Floral display size influences subsequent plant choice by bumble bees


†E-mail: hishii@ucalgary.ca


  • 1The effect of floral display size (number of open flowers per plant) on subsequent plant choice by bumble bees, Bombus hypocrita sapporensis (Cockerell), was investigated using artificial inflorescences with blue or yellow flowers representing different ‘species’.
  • 2Bees flew to an inflorescence of the same colour (constant flight) more often after visiting a large inflorescence. This response occurred when the nectar volume per flower was constant (larger inflorescence offered more nectar), and when nectar volume per inflorescence was constant (nectar volume per flower varied with display size). These results indicate that bees did not exhibit higher constancy after visiting larger inflorescences because they extracted more nectar than from smaller ones.
  • 3Bees spent more time during visits to large vs small inflorescences. Possibly information about the last-visited inflorescence fades from short-term memory during longer visits, which strengthens colour constancy.
  • 4Selection on floral display size is thought to balance the attraction benefits of large displays with the cost of geitonogamy. However, in habitats with several competing plant species, selection for colour constancy may provide a further reason for selection to favour large floral displays.


Pollinators often fly among plants of the same species (hereafter a ‘constant flight’) more often than is expected from plant frequency and distribution (‘flower constancy’: Heinrich 1976, 1979). Constancy may arise partly because of the innate preferences of pollinators, which can differ both within (Lunau et al. 1996) and among species (Heinrich 1976; Chittka et al. 2001). Constancy can also arise as an acquired, often learned preference (Gegear & Laverty 2001) that is shaped as pollinators compare experiences at different plants, and so may differ among individuals with different visitation history (Hill et al. 1997). In addition, pollinators may continue to choose the last-visited plant type because their limited memory of past events (usually from seconds to minutes, Menzel 1999, 2001) is replaced by information on the most recently visited plant type (short-term memory overstrike: Chittka et al. 1999; Ishii 2005). This notion is consistent with hypotheses that limitation of memory for motor patterns (Lewis 1986; Waser 1986; Dukas 1995; Chittka & Thomson 1997) and/or sensory stimuli (Dukas & Real 1993; Wilson & Stine 1996; Goulson 2000) causes pollinators to specialize temporarily on a single or few plant species.

Constant flights benefit plants by promoting conspecific pollination and reducing pollen loss (Waser 1978; Campbell 1985). Consequently, evolutionary divergence in attractive signals, such as flower colour, scent and shape, may evolve to promote constant flights (Jones 2001). For example, Chittka et al. (2001) proposed that species in the same habitat diverge in flower colour to promote constant flight, because bees tend to fly between conspecifics as their colour difference increases. However, as they pointed out, evidence of colour divergence in natural habitats due to the benefit of constant flight is limited (Gumbert et al. 1999). Divergence of other floral recognition cues in relation to constant flight has rarely been studied.

Other than floral signals, plant traits that affect a pollinator's experience on a plant may also promote constant flight. This effect is likely if short-term flower specialization caused by short-term memory overstrike significantly affects a pollinator's choices, because experience at a single plant would alter subsequent plant choice by a pollinator. Pollinators often spend more time probing deep or complex flowers than probing shallow or simple flowers (Harder 1983, 1986; Laverty 1994). Thus if overstrikes on short-term memory progress during long visits, deep or complex flowers may promote constant flight. Alternatively, greater reward per plant may promote constant flight if it reinforces memory.

Pollinators usually probe more flowers per plant as the number of open flowers per plant (floral display size) increases (reviewed by Ohashi & Yahara 2001), and both the duration of an inflorescence visit and nectar availability per plant probably increase with floral display size. Thus floral display size could also affect subsequent plant choice by a pollinator, although such effects have not been investigated previously. To date, two aspects of pollinator behaviour in response to floral display size have been studied that strongly influence pollination and plant mating. First, larger displays benefit plants by attracting more pollinators (reviewed by Ohashi & Yahara 2001), thereby increasing pollen receipt, export and potential mate diversity (Harder & Barrett 1996). Second, large displays can incur mating costs because they allow individual pollinators to visits several flowers on an inflorescence (reviewed by Ohashi & Yahara 2001), which can cause geitonogamous self-pollination, pollen discounting and stigma clogging by self-pollen (de Jong et al. 1993; Harder & Barrett 1995; Snow et al. 1996). Given such relations, selection on floral display size is thought to balance the attraction benefits of large displays with the cost of geitonogamy (de Jong et al. 1992; Harder & Barrett 1996). In habitats with several competing plant species, selection may additionally favour larger displays if it facilitates the fidelity of pollinators to a specific species, increasing pollen transfer among conspecifics.

