Functional Blackwell Publishing, Ltd. Ecology 2005 Effects of floral display size and plant density on pollinator 19, 383–390 visitation rate in a natural population of Digitalis purpurea

J. M. GRINDELAND,* N. SLETVOLD†‡ and R. A. IMS§ Department of Biology, University of Oslo, PO Box 1066 Blindern, N-0316 Oslo, Norway

Summary 1. Pollinator visitation patterns in relation to variation in floral display size may be modified both quantitatively and qualitatively by local plant density. In this study four measures of pollinator response by Bombus spp. (plant visitation rate, bout length, proportion of flowers visited, flower visitation rate) were investigated under two or three different plant densities in two consecutive years in a natural population of Digitalis purpurea L. 2. Plant visitation rate increased with floral display size in both years, and was higher in dense patches compared with sparse ones in 1999. Bout lengths increased with dis- play size in 1999, and bouts were longer in sparse patches. However, the actual rate of increase with display size was independent of plant density for both response measures. 3. The proportion of flowers visited decreased with floral display size in both years, and in 1999 the decline was faster in high-density patches. As a result, the proportion visited was higher in dense patches for the smallest display sizes, and higher in sparse patches for larger display sizes. 4. Flower visitation rate decreased with floral display size in both years. This is incon- sistent with the idea that bees achieve an ideal free distribution across flowers. There was no significant effect of plant density. 5. These results demonstrate that local plant density variation may modify the functional relationship between floral display size and pollinator visitation rate, and potentially influence plant mating patterns. Key-words: Bombus, flower visitation rate, plant density effects, plant–pollinator interactions, pollinator foraging behaviour Functional Ecology (2005) 19, 383–390 doi: 10.1111/j.1365-2435.2005.00988.x

An increase in bout lengths on larger plants is pre- Introduction dicted in optimality models of insect foraging (Pyke Animal-pollinated plants are expected to evolve strat- 1979). From the plant’s point of view this may not be egies that maximize the efficiency of pollen transfer of any selective advantage, since it will result in geito- to and from the plant. One such strategy may be to nogamous pollination (de Jong, Waser & Klinkhamer develop large floral displays that attract more pollina- 1993) and possible inbreeding depression in self- tors. There is usually considerable within-population compatible species (Darwin 1876; Charlesworth & variation in the number of open flowers presented on Charlesworth 1987), or pollen discounting (Ritland a plant, and several studies have reported a positive 1991) and stigma clogging in self-incompatible ones relationship between floral display size and the fre- (Robertson & Macnair 1995). Empirical evidence sup- quency of pollinator visits at the individual plant level port that bout lengths usually increase with plant size, (i.e. plant visitation rate) (e.g. 15 out of 17 studies but this is often accompanied by a decrease in the pro- reviewed by Ohashi & Yahara 1999). portion of open flowers probed (references in Mitchell et al. 2004). As a result, per flower visitation rate has †Author to whom correspondence should be addressed. been found to increase (Klinkhamer, de Jong & de Bruyn E-mail: [email protected] 1989; Andersson 1991), decrease (Andersson 1988; *Present address: Sør-Trøndelag University College, Rotvoll Klinkhamer & de Jong 1990) or show a constant relation Allé 1, N-7004 Trondheim, Norway. with floral display size (e.g. Robertson & Macnair 1995; ‡Present address: Natural History Museum, University of Oslo, PO Box 1172 Blindern, N-0318 Oslo, Norway. Vaughton & Ramsey 1998; Ohashi & Yahara 2002). © 2005 British §Present address: Department of Biology, University of How pollinators respond to differences in display Ecological Society Tromsø, N-9037 Tromsø, Norway. size may be modified by ecological factors, especially

