0 2·0 4·0 6·3 11·0 19 5·5
3-v visitor density (bees per plant) 0 1·0 2·0 2·7  3·5  8 2·5
Regular visitor density (bees per plant) 0 7·0 2·0 4·0  7·0 14 3·9
Total visitation rate (visits per plant)1033·045·065·6103·516445·7
The sum of individual visitation rate of regular visitors (visits per plant) 0 5·020·024·7 37·5 7623·3
The maximum individual visitation rate (visits per plant) 0 4·0 7·07·7 11·0 15 3·9

effects of floral display size and local flower density

Both the floral display size of a plant and the local flower density had positive significant effects on the total visitor density (Table 2). By examining these effects in detail, we found that the visitor densities of all visitation classes significantly increased with floral display size, whereas only the 1-v and 2-v visitor densities increased with local flower density (Table 2). This effect of local flower density on occasional (1-v and 2-v) visitor densities was consistent with its negative significant effects on the sum of individual visitation rate of regular visitors and the maximum individual visitation rate (Table 2). Note that the total visitation rate to a plant did not decrease with the local flower density because the decrease in the sum of individual visitation rate of regular visitors was counterbalanced by the increased density of occasional visitors (Table 2). In contrast, the sum of the individual visitation rate of regular visitors and the maximum individual visitation rate to a plant significantly increased with floral display size (Table 2). As a result of these simultaneous increases in visitor densities and individual visitation rates, the total visitation rate per plant significantly increased with floral display size (Table 2).

Table 2.  Multiple regression for each dependent variable. Significant variables are in bold type (n = 27 for each)
Dependent variableSourceRegression coefficientSEtPStandard regression coefficientR2 (%)PRange of significant radius (m)Radius at which the highest R2 was found (m)
  1. Results were shown only for local flower density within a radius of 10 m, where we most frequently detected significant effects of local flower density on the dependent variables.

Total visitor densityFloral display size3·310·834·100·00040·60    
Local flower density0·100·042·290·03140·3448·40·00045·5–16·0 8·0
Constant−4·26        
1-v visitor densityFloral display size1·250·522·410·02420·39    
Local flower density0·070·032·820·00950·4636·90·00405·5–17·513·5
Constant−1·40        
2-v visitor densityFloral display size0·850·223·930·00060·57    
Local flower density0·030·012·640·01450·3948·80·00036·0–16·512·0
Constant−4·04        
3-v visitor densityFloral display size0·420·113·870·00070·62    
Local flower density0·000·010·550·58870·0939·00·0026  8·0
Constant−0·17        
Regular visitor densityFloral display size0·890·118·08< 0·00010·84    
Local flower density−0·010·011·760·0919−0·1873·8< 0·0001 10·0
Constant1·36        
Total visitation rateFloral display size9·221·685·48< 0·00010·74    
Local flower density0·030·080·400·69580·0555·8< 0·0001  6·5
Constant10·55        
The sum of individualFloral display size4·990·687·38< 0·00010·79    
visitation rate ofLocal flower density−0·110·033·130·0046−0·3472·5< 0·00018·5–20·010·0
regular visitorsConstant20·56        
The maximumFloral display size0·450·143·340·00270·43    
individual visitation rateLocal flower density−0·040·015·22< 0·0001−0·6661·1< 0·00015·0–20·010·0
Constant12·68        

Interplant movements of individual bees

During the 2-day observation period, we followed 24 individual bees. Figure 3 shows the patterns of interplant movements for eight individual bees. We eliminated from our consideration any bees whose data set contained fewer than 100 plant visits, because such data were likely to give unreliable estimates of the size of the bees’ foraging areas (Makino and Sakai 2004).

Figure 3.

