- Email: [email protected]
Parasites and flower choice of bumblebees
Anim. Behav., 1998, 55, 819–825
Parasites and flower choice of bumblebees PAUL SCHMID-HEMPEL & HANS-PETER STAUFFER ETH Zu¨rich, Experimental Ecology (Received 6 May 1997; initial acceptance 9 June 1997; final acceptance 30 July 1997; MS. number: 5540)
Abstract. In a field experiment, we tested whether workers of bumblebees, Bombus pascuorum and B. humilis, parasitized by larvae of conopid flies, Physocephala rufipes and Sicus ferrugineus, differ in their flower choice from unparasitized ones. We collected workers at random in the field and immediately tested them in experimental arenas that offered the choice of a reference plant (red clover, Trifolium pratense) versus a test plant (from five species). The choices of 396 workers were analysed with logistic regression models (logit analysis). We performed all tests in the same field and at the same time where the workers were foraging naturally. On average, the parasitized bees were less likely to visit the reference plant. In addition, they were more likely to switch plant species even after the first visit in the 1998 The Association for the Study of Animal Behaviour experimental sequence.
The presence of parasites often affects the behaviour of hosts. In several cases, such behavioural modifications have been found to provide fitness advantages for the parasite. For example, survival of the host or transmission to a new host may be facilitated to ensure the development and propagation of the parasite (e.g. Holmes & Bethel 1972; Moore 1984; Milinski 1990; Keymer & Read 1991; Mu¨ller 1994). Fitness advantages for the host have also been identified, for example, impediments to parasite growth by altered thermal preferences (e.g. Boorstein & Ewald 1987; Mu¨ller & Schmid-Hempel 1993). However, parasite-induced behavioural changes not only involve conspicuous cases, such as the ‘topping behaviour’ of worker ants infected by liver flukes (where infected ants climb up into vegetation and become exposed to the next host of the parasite; Hohorst & Graefe 1961; Schneider & Hohorst 1971), but may actually often go unnoticed because of their rather subtle effects. Particularly where ecologically important behaviours are involved, parasitism may typically be a subtle factor that nevertheless has important effects for ecological relationships (e.g. Dobson 1988). In this study we investigate an ecologically important behaviour: flower choice by pollinators that is affected by parasitism. Correspondence: P. Schmid-Hempel, ETH Zu¨rich, Experimental Ecology, ETH-Zentrum NW, CH-8092 Zu¨rich, Switzerland (email: [email protected]). 0003–3472/98/040819+07 $25.00/0/ar970661
Bumblebees, Bombus spp., are annual, primitively eusocial species. While foraging for pollen and nectar, workers become exposed to predators and parasites, for example, infectious protozoa that can be contracted on flowers (Durrer & Schmid-Hempel 1994) or conopid flies (Conopidae, Diptera: Schmid-Hempel et al. 1990). Conopids insert an egg inside the abdomen of workers, but also in young queens or males. They develop to pupation within 10–12 days, at which time the host is killed (Schmid-Hempel & SchmidHempel 1996). The pupa stays inside the host and the next generation of flies emerges the following year. Parasitism by conopid flies is very common in natural populations, where larvae and eggs are present in 30–70% of all workers (Schmid-Hempel et al. 1990). Bumblebees are important pollinators in many temperate or cool areas. Their ecology, physiology or pollination strategies have therefore been studied intensively (Inouye 1977; Pyke 1978; Ranta & Tiainen 1982; Pekkarinen 1984; Ellington et al. 1990). Bumblebees have often been cited as a prime example of species co-existence that is driven by passive competition over limited resources (Pyke 1982; Begon et al. 1990). At the same time, recent work has demonstrated substantial degrees of parasitism in Bombus spp. which, moreover, is associated with important ecological characteristics of different species and bee assemblages (Schmid-Hempel & Durrer 1991;
1998 The Association for the Study of Animal Behaviour
819
820
Animal Behaviour, 55, 4
Durrer & Schmid-Hempel 1995; S. Durrer & P. Schmid-Hempel, unpublished data). In addition, preliminary field observations have suggested that flower choice by worker bees is not independent of parasitism (Schmid-Hempel & Schmid-Hempel 1990). Here, we have used field experiments to investigate this phenomenon in more detail. The study was carried out in the flowering meadows commonly visited by bumblebees. We asked to what extent the flower choice of workers of the bumblebees B. pascuorum Scop., an abundant and widespread species, and B. humilis Illiger, a rare and less widely found species in the region, is affected by the presence of endoparasitic larvae of conopid flies. In particular, we analysed whether parasitism affects choice differently when different flowers are offered as an alternative, and to what extent flower constancy, that is, the tendency to choose the same flowering plant species in successive visits, is affected by parasitism. Offering different flowers as alternatives can also generate a preference ‘profile’ of the standard plant against a variety of other available plant species and reveal whether this is affected by parasitism.