383

384 plant density (Charnov 1976). Optimal foraging theory Materials and methods J. M. Grindeland predicts plant visitation rate to be higher in dense et al. patches, but more flowers to be probed per plant visit   in isolated plants (Pyke 1979, 1982). Some studies have documented a positive relationship between plant Foxglove, Digitalis purpurea L., is a facultative biennial visitation rate and plant density (e.g. Klinkhamer & de herb that grows on naturally disturbed soil. The plant Jong 1990; Kunin 1997; Nielsen & Ims 2000), while produces one or a few inflorescences, usually 1–2 m in others have not (e.g. Aizen 1997; Bosch & Waser 2001), height, and normally with 20–100 flowers. The self- and there is at least one report of a decrease in visita- compatible flowers (Darwin 1876) produce nectar and tion rate with plant density (Stout, Allen & Goulson are mainly visited by bumble-bees (Best & Bierzychudek 1998). The longer visitation sequences predicted in 1982). The basal flowers mature first (acropetal flower- isolated plants have been confirmed by several workers ing). There is little spatial separation of the sexes within (Cibula & Zimmerman 1984; Klinkhamer & de Jong flowers, but temporal separation (protandry) is pro- 1990; Cresswell 1997; Bosch & Waser 1999). How these nounced. The combination of acropetal flowering and density effects combine into resulting visitation rates protandry results in inflorescences where the sexes are to individual flowers has received much less attention, spatially separated: the lowest flowers are female, usu- and no clear pattern has emerged from the results so ally followed by neuter flowers, male flowers and, at the far. Sih & Baltus (1987) found the relationship between top open, prefertile flowers. Nectar reward increases density (i.e. patch size) and average flower visitation during flower life (Percival & Morgan 1965) and rate in the perennial Nepeta cataria to depend on visitor consequently induces bumble-bees to visit the lowest type; for honey-bees and bumble-bees it was positive, flowers first. for solitary bees it was negative. No effect of density on flower visitation rate was found either in the facultative   biennial Echium vulgare (Klinkhamer & de Jong 1990), or in the perennial Cirsium purpuratum (Ohashi & Field work was conducted during the summers of 1998 Yahara 2002), while Kunin (1997) found a positive and 1999 in a population on established pasture land effect in the annual Brassica kaber. in Ulvik (Hordaland County) in SW Norway (60°34′ N, In addition to the effects density may have on mean 6°54′ E). An area of approximately 200 m by 100 m on pollinator response to variation in floral display size, the south-facing slopes towards the Hardangerfjord it may also change the actual shape of the response was used (10–50 m above sea level). The population curve. Ohashi & Yahara (2002) compared pollinator consisted of a mosaic of areas with varying plant behaviour in two populations of differing density of density, and focal plants were chosen within patches of Cirsium purpuratum, and showed that plant visitation uniform local density. rate in relation to display size increased at a higher rate In 1998 patches of three plant density categories were in the high-density stand. As far as we are aware this is used: four dense patches with plants less than 0·5 m the only study to address whether density affects the apart, four intermediate patches with plants approxi- functional relationship between floral display size and mately 1 m apart, and four sparse patches with plants pollinator visitation patterns. If pollinators alter their approximately 2 m apart. Ten neighbouring plants behaviour according to density variation at a smaller were haphazardly selected in each patch, yielding a spatial scale (i.e. within populations), this may have large total of 120 focal individuals. Approximately 900 impact on size-dependent pollen flow and plant mating flowering plants were present in the population. Insect patterns in populations with varying local plant density. activity was monitored in each patch for 10 min on Studies that address the effects of combined varia- each of 3 days: 11, 12 and 14 July 1998 (i.e. each patch and tion in floral display size and plant density at the level plant was observed for 30 min in total). The sequence of the individual flower are still few, and very little is of patch observations were changed at random known about how local density-variation may affect between days to avoid observing the same patch at the the shape of pollinator response curves. In this paper same time each day. Total observation time was 6 h. we report the results of an observational study on pol- The relatively low visitation frequency observed in linator visitation patterns in patches of varying spatial 1998 made us increase total observation time in the density in a natural population of Digitalis purpurea L. 1999 season, and the fact that fewer flowering plants (Scrophulariaceae). Two main questions are presently were present (c. 450) made it necessary to reduce the addressed: number of patch density levels. Two density categories were used: seven dense patches with plants less than 1. Is there a positive relationship between four meas- 0·5 m apart, and seven sparse patches with plants more ures of pollinator response (plant visitation rate, than 1 m apart. Neighbouring focal plants were bout length, proportion of flowers visited, flower haphazardly selected in each patch according to the © 2005 British Ecological Society, visitation rate) and floral display size? number available. In dense patches plant number Functional Ecology, 2. How are these relationships affected by plant varied from 6 to 14 with a total of 65, and in sparse 19, 383–390 density? patches from 8 to 11 with a total of 68. Insect visitation