Interplant flights and frequency distributions of visits across plants by eight bees whose observed visits were more than 100 each. The letters in the upper right corner of each segment represent the bee number, the number of observed plant visits, and the Simpson's measure of evenness (E). Data were recorded on 8 and 9 October 1999 and pooled data are shown. A circle indicates the location of a plant and its size expresses the floral display size. A line and its thickness indicate the flight path and the number of flight paths. A cross indicates the centroid of the foraging area of a bee. Graphs inside each segment are frequency distributions of visits among plants. The horizontal axis indicates individual plants, and the vertical axis indicates the number of visits to each plant with scale marks at five visits intervals.

Figure 3 indicates that bees varied in the size and location of their foraging areas. Moreover, bees differed significantly from one another in the shape of distribution of visits among plants (χ2 = 304·6, d.f. = 28, P < 0·0001); some bees allocated their visits evenly among plants (indicated by large values of Simpson's measure of evenness, E), while the others concentrated their visits on a small portion of plants (indicated by small values of E), within their foraging areas (Fig. 3). For example, bees B13 and Y19 had large foraging areas in which they visited plants at similar rates, while bee B59 had a smaller foraging area in which the bee concentrated its visits to a portion of plants.

We found a significant negative correlation between the distance from a plant to the centroid of a bee's foraging area and the number of visits to the plant by the bee (Kendall's blocked τ = −0·19, P < 0·0001, Fig. 4). This means that bees made frequent visits to plants located near the centre of their foraging areas while making infrequent visits to the other peripheral plants. This trend of centralized visitation was still significant when we incorporated floral display size and local flower density into analysis as the other factors (Table 3). On the other hand, we found no significant effect of floral display size on the distance from the plant to the centroid of a bee's foraging area (Table 4); floral display size did not increase the probability of a plant being adopted by a bee as the centre of its foraging area. We found that plants surrounded with higher local flower density tended to be at the centre of a bee's foraging area (P < 0·0001, Table 4).

Figure 4.

Relationship between the number of visits per plant by a bee and the distance from the plant to the centroid of the foraging area of a bee. Kendall's τ with a blocking variable is also shown. Symbols represent individual bees, and numbers in parentheses indicate sample size for each bee.

Table 3. ancova table for the number of visit per plant made by an individual bee
Sourced.f.SSMSFPβ*
  • *

    Regression coefficients in a general linear model.

Floral display size  1 501·0 501·029·5< 0·00010·335
Local flower density  11050·11050·161·8< 0·0001−0·032
Distance to centroid  1 831·8 831·848·9< 0·0001−0·384
Individual bees  71880·1 268·615·8< 0·0001 
Residuals3696271·3  17·0   
Table 4. ancova table for the distance from a plant to the centroid of the foraging area of a bee
Sourced.f.SSMSFPβ*
  • *

    Regression coefficients in a general linear model.

Floral display size  1   0·4   0·4  0·030·87100·009
Local flower density  12037·92037·9134·3< 0·0001−0·039
Individual bees  72567·3 366·8 24·2< 0·0001 
Residuals3705616·2  15·2   

Discussion

effects of floral display size and local flower density on the components of visitation rate per plant

As has been reported in Williams and Thomson (1998), we found that most of the observed plants attracted a few regular visitors and many occasional visitors. Furthermore, we found that the distribution of the visitor density and individual visitation rate varied greatly among plants. These variations were well explained by floral display size and local flower density (R2 = 36·9–73·8%, Table 2).

Large floral display size correlated with both increased visitor density and increased individual visitation rates of regular and occasional visitors, resulting in a higher total visitation rate per plant. The higher attractiveness of larger displays to occasional visitors might have resulted from their conspicuous appearance that makes them more likely to be detected by bees searching in unfamiliar areas (Spaethe, Tautz and Chittka 2001). If larger displays offer a higher rate of reward due to decreased travel time, this could also increase returns of regular individuals, as bumble bees can learn the associations between locations and rewards while foraging in familiar areas (Thomson 1988; Cartar 2004). Similar causes may also explain the positive effect of floral display size on the total visitation rate observed in many previous studies (Klinkhamer and de Jong 1990; Ohara and Higashi 1994; Robertson and Macnair 1995; Goulson et al. 1998; Ohashi and Yahara 1998, 2002; Vrieling et al. 1999; Makino and Sakai 2004; Mitchell et al. 2004; Grindeland et al. 2005; Miyake and Sakai 2005).