MATERIALS AND METHODS Between June and September 1993, we captured foraging workers of B. pascuorum and B. humilis at random along several transects in a dry meadow (mesobrometum) near Blauen, northwestern Switzerland. The majority of species can be readily distinguished in the field, but neither age nor parasitism of bumblebee workers can be detected before dissection in the laboratory. Bees were assigned at random to one of the treatments and were therefore randomized with respect to age, individual history or experience. In particular, with the post hoc examination of our sample, we found no relationship between time of day and the percentage of parasitized bees among those caught and tested. Similarly, the observer was ignorant with respect to parasitism of the bees. After their capture, the bees were deprived of food in a cage for 2 h to standardize hunger levels. They were then immediately transferred to the experiment where each (individually marked) worker was allocated at random to one of eight available test arenas. We used each bee only once. Each test arena was a flight cage (4040
75 cm), covered with fine mesh, that contained flowers of four plants near the four corners of its base. In two opposite corners, the flowers were from the ‘reference plant’, red clover, Trifolium pratense L. Red clover was by far the most abundant plant species in the area and clearly a favourite food plant of the bumblebees. The other two opposite corners were stocked with flowers of the same ‘test plant’ species: the alternative offered. The test plant species were kept the same for each experimental run and cage, but varied systematically over cages and successive runs. With this procedure, we tested the preference of workers, relative to the reference plant, for flowers of Betonica officinalis L., Centaurea jacea L., Prunella grandiflora L., Prunella vulgaris L. and Trifolium repens L. These five plant species were also among the most common ones in the study area. This procedure of pair-wise comparisons with the reference plant generated a preference ‘profile’ relative to the most preferred plant. We restricted the experiment to one such profile only, since it was impractical to test all possible combinations. We picked the flowering plants fresh in the field and assigned the individual plants randomly to the test cages according to schedule. We put all plants in a small tube with water approximately 1 h before the tests. Owing to differences in availability, not all test plant species could be tested with the same number of replicates. We placed the cages containing the bees and the five test arenas in the same meadow from where the bees came. Depending on the weather and bee activity, 5–20 bees could be tested per day. Because the sequence of testing plant species was random with respect to time of day, experimental day or parasitism (as revealed by post hoc dissections), any effect of the depletion of nectar in the surrounding field should have affected all tests in the same way and was thus randomized within the framework of our experiment. In an experimental run, we released the worker into the test arena and subsequently observed it for 1 h. We recorded all flower visits. Bees that made fewer than four visits before the time limit ran out were excluded from the analysis. After each experimental run, we carefully cleaned each cage and replaced the flowers. The bees were frozen for later inspection in the laboratory, and the plants were dry-pressed. The presence or absence of an endoparasite was checked under the
Schmid-Hempel & Stauffer: Parasites and flower choice microscope with the scorer unaware of treatments. We focused on eggs and larvae of conopids. Brood of two widespread and common species, Physocephala rufipes F. and Sicus ferrugineus L. (Schmid-Hempel & Schmid-Hempel 1996), were present in our samples. As no difference could be found between these two groups, we pooled the data for the final analysis presented here. We also noted the occurrence of other parasites, such as Crithidia bombi (Trypanosomatidae; Lipa & Triggiani 1980). However, in our samples the presence of other parasites was not related to parasitism by conopids nor to flower choice. Other parasites were therefore not analysed further. We analysed the data with SAS, procedure CATMOD, using logistic regression for categorical variables. In particular, we analysed the choices of test workers (i.e. test plant or reference plant) with respect to the presence or absence of conopid parasitism, bee species, B. pascuorum, or humilis, and the test plant species used. For the analysis of flower choice, we took only the first visit into account. For the analysis of flower constancies, we coded, with the variable ‘switch’, according to whether the same or a different flower species was visited in the next visit. All bees made four successive visits, so we used the response variables ‘switch1’ and ‘switch2’ to quantify the constancy of the bees on the second (relative to the first) and third (relative to the second) visit, respectively. We could not evaluate the constancy between the third and fourth visit, since only three out of eight possible combinations were observed, and therefore linear dependencies occurred in the analysis. Since the successive visits were made by the same individual worker, we analysed switch1 and switch2 as repeated measures. For this reason, we used the variable ‘Seq’ which was generated by the model according to the two response functions in each combination, that is, the odds ratio of Seq was calculated as the ratio of the odds ratios of switch1 and switch2 (with df=1). The odds ratio of Seq thus indicates how much more likely a bee switches on the second, relative to the first, visit compared with the third, relative to the second, visit. Throughout, odds ratios of 1 imply no effect of the tested factor. We chose the best statistical model by comparing all possible models (flower choice), or a number of plausible models (flower constancy), starting from the saturated model (with all factors and interactions present), and
821
then removing all main effects and interaction terms that produced non-significant deviations. SAS estimates the regression coefficients of such logit models with maximum-likelihood for the flower choice analysis and weighted least squares for the analysis of flower constancies (Anderson et al. 1980; SAS Institute 1990; Trexler & Travis 1993). All reported significance levels are two tailed.
RESULTS We tested 540 bees, of which 247 B. pascuorum and 149 B. humilis made at least four choices in the test arenas and were thus included in the final analysis. A total of 115 workers of B. pascuorum (46.6%) and 81 of B. humilis (54.54%) were parasitized. Flower Choice Table I summarizes the best model for the analysis of flower choice, that is, the first visit made to a reference or test plant. None of the interaction terms nor the main effect ‘bee species’ (B. pascuorum versus B. humilis) generated a significant deviation. The best model of Table I therefore contains only the (additive) main effects of ‘parasitism’ (yes/no) and ‘plant species’ (the five test plants). It was 1.54 times less likely that a parasitized bee was visiting the reference plant, T. pratense, than any test plant (Table I). Since no interactions occurred, this effect was independent of the test plant species offered (Fig. 1). On the other hand, the kind of test plant that was offered as an alternative to the reference plant affected the bees’ preferences, independent of parasitism. For example, B. officinalis, C. jacea or P. vulgaris had almost no effect on choice, but T. repens was clearly avoided and P. grandiflora preferred over the reference plant (Table I; see 50% preference in Fig. 1). Flower Constancy The response variable, that is, to visit the same plant species or to switch to the alternative test plant offered, was analysed with a logit model in relation to the same factors as before, that is, parasitism, test plant species, bee species and their
Animal Behaviour, 55, 4
822
Table I. Logit model for the effects of Parasite and Plant species on first flower choice Factor*
Coefficient†
Odds ratio‡
95% C.I.