This study tests whether bumble bees are more likely to fly to the same-coloured inflorescence after visiting large inflorescences than after visiting small ones, and whether this response, if true, depends on the amount of nectar extracted from an inflorescence.

Materials and methods

study design

The experiments were conducted during August 2004 in a 4 (width) × 5 (depth) × 2-m (height) screen cage. Responses to artificial inflorescences were studied in 12 worker bumble bees, Bombus hypocrita sapporensis (Cockerell), from two colonies (six individuals from each colony). The colonies were founded by queens caught during May 2004 in a natural deciduous forest in Sapporo, Hokkaido, northern Japan. Artificial inflorescences presented artificial flowers constructed from the closed end of 1·5-ml Eppendorf centrifuge tubes. Each flower was coloured blue or yellow with a marker pen. The diameter and depth of each flower were 10 and 15 mm. Seven inflorescence types were prepared with two to eight flowers arranged vertically to form a one-sided, spike-like inflorescence (Fig. 1a): inflorescences with two, four and eight blue flowers (B2, B4 and B8, respectively); two, four and eight yellow flowers (Y2, Y4 and Y8, respectively); and two blue and two yellow flowers (B2Y2). Prior to testing, each bee was trained on a patch of 20–30 B2Y2 inflorescences to preclude the establishment of any colour preference other than innate preference. Individual flowers contained 2 µl 30% sucrose solution. When bees visited a B2Y2 inflorescence, they usually probed all four flowers from bottom to top, and did not avoid either colour.

Figure 1.

Aspects of the study design, including the characteristics of (a) artificial inflorescences; (b) foraging arena; (c) example of an experimental booth. •, Blue (B2, B4, B8); ○, yellow (Y2, Y4, Y8); half-filled circles, mixed-colour inflorescences (B2Y2). The foraging arena included four experimental booths arranged around a buffer zone of 20–30 B2BY inflorescences. An experimental booth presented four inflorescences of different colours at positions A and B, and C and D.

During testing, 20–30 B2Y2 inflorescences were arranged randomly within an 80 × 160-cm area as a buffer zone, and four experimental booths were arranged around them (Fig. 1b). All sides of an experimental booth, except for the side facing the buffer zone, were closed with clear acryl glass to encourage a bee entering the booth to visit inflorescences in a specific sequence. Each booth contained four artificial inflorescences arranged in a Y configuration with 10 cm between adjacent inflorescences (Fig. 1c). Because of this arrangement, a bee entering a booth first visited inflorescence A at the base of the Y (96·6% of entries), followed by inflorescence B in the middle of the Y (93·3% of transitions from A), before choosing between inflorescences C and D at the tips of the Y (95·1% of transitions from B). All bees, except one test bee, were removed to eliminate interactions between them. Inflorescence A presented either four yellow or four blue flowers (Y4 or B4), each containing 2 µl 30% sucrose solution. Inflorescence B presented flowers of the alternate colour to that of inflorescence A (if inflorescence A was yellow, inflorescence B was blue, and vice versa). The characteristics of inflorescence B were varied to assess the relative importance of inflorescence size and nectar volume on a bee's subsequent choice (see below). One inflorescence at positions C and D presented four yellow flowers (Y4), whereas the other presented four blue flowers (B4), each of which contained 2 µl 30% sucrose solution. Thus a bee could visit either the same or a different inflorescence colour after visiting inflorescence B. At each booth I recorded a bee's choice among inflorescences C and D (constant vs non-constant flight) after visiting inflorescences A and B, and the number of flowers visited per inflorescence. So that none of the data analysed used visits to booths where inflorescences A and B were previously visited, if a bee did not visit inflorescences in the order A–B–C or A–B–D, I replaced inflorescences A and B in the booth and repeated the trial. Most deviations occurred during the foraging sequence that immediately preceded a focal bee's return to her nest. All inflorescences in the experimental booth were replaced after being visited.