385 was monitored for 15 min on 5 days between 29 June tively small integer counts) were analysed by log-linear Floral display size and 8 July 1999 (i.e. total observation time for each patch models, i.e. Poisson response distribution and log-link and pollinator and plant was 75 min). It was not possible to monitor function (Crawley 1993; Agresti 1996). Proportion of visitation rate every patch on each day due to weather conditions: On flowers visited was analysed by logit models, i.e. bino- 30 June, and 7 and 8 July, 14 patches were monitored, mial response distribution and logit link function on 29 June and 4 July, 10 patches, and on 3 July, 8 patches. (Agresti 1996). Flower visitation rate met assumptions Total observation time was 17·5 h. Observations were of normal distribution and constant variance and was timed to coincide with peak pollinator activity (between analysed by linear models (Sokal & Rohlf 1995). In 10·00 hours and 17·00 hours) in both study years. all models, plant density and observation date were Inter-patch distances varied from 5 to 20 m. included as fixed categorical factors and floral display For each plant visit the species identity of the visitor size as a continuous covariate. In all analyses backward and the number of flowers probed was recorded on selection and Type III tests (retain criterion P < 0·05) a mini-cassette recorder. It was not always possible were used to find the most appropriate model starting to determine which species of bumble-bee the visitor with all pairwise and three-way interactions. The belonged to (the morphology of Digitalis flowers makes goodness of fit of logit and log-linear models were it possible to see the insect only when it enters or leaves assessed by Pearson chi-square statistics and always a flower). A visit was recorded if the insect entered found appropriate. This validates and confirms the the corolla tube. Before each observation period the implicit assumption of our modelling strategy, that number of open flowers on the inflorescence (= floral random patch and plant effects were relatively un- display size) was noted for each plant in the patch. A important and consequently not needed in the models. flower was classified as ‘open’ when the corolla opening The linear models were performed with SPSS 11·0 was wider than 1 cm (no bumble-bees were ever seen (SPSS Inc. 2001), the logit and log-linear models with trying to enter a flower classified as ‘not open’). From the Insight module in SAS 8·02 (SAS Institute 2001). these data, pollinator visitation rate per hour to plants and to flowers, bout lengths (number of flowers probed Results during a single plant visit) and proportion of flowers visited (bout length divided by floral display size) were  determined. In the 1999 season several plants were infested by All visitors to D. purpurea observed were bumble-bees larvae of the species Eupithecia pulchellata Stephens (Bombus spp.) except one visit by a honey-bee (Apis (Lepidoptera: Geometridae). Oviposition takes place mellifera L.) in 1998. Bumble-bees were mainly forag- on immature buds and the presence of the larva inhibits ing for nectar, with a handful of observations of pollen flower opening, although the ‘bud’ grows to a size well collecting (i.e. turning upside down once inside the beyond the point where it would normally open (J. M. corolla tube). In total for both years four species of bumble- Grindeland, unpublished observation). Infested flowers bees were observed (Table 1): Bombus hortorum (L.), situated among normal open flowers were considered B. pascuorum (Scop.), B. lucorum (L.) and B. hypnorum to be part of the floral display and consequently treated (L.). The most frequent visitor by far was B. hortorum, as ‘open’. When movements within an inflorescence which accounted for more than half of all visits in both were considered only truly open flowers were included. years (Table 1). The other species varied in frequency both within year and between years. B. lucorum was a   relatively frequent visitor in 1998, but nearly absent the second year. For B. pascuorum the opposite was the Because of the slightly different methods (i.e. length of case. The fourth species, B. hypnorum, was observed in the observation periods and different categorization of 1998 only and then at fairly low frequency (Table 1). plant density), we analysed the data from the two years The frequencies are conservative estimates as not separately. Plant visitation rate and bout length (rela- all visitors were identified to species (Bombus sp. 12%