P for factor level§
Intercept Parasite Plant species** B. officinalis C. jacae P. grandiflora P. vulgaris T. repens
0.062 0.429
0.121 0.216
1.064 1.535
0.742–1.191 1.006–2.343
0.048
0.015 0.481 0.925 0.001 1.392
0.184 0.244 0.203 0.253 0.290
1.015 0.618 0.397 0.999 4.022
0.708–1.456 0.383–0.997 0.267–0.590 0.608–1.640 2.277–7.105
0.049 <0.0001 <0.0001
ones with respect to switching (odds ratio=1.06, ). There was, however, a significant difference in that parasitized bees were more likely to switch between the first and second visits but less likely between the second and third visits, compared with their unparasitized counterparts (Fig. 2; see interaction term Parasite*Seq in Table II, odds ratio=0.745). None of the other factors or interactions produced a significant effect.
100
75
50
25
ffi B. o
200
196
M ea n
25
26
28
21
55
60
29
29
al is C. P. ja c gr an ea di flo ra T. vu lg ar is T. re pe ns
0
63
60
DISCUSSION
ci n
Observed preference for test plant (% C.I.)
Parasite: the presence of conopid eggs/larvae; Plant species: the species offered as an alternative to the reference plant. *A total of 396 workers analysed. †Regression coefficient () in the logit model. ‡Odds ratio >1 for Parasite indicates preference of parasitized bees for test plant. For Plant species it indicates preference for reference plant. §Wald’s test (=P>0.1). **Significance for effect over all plant species: P<0.0001.
Test plant Figure 1. Observed frequencies of visits (with 95% C.I. for observed percentages) to the test plant when offered as an alternative to the reference plant, T. pratense, and in relation to the presence () or absence ( ) of endoparasitic eggs or larvae of conopid flies. Sample sizes (number of workers tested) are given within the bars. For statistics of the logit model, see Table I. ‘Mean’ displays the average effect of parasitism, independent of test plant species.
interactions. Table II shows the best model. According to this analysis, parasitized bees behaved in the same manner as non-parasitized
Flower choice by pollinators, and by bumblebees in particular, is affected by a variety of factors. These include, for example, the availability of flowering plants (Heinrich et al. 1977; Heinrich 1979a, b; Inouye 1980), experience (Heinrich 1976, 1979b; Laverty 1994), nutritional status of the colony (Cartar & Dill 1990; Cartar 1991), weather conditions (Heinrich 1979a) or the presence of competitors (Inouye 1978; Pyke 1982). We now show that parasitism is also associated with differences in flower choice and with a difference in flower constancy. At least for the sample tested here, parasitized bees on average shifted away from the most common food plant in the area (i.e. our reference plant, T. pratense), independently of the alternative (‘test’) plant species offered (Table I, Fig. 1). Parasitized bees were also less constant in their choices. They tended to switch plant species on the first occasion, that is, between the
Schmid-Hempel & Stauffer: Parasites and flower choice
823
Table II. Logit model for the effects of Parasite, Plant species and Seq on flower constancy Factor*
Odds ratio‡
95% C.I.
P for factor level§
0.588 0.061 0.082 0.294
0.088 0.162 0.140 0.140
1.800 1.063 0.921 0.745
1.515–2.139 0.774–1.461 0.700–1.213 0.566–0.981
0.036
0.167 0.299 0.183 0.362 1.005
0.145 0.178 0.145 0.189 0.213
1.175 1.348 1.201 1.436 0.366
0.885–1.561 0.952–1.909 0.903–1.596 0.992–2.709 0.241–0.556
0.093 0.055 <0.0001
Coefficient†
Intercept Parasite Seq Parasite*Seq Plant species** B. officinalis C. jacae P. grandiflora P. vulgaris T. repens
80
70
50
Second
200
196
200
60
196
Proportion of bees switching plant species (% C.I.)
Parasite: the presence of conopid eggs/larvae; Plant species: the species offered as an alternative to the reference plant; Seq: the sequence of plant visited. *A total of 396 workers analysed. †Regression coefficient () in the logit model. ‡Odds ratio for Seq >1 indicate that the bee is more likely to switch between first and second visit than between second and third visit (see Methods). §Wald’s test (=P>0.1). **Significance for effect over all plant species: P<0.0001.
Third Visit
Figure 2. Observed percentage of bees (95% C.I.) switching plant species on the second or third visit, respectively, compared to the previous visit. : Parasitized;
: non-parasitized. Sample sizes (number of workers tested) are given within the bars.