In contrast to floral display size, local flower density had a negative effect on the visitation rate of regular visitors, while it was positively correlated with the density of occasional visitors. This may suggest that plants in dense patches had more chances to be ‘sampled’ by bees foraging on neighbouring plants. Moreover, such occasional visitors would depress the reward level on a plant, which might discourage regular visitors from returning to the plant (Thomson et al. 1987; Thomson 1988; Carter 2004; Makino and Sakai 2005). Our finding that the individual visitation rate of regular visitors decreased with local flower density seems consistent with this interpretation.

patterns of spatial use by individual bees

Within their own foraging areas, individual bees visited particular sets of plants more frequently than the others (Fig. 3). This result means that a regular visitor on one plant could have been categorized as an occasional visitor on another plant in Fig. 2 and Table 2. Similar behavioural trends in bumble bees have been reported in previous studies (Heinrich 1976; Thomson et al. 1982, 1987, 1997; Comba 1999; Makino and Sakai 2004, 2005). Moreover, results obtained here suggest that a bee makes frequent visits to plants near the centre of its foraging area while it makes infrequent visits to peripheral plants. This effect on visits was significant independently of the effects of floral displays size and local flower density.

Although individual bees concentrated their visits to the central part of their own foraging areas, these central plants did not always have large floral displays. This is simply because the relation between distance to the centroid and floral display size of each plant inevitably reflect the spatial distribution of floral displays, or because the locations of a bee's foraging area is determined not only by the profitability of plants but also by competition with other foragers (Makino and Sakai 2004, 2005). On the other hand, plants with higher local flower density were on average closer to the centre of a bee's foraging area. This is probably because peripheral plants of bees’ foraging areas were often located near the edge of our Cirsium population where local flower density should be low (Fig. 3). Note, however, that this explanation does not contradict our finding that local flower density had a negative effect on the attraction of individual bees (Table 2), because we found the same trend for individual bees even when we held the distance from the centre constant (Table 3).

Conclusions

Williams and Thomson (1998) were the first to find by marking individual bumble bees that pollinators visiting a single plant contained both regular (frequent) and occasional (infrequent) visitors. Here we conducted similar observations for multiple plants in a population, and found that large floral displays increased both regular and occasional bumble bee visitors while high local flower density (or large number of neighbouring flowers) increased only occasional visitors. To our knowledge, this is the first time that plant characters have been demonstrated to differ in the attraction of these two types of individual visitors.

Another novel finding of this study is that individual bees returned to central plants more often than to peripheral plants within their own foraging areas. Irrespective of the location in its foraging area, moreover, each bee preferred to visit plants with larger floral displays as well as plants surrounded by fewer flowers. Although floral display size did not specify the location of a plant within a bee's foraging area, our data illustrate how a bee differentiate its visitation to plants, depending on their display sizes, local flower densities, and locations within its foraging area.

The results obtained here strongly support our basic idea that we need to dissect the collective data of ‘pollinator attraction’ into visits made by individual pollinators to fully understand the dynamics of pollen flow within a plant population. Our data suggest that producing more flowers differs from having more flowers in its neighbourhood in terms of its effects on pollen flow of the plant; although both traits will enhance mate diversity by increasing the number of occasional visitors, only large floral display will assure mating with particular plants by increasing returns of each individual. Further studies using methods employed here are needed to clarify how such visitation patterns are widespread among plant–pollinator systems, as well as how they can affect the evolution of plants mediated through their effects on pollen dynamics among plants (reviewed by Thomson and Chittka 2001).

Acknowledgements

We are grateful to H. Taneda, M.C. Kato, T. Kimura and T. Itagaki for their help in our fieldwork. We also thank J.E. Cresswell, A.W. Robertson and an anonymous reviewer for helpful comments on the manuscript.

Ancillary