I considered two different relations of display size and nectar availability per inflorescence. First, to examine the effect of an association between display size and an inflorescence's nectar availability on a bee's subsequent choice, the display size of inflorescence B was varied from two to eight flowers with a constant nectar volume per flower (2 µl 30% sucrose solution per flower). Therefore inflorescences with two, four and eight flowers presented 4, 8 and 16 µl sucrose solution, respectively. Because bees usually probed all flowers on an inflorescence (see Results), nectar intake varied with display size. Second, to determine whether display size influenced choice independently of nectar availability, I varied the display size of inflorescence B without changing the nectar volume per inflorescence (8 µl 30% sucrose solution per inflorescence). In this case, flowers contained 4, 2 or 1 µl sucrose solution on inflorescences with two, four and eight flowers, respectively. For both nectar situations, yellow and blue were used equally as the flower colours of inflorescences A, B, C and D. Eight replicate booth visits were conducted for each situation for each of the 12 test bees, resulting in 1152 trials (12 bees × two nectar settings × two colours × three display sizes × eight replicates). The four experimental booths were used randomly to preclude trapline foraging of bees (Thomson et al. 1997). The order in which each bee experienced the different treatment was also assigned randomly to exclude learning effects.

In addition to the above 1152 tests, the duration of bees’ visits to inflorescences was timed to relate it to display size. Inflorescences were arranged as in the above experiments and a bee's visit duration at inflorescence B was recorded. For these measurements I did not identify individual bees, but at least 15 bees from two colonies were used. Nectar-extraction rate (amount of sucrose solution extracted from an inflorescence per visit duration) was then estimated based on the assumption that bees ingest all sucrose solution from flowers that they visit.

statistical analysis

The influence of characteristics of inflorescence B (colour, display sizes, nectar settings) and their interactions on a bee's subsequent inflorescence choice was tested with a repeated-measures generalized linear model (Genmod procedure of sas ver. 8·2, 2001). This model considered binominal errors and logistic transformation of the dependent variable and used generalized estimation equations with exchangeable correlation matrix to account for repeated design of individual bees (Liang & Zeger 1986). A bee's choice after visiting inflorescence B (same or different colour) was treated as a dichotomous response variable.

One-factor anovas were used to test how an inflorescence's display size affected a bee's visit duration per inflorescence and nectar extraction rate. For these analyses, data for blue and yellow inflorescences were pooled because colour did not affect responses.


Bees flew to the same inflorescence type more often after visiting a larger inflorescence, at both types of nectar setting (Fig. 2; Table 1). Based on an overall comparison, display sizes had significant but weak effect on the percentage of subsequent constant flight (χ2 = 6·86, df = 2, P = 0·032; Table 1); however, a test of the more specific hypothesis that constant flight increased linearly with display size was more strongly significant (χ2 = 6·85, df = 1, P < 0·01). The colour of inflorescence B also affected a subsequent bee's choice, with constant flights occurring more often after a visit to a blue than a yellow inflorescence. The effect of display size on a bee's subsequent choice did not depend on nectar setting (display size × nectar setting interaction). Bees flew to the same-coloured inflorescence more often after visiting a larger inflorescence, regardless of whether or not nectar availability per inflorescence varied with display size.

Figure 2.

Least-square means and standard errors (error bars) of percentage of constant flights after visiting inflorescence B estimated by a repeated-measures generalized linear model. Back-transformation resulted in asymmetrical standard errors. Closed symbols, blue; open symbols, yellow; circles, constant nectar volume per flower; squares, constant nectar volume per inflorescence. Dotted horizontal line, 50% of constant flights.

Table 1.  Result of a repeated-measures generalized linear model showing the effect of display size, colour and nectar settings on the probability that a bee flew to the same inflorescence type after visiting inflorescence B
Source of variationdfχ2P
Display size of inflorescence B2  6·86<0·05
Colour of inflorescence B1  7·32<0·01
Nectar setting1<0·01>0·95
Display size × colour2  0·56>0·75
Display size × nectar setting2  0·45>0·75
Colour × nectar setting1  0·98>0·25
Display size × colour × nectar setting2  0·77>0·50

When bees visited an inflorescence, they typically probed all flowers (>89% in all cases). Accordingly, the amount of nectar extracted from an inflorescence increased almost proportionally with display size for inflorescences with a constant nectar volume per flower, whereas it was almost constant with a constant nectar volume per inflorescence (Fig. 3a).

Figure 3.

Mean (± SE) (a) estimated amount of nectar extracted (n = 192 inflorescences per mean); (b) visit duration at inflorescences (n = 40); (c) estimated nectar extraction rate (n = 40). •, Constant nectar volume per flower; □, constant nectar volume per inflorescence.