first and second visit (Table II). An earlier, observational study (Schmid-Hempel & SchmidHempel 1990) similarly showed a systematic bias in flower visit rates in that parasitized workers of
B. pascuorum also shifted away from the more common of two species of flowering plants available in the area (P. grandiflora was more common than B. officinalis in this earlier study). We used randomly selected workers from the field that had become naturally infested by conopid flies and tested them immediately in the field. Hence, we can assume that they behaved in the natural context against their background experience. Our experimental protocol has the advantage of testing an actual field situation but poses the problem of causation, since naturally occurring parasitism was used. Experimental infections would have been preferred but are impractical for the necessarily large sample sizes. However, several lines of reasoning suggest that conopid parasitism is indeed responsible for the observed differences. For example, other known behavioural changes relate to parasitism rather than, for example, worker age (Schmid-Hempel & Schmid-Hempel 1991; Mu¨ller & Schmid-Hempel 1993). From many years of field studies, we find that conopid attacks are not associated with particular flowering plant species. Hence individual plant preferences of workers (e.g. Heinrich 1976) do not affect their risk of parasitism. Rather, our observations show that conopid females wait near the ground in flower fields and attack bees flying by. Together with the typical fine-grained mosaic of plant species, relative to the activity range of
824
Animal Behaviour, 55, 4
conopids and bees in the field, this behaviour makes it very unlikely that workers are only at risk at certain plant species but not at others. Conopids do, however, generally prefer larger hosts (Mu¨ller et al. 1996). No such difference in body size could be found in the sample used in the present experiments, nor did body size relate to flower choices. We have not investigated what underlying mechanisms cause the observed behavioural differences. Previous studies have shown that workers parasitized by conopids do not maintain thoracic temperature to the same degree (Heinrich & Heinrich 1983) and have reduced nectarcarrying capacities (Schmid-Hempel & SchmidHempel 1991). Given the large physical dimensions of conopid larvae inside their host’s abdomen (Schmid-Hempel & Schmid-Hempel 1996), it is likely that parasitized bees are simply physiologically stressed, for example, in a permanent state of nutritional deficit. This may cause the behavioural changes. However, it remains unexplained why certain flowers rather than others become less preferred or to what extent this might reflect an advantageous behavioural decision. On the other hand, the observed behavioural changes have potentially far-reaching effects at different levels. For example, the present and earlier (Schmid-Hempel & Schmid-Hempel 1990) results suggest that as parasite frequency increases in a given bee assemblage, certain plant species are expected to be visited more often, for example the less common species such as Betonica in the earlier study, or Prunella spp. and T. repens in this study (Fig. 1). Besides the straightforward effects on the pollination of these plants, the interactions among competing bee species in the local assemblage should also be affected. In addition, the transmission of other parasites will be different when a major pathway, that is, via the flowers (Durrer & Schmid-Hempel 1994), is modified. Such secondary consequences in communities of interacting species are not well understood, but in cases such as the present one parasite-induced behavioural changes may be at the root of the process.
ACKNOWLEDGMENTS We thank the community of Blauen for permission to work on the site. S. Cameron com-
mented on the manuscript. H.-U. Reyer (Zu¨rich University) provided logistic support for H.-P. S. M. Ma¨chler (ETH) helped to analyse the data. The study was supported by grants of the Swiss National Science Foundation (no. 3100-049040.95 to P.S.-H.).
REFERENCES Anderson, S., Auquier, A., Hauk, W. H., Oakes, D., Vandaele, W. & Weisberg, H. I. 1980. Statistical Methods for Comparative Studies. New York: John Wiley. Begon, M., Harper, J. L. & Townsend, C. R. 1990. Ecology: Individuals, Populations and Communities, 2nd edn. Oxford: Blackwell. Boorstein, S. M. & Ewald, P. W. 1987. Costs and benefits of behavioral fever in Melanoplus sanguinipes infected by Nosema acridophagus. Physiol. Zool., 60, 586–595. Cartar, R. V. 1991. Colony energy requirements affect response to predation risk in foraging bumble bees. Ethology, 87, 90–96. Cartar, R. V. & Dill, L. M. 1990. Why are bumble bees risk-sensitive foragers?. Behav. Ecol. Sociobiol., 26, 121–127. Dobson, A. P. 1988. The population dynamics of parasite-induced changes in host behavior. Q. Rev. Biol., 63, 139–165. Durrer, S. & Schmid-Hempel, P. 1994. Shared use of flowers leads to horizontal pathogen transmission. Proc. R. Soc. Lond. Ser. B, 258, 299–302. Durrer, S. & Schmid-Hempel, P. 1995. Parasites and the regional distribution of bumble bee species. Ecography, 18, 114–122. Ellington, C. P., Machin, K. E. & Casey, T. M. 1990. Oxygen consumption of bumblebees in forward flight. Nature, Lond., 347, 472–473. Heinrich, B. 1976. The foraging specializations of individual bumblebees. Ecol. Monogr., 46, 105–128. Heinrich, B. 1979a. Bumblebee Economics. Cambridge, Massachusetts: Harvard University Press. Heinrich, B. 1979b. ‘Majoring’ and ‘minoring’ by foraging bumblebees, Bombus vagans: an experimental analysis. Ecology, 60, 245–255. Heinrich, B. & Heinrich, M. J. E. 1983. Heterothermia in foraging workers and drones of the bumblebee Bombus terricola. Physiol. Zool., 56, 563–567. Heinrich, B., Mudge, P. R. & Deringis, P. G. 1977. Laboratory analysis of flower constancy in foraging bumblebees: Bombus ternarius and B. terricola. Behav. Ecol. Sociobiol., 2, 247–265. Hohorst, W. & Graefe, G. 1961. Ameisen: obligate Zwischenwirte des Lanzettegels (Dicrocoelium dendriticum). Naturwissenschaften, 48, 229–230. Holmes, J. C. & Bethel, W. M. 1972. Modification of intermediate host behavior by parasites. In: Behavioral Aspects of Parasitic Transmission (Ed. by
Schmid-Hempel & Stauffer: Parasites and flower choice E. U. Canning & C. A. Wright), pp. 123–149. New York: Academic Press. Inouye, D. W. 1977. Species structure of bumblebee communities in North America and Europe. In: The Role of Arthropods in Forest Ecosystems (Ed. by W. J. Mattson), pp. 35–40. New York: Springer. Inouye, D. W. 1978. Resource partitioning in bumblebees: experimental studies of foraging behavior. Ecology, 59, 672–678. Inouye, D. W. 1980. The effect of proboscis and corolla tube lengths on patterns and rates of flower visitation by bumblebees. Oecologia (Berl.), 45, 197–201. Keymer, A. E. & Read, A. F. 1991. Behavioural ecology: the impact of parasitism. In: Parasite–Host Associations: Coexistence or Conflict? (Ed. by C. A. Toft, A. Aeschlimann & L. Bolis), pp. 37–61. Oxford: Oxford University Press. Laverty, T. M. 1994. Bumblebee learning and flower morphology. Anim. Behav., 47, 531–545. Lipa, J. J. & Triggiani, O. 1980. Crithidia bombi sp.n. a flagellated parasite of a bumble-bee Bombus terrestris L. (Hymenoptera, Apidae). Acta Protozool., 27, 287– 290. Milinski, M. 1990. Parasites and host decision-making. In: Parasitism and Host Behaviour (Ed. by C. J. Barnard & J. M. Behnke), pp. 95–116. London: Taylor & Francis. Moore, J. 1984. Altered behavioral responses in intermediate hosts: an acantocephalan parasite strategy. Am. Nat., 123, 572–577. Mu¨ller, C. B. 1994. Parasite-induced digging behaviour in bumblebee workers. Anim. Behav., 48, 961–966. Mu¨ller, C. B. & Schmid-Hempel, P. 1993. Exploitation of cold temperature as defence against parasitoids in bumblebees. Nature, Lond., 363, 65–67. Mu¨ller, C. B., Blackburn, T. M. & Schmid-Hempel, P. 1996. Field evidence that host selection by conopid parasitoids is related to host body size. Insectes soc., 43, 227–233.