Table 1. Plant visits to Digitalis purpurea in 1998 and 1999 with individual visitor species

1998 1999 Total

Visitor species Number of visits % Number of visits % Number of visits %

Bombus hortorum 58 53·7 496 52·0 558 52·2 Bombus hypnorum 7 6·5 0 0·0 7 0·7 Bombus lucorum 26 24·1 36 3·8 63 5·9 Bombus pascuorum 3 2·8 358 37·6 363 34·0 © 2005 British Bombus sp. 13 12·0 63 6·6 76 7·1 Ecological Society, Apis mellifera 1 0·9 0 0·0 1 0·1 Functional Ecology, Total 108 953 1068 19, 383–390

386 in 1998 and 6·6% in 1999); however, most of these In 1998 only floral display size affected plant visitation J. M. Grindeland belonged to one of the two yellow-banded species (i.e. rate significantly (Table 2). The visitation rate increased et al. B. hortorum and B. lucorum). moderately with number of open flowers (Fig. 1a). In In 1998 the average floral display size was substantially 1999, floral display size, density, date and the plant higher than in 1999, 9·4 ± 4·1 (SD) as compared with 5·5 ± density–date interaction affected plant visitation rate 2·9 (SD). This difference was partly due to the presence significantly (Table 2). Plant visitation rate showed an of parasitized flowers in 1999, but also the fact that accelerating increase with floral display size, and it was inflorescences on average were 20 cm shorter in 1999. higher in dense patches than in sparse patches (Fig. 1b). The range in floral display size was also somewhat higher Mean visitation rate increased c. fourfold from the in 1998. Since in 1998 there were twice as many flowering first to the last observation date, and it also differed plants each on average with nearly twice as many open increasingly between the two density levels at the later flowers, a crude estimate gives c. four times as many observation dates. simultaneously open flowers in the population this year.      The average bout length per plant was slightly higher In 1998 a total of 108 plant visits were recorded, while in 1999 than in 1998 (2·1 ± 1·3 (SD) vs 1·8 ± 1·2 (SD)), the total for 1999 was 953 (Table 1). The mean fre- and the maximum bout length was five flowers longer. quency of visits varied more than threefold between On average the most frequent visitor, B. hortorum, had years: 1·8 visits per plant per hour in 1998 and 5·9 in the longest bout lengths in both years. In 1998 59% 1999. The range in number of visits was nearly twice as of all visits was to one flower only, in 1999 43%. In high in 1999; 0–32 visits per hour compared with 0–18 1998 neither floral display size, plant density nor date visits per hour in 1998. The proportion of plants left affected bout lengths significantly (Table 2, Fig. 2a). In unvisited was also lower. 1999 there was a highly significant main effect of

Table 2. Generalized linear model analyses of the effects of floral display size, density and date on plant visitation rate and bout length (Poisson errors, log link), and on the proportion of flowers visited (binomial errors, logit link). Only significant interactions are included

Plant visitation rate Bout length Proportion visited

Source of variation df χ2 P df χ2 P df χ2 P

1998 Display size 1 8·70 0·003 1 1·07 0·301 1 14·45 <0·001 Density 2 2·36 0·307 2 1·35 0·510 2 2·80 0·247 Date 2 5·12 0·078 2 2·16 0·340 2 2·09 0·351 1999 Display size 1 54·13 <0·001 1 91·89 <0·001 1 84·07 <0·001 Density 1 6·27 0·012 1 7·91 0·005 1 11·65 <0·001 Date 5 174·27 <0·001 5 18·15 0·003 5 6·97 0·223 Density * Date 5 11·85 0·037 5 5·75 0·331 Display * Density 1 17·36 <0·001 Display * Date 5 1·99 0·851 Display * Density * Date 5 12·99 0·023

© 2005 British Ecological Society, Functional Ecology, Fig. 1. The observed (mean ± SE) and predicted relationship between floral display size and plant visitation rate in (a) 1998 and 19, 383–390 (b) 1999.