Visit duration on an inflorescence increased with display size for both a constant nectar volume per flower (one-factor anova: F2,117 = 549·21, P < 0·001; Fig. 3b) and per inflorescence (F2,117 = 122·61, P < 0·001). Consequently, the extraction rate of sucrose solution on an inflorescence declined with display size at a constant nectar volume per inflorescence (one-factor anova: F2,117= 104·45, P < 0·001; Fig. 3c) whereas it did not change significantly at a constant nectar volume per flower (F2,117 = 0·594, P = 0·554).


This study provides the first evidence that display size can affect flower constancy by pollinators. In particular, bees flew preferentially to the same inflorescence type after visiting a large display. Thus a trait of the single most recently visited plant (or inflorescence) affected subsequent plant choice.

This result supports the hypothesis that pollinators tend to fly to the same plant type as they last visited, because their short-term memory is overstruck by the information gained from the most recently visited plant type (Chittka et al. 1999; Ishii 2005). Surprisingly, this result was not affected by the amount of nectar extracted from large vs small inflorescences, or inflorescence B vs inflorescence A, as shown by the lack of interaction between display size and nectar setting on a bee's subsequent choice: the percentage of constant flights increased with display size although nectar extraction rate on an inflorescence declined with display size for a constant nectar volume per inflorescence.

Longer visit duration on large displays may be responsible for these results. Chittka et al. (1997) showed that a long-distance flight between plants reduces the probability of constant flight, possibly because information stored in short-term memory of the last-visited plant fades while travelling longer distances. In my experiment, information of the last-visited inflorescence might fade during longer visit duration and be easily replaced by information on the plant type being visited. Variation in visit duration among display sizes in this experiment (3·6–31·5 s) may have affected this overstrike process, because short-term memory decays within seconds to minutes (Menzel 2001). Alternatively, the larger vertical plane of larger displays may facilitate subsequent constant flight. A larger vertical plane might act as a stronger stimulus to bees, enhancing memory assimilation. Further experiments, considering both visit duration and optical stimuli, will be needed to relate floral display size and subsequent plant choice by bees.

Constant flight occurred more often after bees visited blue inflorescences than after they visited yellow ones. This colour effect may reflect an innate preference by bees for blue over yellow (Chittka et al. 2001; Gumbert 2000), or a stronger contrast of blue inflorescences against the background (Giurfa et al. 1996).

Individual plant traits that affect pollinator choice could influence floral evolution. Constant flights benefit plants by reducing improper pollen transfer between different species (Rathcke 1983). Accordingly, traits that influence subsequent constant flights benefit the plant itself. As noted above, selection on floral display size is thought to balance the attraction benefits of large displays with the cost of geitonogamy (de Jong et al. 1992; Harder & Barrett 1996). This experiment suggests that, in habitats with several competing plant species, selection might favour larger displays because it facilitates constant flights and benefits the plant. Other floral traits might also be able to promote constant flight. For example, pollinators often spend longer probing deep or complex flowers than shallow or simple flowers (Harder 1983, 1986; Laverty 1994). Thus if longer visit duration on a plant encourages subsequent constant flight, deep or complex flowers may also promote constant flight.

Further testing is needed of the relationship of constant flight to display size in natural situations. In this study, I excluded differences in floral traits other than colour, whereas in natural situations traits such as flower size, scent and shape also vary among species. As Gegear & Laverty (2001) indicated, pollinators become more selective if flowers differ in two or more traits. Accordingly, bees may respond more strongly to floral display size in natural conditions than was observed in this experiment. Alternatively, floral display size may have a limited effect on flower constancy. In natural habitats, plant species are not usually distributed uniformly as in this experiment, and sometimes form exclusive patches. Pollinators tend to visit the nearest plants in succession (Marden & Waddington 1981; Chittka et al. 1997). Thus the effect of spatial arrangement and the proximity of similar or different plants may overwhelm the effect of display size. Because several factors affect plant choices, including the spatial arrangement of plants, learning, innate preference and the overstrike effect on short-term memory (Ishii 2005), comprehensive analysis is needed to understand how floral display size affects plant choices by pollinators in natural situations.


I thank G. Kudo for his guidance in planning and fruitful discussions of this study, and L.D. Harder for his critical comments on the manuscript and assistance with statistical analysis. I especially acknowledge the support of Y. Hirabayashi throughout this study. I also thank S.F. Hasegawa and the members of Regional Ecosystems in Hokkaido University for building the flight cage. This study was supported by a fellowship of the Japan Society for the Promotion of Science for Young Scientists (no. 0124) and a Grant-in-Aid from the Ministry of Education, Culture, Sports, Science and Technology, Japan (no. 15370006).