825
Pekkarinen, A. 1984. Resource partitioning and coexistence in bumblebees (Hymenoptera, Bombinae). Annls Entomol. Fenn., 50, 97–107. Pyke, G. H. 1978. Optimal foraging in bumblebees and coevolution with their plants. Oecologia (Berl.), 36, 281–293. Pyke, G. H. 1982. Local geographic distributions of bumble bees near Crested Butte, Colorado: competition and community structure. Ecology, 63, 555–573. Ranta, E. & Tiainen, M. 1982. Structure in seven bumblebee communities in eastern Finland in relation to resource availability. Holarct. Ecol., 5, 48–54. SAS Institute 1990. SAS User’s Guide: Statistics, Version 6. Cary, North Carolina: SAS Institute. Schmid-Hempel, P. & Durrer, S. 1991. Parasites, floral resources and reproduction in natural populations of bumblebees. Oikos, 62, 342–350. Schmid-Hempel, P. & Schmid-Hempel, R. 1990. Endoparasitic larvae of conopid flies alter pollination behaviour of bumblebees. Naturwissenschaften, 27, 50–52. Schmid-Hempel, P., Mu¨ller, C., Schmid-Hempel, R. & Shykoff, J. A. 1990. Frequency and ecological correlates of parasitism by conopid flies (Conopidae, Diptera) in populations of bumblebees. Insectes soc, 37, 14–30. Schmid-Hempel, R. & Schmid-Hempel, P. 1991. Endoparasitic flies, pollen collection by bumblebees and a potential host-parasite conflict. Oecologia (Berl.), 87, 227–232. Schmid-Hempel, R. & Schmid-Hempel, P. 1996. Larval development of two parasitic flies (Conopidae) in the common host Bombus pascuorum. Ecol. Entomol., 21, 63–70. Schneider, G. & Hohorst, W. 1971. Wanderungen der Metacercarien des Lanzett-Egels in Ameisen. Naturwissenschaften, 58, 327–328. Trexler, J. C. & Travis, J. 1993. Nontraditional regression analyses. Ecology, 74, 1629–1637.
Parasites and flower choice of bumblebees PAUL SCHMID-HEMPEL & HANS-PETER STAUFFER ETH Zu¨rich, Experimental Ecology (Received 6 May 1997; initial acceptance 9 June 1997; final acceptance 30 July 1997; MS. number: 5540)
Abstract. In a field experiment, we tested whether workers of bumblebees, Bombus pascuorum and B. humilis, parasitized by larvae of conopid flies, Physocephala rufipes and Sicus ferrugineus, differ in their flower choice from unparasitized ones. We collected workers at random in the field and immediately tested them in experimental arenas that offered the choice of a reference plant (red clover, Trifolium pratense) versus a test plant (from five species). The choices of 396 workers were analysed with logistic regression models (logit analysis). We performed all tests in the same field and at the same time where the workers were foraging naturally. On average, the parasitized bees were less likely to visit the reference plant. In addition, they were more likely to switch plant species even after the first visit in the 1998 The Association for the Study of Animal Behaviour experimental sequence.