387 Floral display size and pollinator visitation rate

Fig. 2. The observed (mean ± SE) and predicted relationship between floral display size and bout length (number of flowers probed in a single plant visit) in (a) 1998 and (b) 1999.

Fig. 3. The observed (mean ± SE) and predicted relationship between floral display size and proportion of flowers visited in (a) 1998 and (b) 1999.

floral display size, with bout lengths increasing with play size in both years (Fig. 4a,b). In 1998 there was a the number of open flowers (Fig. 2b). Bouts were also significant main effect of floral display size and no longer in sparse than in dense patches (Fig. 2b), and significant interactions (Table 3). The effect of density there was a significant effect of date (Table 2). No was marginally non-significant, with higher flower visita- interactions were significant. tion rates in low density patches. In 1999 the main effects of floral display size and observation date were     highly significant (Table 3), mean flower visitation rate almost doubled from the first to the last observation The average proportion of flowers visited was higher date. There was no effect of plant density, and neither in 1999 than in 1998 (0·42 ± 0·26 (SD) vs 0·20 ± 0·15 of the interactions were significant (Table 3). (SD)). It declined with increasing floral display size in both years (Fig. 3a,b). In 1998 only floral display size Discussion affected the proportion of flowers visited significantly (Table 2). In 1999 there was a significant three-way This study demonstrates that plant density-variation at interaction between floral display size, plant density the patch level may alter pollinator response to variation and date, and the main effects of display size and plant in floral display size. The proportion of flowers visited on density were highly significant (Table 2). The proportion equally sized plants of Digitalis purpurea declined at a visited declined at a faster rate in high density patches faster rate in dense patches compared with sparse ones. (Fig. 3b), and this relationship was strongest at the This density-dependent difference in foraging behaviour earlier observation dates. may affect plant mating patterns at a local scale, in particular in systems where geitonogamy is common.    © 2005 British    Ecological Society, Mean flower visitation rate was considerably higher in Functional Ecology, 1999 than in 1998: 3·1 visits per flower per hour in 1999 Plant visitation rates were substantially higher in the 19, 383–390 and 1·3 in 1998. It decreased with increasing floral dis- 1999 season compared with 1998, probably because of

388 J. M. Grindeland et al.

Fig. 4. The observed (mean ± SE) and predicted relationship between floral display size and flower visitation rate in (a) 1998 and (b) 1999.

Table 3.  of the effects of floral display size, density and patches at later dates. However, the bumble-bees’ actual date on flower visitation rate. No interactions were significant response curve to floral display size was independent

of density: dense patches attracted more visitors, but Source of variation SS df FP once in the patch the bumble-bee would discriminate 1998 between displays of different size in the same manner Display size 0·20 1 8·11 0·005 as in sparse patches. Density 0·12 2 2·49 0·089 Date 0·054 2 1·10 0·336 Residual 2·22 90     1999 Display size 4·33 1 20·03 <0·001 In 1998 bout lengths of foraging bumble-bees were Density 0·11 1 0·52 0·469 unaffected by the number of open flowers of D. purpurea, Date 9·65 5 8·93 <0·001 while in the 1999 season there was a marked increase Residual 94·27 436 in bout lengths with floral display size, and bouts were

also longer in sparse patches. We expected lower rates of increase in dense patches, as the difference in flight the combined effect of a higher number of pollinators costs between inter- and intraplant movements should foraging in the population and fewer plants available to be smaller here. Although such a trend was present in them. Bombus hortorum was probably the most impor- 1999, no significant effect of density was found. As insect tant pollinator both years, since visitation rates from activity increased during the 1999 season we expected this species was high and fairly constant throughout bouts to be shorter at later dates, but though date was the observation periods. Plant visitation rate increased significant in the analyses, bout length showed no sys- with floral display size in both years in our study tematic variation with date. population. The relationship was slightly accelerating, Bouts increased less than proportionally with the whereas most earlier studies have found this relation number of open flowers presented on a plant, and the to be decelerating (reviewed by Iwasa, de Jong & resulting proportion of flowers visited declined with Klinkhamer 1995; Ohashi & Yahara 2001). floral display size in both years. The proportion probed In 1998 plant visitation rate was independent of was expected to be lower in dense patches, but the density, while in 1999 pollinators visited plants in effect was absent in 1998 and proved to be dependent dense patches more frequently (on average 6·6 times on display size in 1999. At small display sizes the pro- per hour vs 5·3 times per hour in sparse patches). The portion in fact was higher in high-density patches, lack of any significant effect of density in 1998 could but as the relationship also declined at a faster rate, be due to weaker competition for nectar this year, as the resulting response curves crossed at a display bumble-bees should be less selective in their visitation size of approximately four flowers, which was a little pattern when reward levels are high. There was a gen- below mean display size (5·5 this year). This means eral increase in pollinator activity during the 1999 that on equally sized plants below this size, pollinators season, with much higher overall visitation rates at the would show greater tenacity in dense patches, whereas later observation dates, and this was accompanied by above this size tenacity would be highest in sparse increasing differences in visitation rate between the patches. A similar pattern was found by Ohashi & two patch density levels. This suggests that the stand- Yahara (2002) in their study of Cirsium purpuratum, © 2005 British Ecological Society, ing crop of nectar was increasingly depleted, and low but they compared bumble-bees foraging in either Functional Ecology, reward levels induced bumble-bees to keep interplant high or low-density areas. A bumble-bee foraging in a 19, 383–390 movements at low cost, and consequently choose dense population of D. purpurea will frequently move between