The presence of parasites often affects the behaviour of hosts. In several cases, such behavioural modifications have been found to provide fitness advantages for the parasite. For example, survival of the host or transmission to a new host may be facilitated to ensure the development and propagation of the parasite (e.g. Holmes & Bethel 1972; Moore 1984; Milinski 1990; Keymer & Read 1991; Mu¨ller 1994). Fitness advantages for the host have also been identified, for example, impediments to parasite growth by altered thermal preferences (e.g. Boorstein & Ewald 1987; Mu¨ller & Schmid-Hempel 1993). However, parasite-induced behavioural changes not only involve conspicuous cases, such as the ‘topping behaviour’ of worker ants infected by liver flukes (where infected ants climb up into vegetation and become exposed to the next host of the parasite; Hohorst & Graefe 1961; Schneider & Hohorst 1971), but may actually often go unnoticed because of their rather subtle effects. Particularly where ecologically important behaviours are involved, parasitism may typically be a subtle factor that nevertheless has important effects for ecological relationships (e.g. Dobson 1988). In this study we investigate an ecologically important behaviour: flower choice by pollinators that is affected by parasitism. Correspondence: P. Schmid-Hempel, ETH Zu¨rich, Experimental Ecology, ETH-Zentrum NW, CH-8092 Zu¨rich, Switzerland (email: [email protected]). 0003–3472/98/040819+07 $25.00/0/ar970661
Bumblebees, Bombus spp., are annual, primitively eusocial species. While foraging for pollen and nectar, workers become exposed to predators and parasites, for example, infectious protozoa that can be contracted on flowers (Durrer & Schmid-Hempel 1994) or conopid flies (Conopidae, Diptera: Schmid-Hempel et al. 1990). Conopids insert an egg inside the abdomen of workers, but also in young queens or males. They develop to pupation within 10–12 days, at which time the host is killed (Schmid-Hempel & SchmidHempel 1996). The pupa stays inside the host and the next generation of flies emerges the following year. Parasitism by conopid flies is very common in natural populations, where larvae and eggs are present in 30–70% of all workers (Schmid-Hempel et al. 1990). Bumblebees are important pollinators in many temperate or cool areas. Their ecology, physiology or pollination strategies have therefore been studied intensively (Inouye 1977; Pyke 1978; Ranta & Tiainen 1982; Pekkarinen 1984; Ellington et al. 1990). Bumblebees have often been cited as a prime example of species co-existence that is driven by passive competition over limited resources (Pyke 1982; Begon et al. 1990). At the same time, recent work has demonstrated substantial degrees of parasitism in Bombus spp. which, moreover, is associated with important ecological characteristics of different species and bee assemblages (Schmid-Hempel & Durrer 1991;
1998 The Association for the Study of Animal Behaviour
819
820
Animal Behaviour, 55, 4
Durrer & Schmid-Hempel 1995; S. Durrer & P. Schmid-Hempel, unpublished data). In addition, preliminary field observations have suggested that flower choice by worker bees is not independent of parasitism (Schmid-Hempel & Schmid-Hempel 1990). Here, we have used field experiments to investigate this phenomenon in more detail. The study was carried out in the flowering meadows commonly visited by bumblebees. We asked to what extent the flower choice of workers of the bumblebees B. pascuorum Scop., an abundant and widespread species, and B. humilis Illiger, a rare and less widely found species in the region, is affected by the presence of endoparasitic larvae of conopid flies. In particular, we analysed whether parasitism affects choice differently when different flowers are offered as an alternative, and to what extent flower constancy, that is, the tendency to choose the same flowering plant species in successive visits, is affected by parasitism. Offering different flowers as alternatives can also generate a preference ‘profile’ of the standard plant against a variety of other available plant species and reveal whether this is affected by parasitism.
MATERIALS AND METHODS Between June and September 1993, we captured foraging workers of B. pascuorum and B. humilis at random along several transects in a dry meadow (mesobrometum) near Blauen, northwestern Switzerland. The majority of species can be readily distinguished in the field, but neither age nor parasitism of bumblebee workers can be detected before dissection in the laboratory. Bees were assigned at random to one of the treatments and were therefore randomized with respect to age, individual history or experience. In particular, with the post hoc examination of our sample, we found no relationship between time of day and the percentage of parasitized bees among those caught and tested. Similarly, the observer was ignorant with respect to parasitism of the bees. After their capture, the bees were deprived of food in a cage for 2 h to standardize hunger levels. They were then immediately transferred to the experiment where each (individually marked) worker was allocated at random to one of eight available test arenas. We used each bee only once. Each test arena was a flight cage (4040
75 cm), covered with fine mesh, that contained flowers of four plants near the four corners of its base. In two opposite corners, the flowers were from the ‘reference plant’, red clover, Trifolium pratense L. Red clover was by far the most abundant plant species in the area and clearly a favourite food plant of the bumblebees. The other two opposite corners were stocked with flowers of the same ‘test plant’ species: the alternative offered. The test plant species were kept the same for each experimental run and cage, but varied systematically over cages and successive runs. With this procedure, we tested the preference of workers, relative to the reference plant, for flowers of Betonica officinalis L., Centaurea jacea L., Prunella grandiflora L., Prunella vulgaris L. and Trifolium repens L. These five plant species were also among the most common ones in the study area. This procedure of pair-wise comparisons with the reference plant generated a preference ‘profile’ relative to the most preferred plant. We restricted the experiment to one such profile only, since it was impractical to test all possible combinations. We picked the flowering plants fresh in the field and assigned the individual plants randomly to the test cages according to schedule. We put all plants in a small tube with water approximately 1 h before the tests. Owing to differences in availability, not all test plant species could be tested with the same number of replicates. We placed the cages containing the bees and the five test arenas in the same meadow from where the bees came. Depending on the weather and bee activity, 5–20 bees could be tested per day. Because the sequence of testing plant species was random with respect to time of day, experimental day or parasitism (as revealed by post hoc dissections), any effect of the depletion of nectar in the surrounding field should have affected all tests in the same way and was thus randomized within the framework of our experiment. In an experimental run, we released the worker into the test arena and subsequently observed it for 1 h. We recorded all flower visits. Bees that made fewer than four visits before the time limit ran out were excluded from the analysis. After each experimental run, we carefully cleaned each cage and replaced the flowers. The bees were frozen for later inspection in the laboratory, and the plants were dry-pressed. The presence or absence of an endoparasite was checked under the
Schmid-Hempel & Stauffer: Parasites and flower choice microscope with the scorer unaware of treatments. We focused on eggs and larvae of conopids. Brood of two widespread and common species, Physocephala rufipes F. and Sicus ferrugineus L. (Schmid-Hempel & Schmid-Hempel 1996), were present in our samples. As no difference could be found between these two groups, we pooled the data for the final analysis presented here. We also noted the occurrence of other parasites, such as Crithidia bombi (Trypanosomatidae; Lipa & Triggiani 1980). However, in our samples the presence of other parasites was not related to parasitism by conopids nor to flower choice. Other parasites were therefore not analysed further. We analysed the data with SAS, procedure CATMOD, using logistic regression for categorical variables. In particular, we analysed the choices of test workers (i.e. test plant or reference plant) with respect to the presence or absence of conopid parasitism, bee species, B. pascuorum, or humilis, and the test plant species used. For the analysis of flower choice, we took only the first visit into account. For the analysis of flower constancies, we coded, with the variable ‘switch’, according to whether the same or a different flower species was visited in the next visit. All bees made four successive visits, so we used the response variables ‘switch1’ and ‘switch2’ to quantify the constancy of the bees on the second (relative to the first) and third (relative to the second) visit, respectively. We could not evaluate the constancy between the third and fourth visit, since only three out of eight possible combinations were observed, and therefore linear dependencies occurred in the analysis. Since the successive visits were made by the same individual worker, we analysed switch1 and switch2 as repeated measures. For this reason, we used the variable ‘Seq’ which was generated by the model according to the two response functions in each combination, that is, the odds ratio of Seq was calculated as the ratio of the odds ratios of switch1 and switch2 (with df=1). The odds ratio of Seq thus indicates how much more likely a bee switches on the second, relative to the first, visit compared with the third, relative to the second, visit. Throughout, odds ratios of 1 imply no effect of the tested factor. We chose the best statistical model by comparing all possible models (flower choice), or a number of plausible models (flower constancy), starting from the saturated model (with all factors and interactions present), and
821
then removing all main effects and interaction terms that produced non-significant deviations. SAS estimates the regression coefficients of such logit models with maximum-likelihood for the flower choice analysis and weighted least squares for the analysis of flower constancies (Anderson et al. 1980; SAS Institute 1990; Trexler & Travis 1993). All reported significance levels are two tailed.
RESULTS We tested 540 bees, of which 247 B. pascuorum and 149 B. humilis made at least four choices in the test arenas and were thus included in the final analysis. A total of 115 workers of B. pascuorum (46.6%) and 81 of B. humilis (54.54%) were parasitized. Flower Choice Table I summarizes the best model for the analysis of flower choice, that is, the first visit made to a reference or test plant. None of the interaction terms nor the main effect ‘bee species’ (B. pascuorum versus B. humilis) generated a significant deviation. The best model of Table I therefore contains only the (additive) main effects of ‘parasitism’ (yes/no) and ‘plant species’ (the five test plants). It was 1.54 times less likely that a parasitized bee was visiting the reference plant, T. pratense, than any test plant (Table I). Since no interactions occurred, this effect was independent of the test plant species offered (Fig. 1). On the other hand, the kind of test plant that was offered as an alternative to the reference plant affected the bees’ preferences, independent of parasitism. For example, B. officinalis, C. jacea or P. vulgaris had almost no effect on choice, but T. repens was clearly avoided and P. grandiflora preferred over the reference plant (Table I; see 50% preference in Fig. 1). Flower Constancy The response variable, that is, to visit the same plant species or to switch to the alternative test plant offered, was analysed with a logit model in relation to the same factors as before, that is, parasitism, test plant species, bee species and their
Animal Behaviour, 55, 4
822
Table I. Logit model for the effects of Parasite and Plant species on first flower choice Factor*
Coefficient†
Odds ratio‡
95% C.I.
P for factor level§
Intercept Parasite Plant species** B. officinalis C. jacae P. grandiflora P. vulgaris T. repens
0.062 0.429
0.121 0.216
1.064 1.535
0.742–1.191 1.006–2.343
0.048
0.015 0.481 0.925 0.001 1.392
0.184 0.244 0.203 0.253 0.290
1.015 0.618 0.397 0.999 4.022
0.708–1.456 0.383–0.997 0.267–0.590 0.608–1.640 2.277–7.105
0.049 <0.0001 <0.0001
ones with respect to switching (odds ratio=1.06, ). There was, however, a significant difference in that parasitized bees were more likely to switch between the first and second visits but less likely between the second and third visits, compared with their unparasitized counterparts (Fig. 2; see interaction term Parasite*Seq in Table II, odds ratio=0.745). None of the other factors or interactions produced a significant effect.
100
75
50
25
ffi B. o
200
196
M ea n
25
26
28
21
55
60
29
29
al is C. P. ja c gr an ea di flo ra T. vu lg ar is T. re pe ns
0
63
60
DISCUSSION
ci n
Observed preference for test plant (% C.I.)
Parasite: the presence of conopid eggs/larvae; Plant species: the species offered as an alternative to the reference plant. *A total of 396 workers analysed. †Regression coefficient () in the logit model. ‡Odds ratio >1 for Parasite indicates preference of parasitized bees for test plant. For Plant species it indicates preference for reference plant. §Wald’s test (=P>0.1). **Significance for effect over all plant species: P<0.0001.
Test plant Figure 1. Observed frequencies of visits (with 95% C.I. for observed percentages) to the test plant when offered as an alternative to the reference plant, T. pratense, and in relation to the presence () or absence ( ) of endoparasitic eggs or larvae of conopid flies. Sample sizes (number of workers tested) are given within the bars. For statistics of the logit model, see Table I. ‘Mean’ displays the average effect of parasitism, independent of test plant species.
interactions. Table II shows the best model. According to this analysis, parasitized bees behaved in the same manner as non-parasitized
Flower choice by pollinators, and by bumblebees in particular, is affected by a variety of factors. These include, for example, the availability of flowering plants (Heinrich et al. 1977; Heinrich 1979a, b; Inouye 1980), experience (Heinrich 1976, 1979b; Laverty 1994), nutritional status of the colony (Cartar & Dill 1990; Cartar 1991), weather conditions (Heinrich 1979a) or the presence of competitors (Inouye 1978; Pyke 1982). We now show that parasitism is also associated with differences in flower choice and with a difference in flower constancy. At least for the sample tested here, parasitized bees on average shifted away from the most common food plant in the area (i.e. our reference plant, T. pratense), independently of the alternative (‘test’) plant species offered (Table I, Fig. 1). Parasitized bees were also less constant in their choices. They tended to switch plant species on the first occasion, that is, between the
Schmid-Hempel & Stauffer: Parasites and flower choice
823
Table II. Logit model for the effects of Parasite, Plant species and Seq on flower constancy Factor*
Odds ratio‡
95% C.I.