389 areas of differing local density, and our study indicates variability. The higher profitability of large plants Floral display size that pollinators may adjust their behaviour according attracts more visitors, but the occurrence of newly and pollinator to density-variation at this small, within-population emptied flowers within the display may cause the bee visitation rate scale. to leave after just a few flowers have been probed. Detailed measurements on the amount of nectar gained    at the flower level will be necessary to test whether this really is the case. Our results also suggest that plant The fact that pollinators tend to visit large displays density variation at a within-population scale may more often, but leave a greater proportion of the flowers influence pollinator foraging behaviour, and conse- behind has repeatedly been observed (Ohashi & Yahara quently size-dependent plant mating patterns (i.e. 2001). These countervailing factors have in most stud- selfing rates). Further work is required to determine if ies resulted in a constant relationship between display this is common in other systems, and the inclusion of size and flower visitation rate (e.g. Robertson & plant density in studies on pollination ecology is Macnair 1995; Goulson et al. 1998; Ohashi & Yahara recommended. 2002; Mitchell et al. 2004), whereas in D. purpurea flower visitation rate declined with increasing floral display size in both study years. Unless nectar production rate Acknowledgements (NPR) per flower is negatively correlated with display We thank G. Flatabø for assistance in locating a suita- size, this contradicts the idea that pollinators achieve ble field population, N. Øydvin for allowing us to work an ideal free distribution (IFD) across flowers (Fretwell on his land and L. O. Hansen for identifying the & Lucas 1970; Dreisig 1995). Since we lack data on Lepidoptera species. Helpful comments from C. Fox, NPR from this population, we cannot completely dis- K. Ohashi and an anonymous reviewer improved the miss the possibility that bumble-bees approximated an manuscript. This research was financially supported IFD. However, very few studies have found a negative by the Norwegian Research Council (grant no. relation between NPR per flower and display size 115536/410 to J.M.G. and grant no. 123845/410 to (references in Harder & Cruzan 1990), and in a com- N.S.), and by the Nansen Foundation of the Norwe- mon garden experiment on D. purpurea there was no gian Academy of Sciences (grant no. 31/98 to J.M.G.). significant correlation between plant size and NPR (J. M. Grindeland & N. Sletvold, unpublished observation). The decrease in flower visitation rate with display References size indicates that average reward levels were higher Agresti, A. (1996) An Introduction to Categorical Data Analysis. on large plants. In view of this it is puzzling that bout John Wiley & Sons, New York. lengths tended to be so short on large displays. One Aizen, M.A. (1997) Influence of local floral density and sex possible explanation could be that the unpredictable ratio on pollen receipt and seed output: empirical and variation in per flower reward level increases with dis- experimental results in dichogamous Alstroemeria aurea play size (Biernaskie & Cartar 2004). D. purpurea has (Alstroemeriaceae). Oecologia 111, 404–412. Andersson, S. (1988) Size-dependent pollination efficiency in a highly organized, vertical inflorescence, but as flower Anchusa officinalis (Boraginaceae): causes and consequences. number increases, more flowers will be situated at Oecologia 76, 125–130. approximately identical height. The likelihood that Andersson, S. (1991) Floral display and pollination success in some flowers will be missed during a foraging bout Achillea ptarmica (Asteraceae). Holarctic Biology 14, 186– is thus higher on large displays (J. M. Grindeland, 191. Best, L. & Bierzychudek, P. (1982) Pollinator foraging on fox- unpublished observation; Best & Bierzychudek 1982). glove (Digitalis purpurea): a test of a new model. Evolution If pollinators use reward levels gained at each flower to 36, 70–79. decide whether to leave a plant (Hodges 1985), a larger Biernaskie, J.M. & Cartar, R.V. (2004) Variation in rate of within-variability in reward level on large plants nectar production depends on floral display size: a pollinator could result in early departures (Pappers et al. 1999). manipulation hypothesis. Functional Ecology 18, 125–129. Bosch, M. & Waser, N.M. (1999) Effects of local density on We found no effects of density on flower visitation pollination and reproduction in Delphinium nuttallianum rate. The higher plant visitation rate observed in dense and Aconitum columbianum (Ranuculaceae). American Journal patches was counterbalanced by the higher decline of Botany 86, 871–879. in proportion of flowers visited. This is similar to the Bosch, M. & Waser, N.M. (2001) Experimental manipulation results of Ohashi & Yahara (2002), where average of plant density and its effect on pollination and reproduction of two confamilial montane herbs. Oecologia 126, 76–83. visitation rate per head in Cirsium purpuratum was Charlesworth, D. & Charlesworth, B. (1987) Inbreeding independent of plant density. depression and its evolutionary consequences. Annual Review of Ecology and Systematics 18, 237–268. Charnov, E.L. (1976) Optimal foraging, the marginal value Conclusion theorem. Theoretical Population Biology 9, 129–135. © 2005 British Cibula, D.A. & Zimmerman, M. (1984) The effect of plant Ecological Society, In view of the results of our study, we hypothesize that density on departure decisions: testing the marginal value Functional Ecology, bee behaviour on D. purpurea inflorescences is governed theorem using bumblebees and Delphinium nelsonii. Oikos 19, 383–390 by correlated changes in reward level and reward 43, 154–158. 390 Crawley, M.J. (1993) GLIM for Ecologists. Blackwell, Oxford. Ohashi, K. & Yahara, T. (1999) How long to stay on, and how J. M. Grindeland Cresswell, J.E. (1997) Spatial heterogeneity, pollinator often to visit a flowering plant? – a model for foraging strategy et al. behaviour and pollinator-mediated gene flow: bumblebee when floral displays vary in size. Oikos 86, 386–392. movements in variously aggregated rows of oil-seed rape. Ohashi, K. & Yahara, T. (2001) Behavioral responses of Oikos 78, 546–556. pollinators to variation in floral display size and their influ- Darwin, C. (1876) The Effects of Cross and Self Fertilisation ences on the evolution of floral traits. Cognitive Ecology of in the Vegetable Kingdom. Murray, London. Pollination: Animal Behaviour and Floral Evolution (eds Dreisig, H. (1995) Ideal free distributions of nectar foraging L. Chittka & J.D. Thomson), pp. 274–296. Cambridge bumblebees. Oikos 72, 161–172. University Press, New York. Fretwell, S.D. & Lucas, H.L. (1970) On territorial behavior Ohashi, K. & Yahara, T. (2002) Visit larger displays but probe and other factors influencing habitat distribution in birds. proportionally fewer flowers: counterintuitive behaviour Acta Biotheoretica 19, 16–36. of nectar-collecting bumble bees achieves an ideal free Goulson, D., Stout, J.C., Hawson, S.A. & Allen, J.A. (1998) distribution. Functional Ecology 16, 492–503. Floral display size in comfrey, Symphytum officinale L. Pappers, S.M., de Jong, T.J., Klinkhamer, P.G.L. & Meelis, E. (Boraginaceae): relationships with visitation by three (1999) Effects of nectar content on the number of bumble- bumblebee species and subsequent seed set. Oecologia 113, bee approaches and the length of visitation sequences in 502–508. Echium vulgare (Boraginaceae). Oikos 87, 580–586. Harder, L.D. & Cruzan, M.B. (1990) An evaluation of the Percival, M. & Morgan, P. (1965) Observations on the floral physiological and evolutionary influences of inflorescence biology of Digitalis species. New Phytologist 64, 1–22. size and flower depth on nectar production. Functional Pyke, G.H. (1979) Optimal foraging in bumblebees: rule of Ecology 4, 559–572. movement between flowers within inflorescences. Animal Hodges, C.M. (1985) Bumble bee foraging: energetic conse- Behaviour 27, 1167–1181. quences of using a threshold departure rule. Ecology 66, Pyke, G.H. (1982) Foraging in bumblebees: rule of departure 188–197. from an inflorescence. Canadian Journal of Zoology 60, Iwasa, Y., de Jong, T.J. & Klinkhamer, P.G.L. (1995) Why 417–428. pollinators visit only a fraction of the open flowers on Ritland, K. (1991) A genetic approach to measuring pollen a plant and its consequences for fitness curves of plants. discounting in natural plant populations. American Journal of Evolutionary Biology 8, 439–453. Naturalist 138, 1049–1057. de Jong, T.J., Waser, N.M. & Klinkhamer, P.G.L. (1993) Robertson, A.W. & Macnair, M.R. (1995) The effects of Geitonogamy: the neglected side of selfing. Trends in floral display size on pollinator service to individual flowers Ecology and Evolution 8, 321–325. of Myosotis and Mimulus. Oikos 72, 106–114. Klinkhamer, P.G.L. & de Jong, T.J. (1990) Effects of plant SAS Institute Inc. (2001) The SAS System for Windows, size, plant density and sex differential nectar reward on Release 8·0. SAS Institute Inc, Cary, NC. pollinator visitation in the protandrous Echium vulgare Sih, A. & Baltus, M.S. (1987) Patch size, pollinator behavior, (Boraginaceae). Oikos 57, 399–405. and pollinator limitation in catnip. Ecology 68, 1679–1690. Klinkhamer, P.G.L., de Jong, T.J. & de Bruyn, G.-J. (1989) Sokal, R.R. & Rohlf, F.J. (1995) Biometry, 3rd edn. W.H. Plant size and pollinator visitation in Cynoglossum officinale. Freeman, New York. Oikos 54, 201–204. SPSS Inc. (2001) SPSS for MS Windows, release 11·0, SPSS Kunin, W.E. (1997) Population size and density effects in Inc., Chicago, IL. pollination: pollinator foraging and plant reproductive Stout, J.C., Allen, J.A. & Goulson, D. (1998) The influence of success in experimental arrays of Brassica kaber. Journal of relative plant density and floral morphological complexity Ecology 85, 225–234. on the behaviour of bumblebees. Oecologia 117, 543–550. Mitchell, R.J., Karron, J.D., Holmquist, K.G. & Bell, J.M. Vaughton, G. & Ramsey, M. (1998) Floral display, pollinator (2004) The influence of Mimulus ringens floral display size visitation and reproductive success in the dioecious on pollinator visitation patterns. Functional Ecology 18, perennial herb Wurmbea dioica (Liliaceae). Oecologia 115, 116–124. 93–101. Nielsen, A. & Ims, R.A. (2000) Bumble bee pollination of the sticky catchfly in a fragmented agricultural landscape. Received 1 November 2004; revised 2 February 2005; accepted Ecoscience 7, 157–165. 17 February 2005

© 2005 British Ecological Society, Functional Ecology, 19, 383–390