P for factor level§
0.588 0.061 0.082 0.294
0.088 0.162 0.140 0.140
1.800 1.063 0.921 0.745
1.515–2.139 0.774–1.461 0.700–1.213 0.566–0.981
0.036
0.167 0.299 0.183 0.362 1.005
0.145 0.178 0.145 0.189 0.213
1.175 1.348 1.201 1.436 0.366
0.885–1.561 0.952–1.909 0.903–1.596 0.992–2.709 0.241–0.556
0.093 0.055 <0.0001
Coefficient†
Intercept Parasite Seq Parasite*Seq Plant species** B. officinalis C. jacae P. grandiflora P. vulgaris T. repens
80
70
50
Second
200
196
200
60
196
Proportion of bees switching plant species (% C.I.)
Parasite: the presence of conopid eggs/larvae; Plant species: the species offered as an alternative to the reference plant; Seq: the sequence of plant visited. *A total of 396 workers analysed. †Regression coefficient () in the logit model. ‡Odds ratio for Seq >1 indicate that the bee is more likely to switch between first and second visit than between second and third visit (see Methods). §Wald’s test (=P>0.1). **Significance for effect over all plant species: P<0.0001.
Third Visit
Figure 2. Observed percentage of bees (95% C.I.) switching plant species on the second or third visit, respectively, compared to the previous visit. : Parasitized;
: non-parasitized. Sample sizes (number of workers tested) are given within the bars.
first and second visit (Table II). An earlier, observational study (Schmid-Hempel & SchmidHempel 1990) similarly showed a systematic bias in flower visit rates in that parasitized workers of
B. pascuorum also shifted away from the more common of two species of flowering plants available in the area (P. grandiflora was more common than B. officinalis in this earlier study). We used randomly selected workers from the field that had become naturally infested by conopid flies and tested them immediately in the field. Hence, we can assume that they behaved in the natural context against their background experience. Our experimental protocol has the advantage of testing an actual field situation but poses the problem of causation, since naturally occurring parasitism was used. Experimental infections would have been preferred but are impractical for the necessarily large sample sizes. However, several lines of reasoning suggest that conopid parasitism is indeed responsible for the observed differences. For example, other known behavioural changes relate to parasitism rather than, for example, worker age (Schmid-Hempel & Schmid-Hempel 1991; Mu¨ller & Schmid-Hempel 1993). From many years of field studies, we find that conopid attacks are not associated with particular flowering plant species. Hence individual plant preferences of workers (e.g. Heinrich 1976) do not affect their risk of parasitism. Rather, our observations show that conopid females wait near the ground in flower fields and attack bees flying by. Together with the typical fine-grained mosaic of plant species, relative to the activity range of
824
Animal Behaviour, 55, 4
conopids and bees in the field, this behaviour makes it very unlikely that workers are only at risk at certain plant species but not at others. Conopids do, however, generally prefer larger hosts (Mu¨ller et al. 1996). No such difference in body size could be found in the sample used in the present experiments, nor did body size relate to flower choices. We have not investigated what underlying mechanisms cause the observed behavioural differences. Previous studies have shown that workers parasitized by conopids do not maintain thoracic temperature to the same degree (Heinrich & Heinrich 1983) and have reduced nectarcarrying capacities (Schmid-Hempel & SchmidHempel 1991). Given the large physical dimensions of conopid larvae inside their host’s abdomen (Schmid-Hempel & Schmid-Hempel 1996), it is likely that parasitized bees are simply physiologically stressed, for example, in a permanent state of nutritional deficit. This may cause the behavioural changes. However, it remains unexplained why certain flowers rather than others become less preferred or to what extent this might reflect an advantageous behavioural decision. On the other hand, the observed behavioural changes have potentially far-reaching effects at different levels. For example, the present and earlier (Schmid-Hempel & Schmid-Hempel 1990) results suggest that as parasite frequency increases in a given bee assemblage, certain plant species are expected to be visited more often, for example the less common species such as Betonica in the earlier study, or Prunella spp. and T. repens in this study (Fig. 1). Besides the straightforward effects on the pollination of these plants, the interactions among competing bee species in the local assemblage should also be affected. In addition, the transmission of other parasites will be different when a major pathway, that is, via the flowers (Durrer & Schmid-Hempel 1994), is modified. Such secondary consequences in communities of interacting species are not well understood, but in cases such as the present one parasite-induced behavioural changes may be at the root of the process.
ACKNOWLEDGMENTS We thank the community of Blauen for permission to work on the site. S. Cameron com-
mented on the manuscript. H.-U. Reyer (Zu¨rich University) provided logistic support for H.-P. S. M. Ma¨chler (ETH) helped to analyse the data. The study was supported by grants of the Swiss National Science Foundation (no. 3100-049040.95 to P.S.-H.).
REFERENCES Anderson, S., Auquier, A., Hauk, W. H., Oakes, D., Vandaele, W. & Weisberg, H. I. 1980. Statistical Methods for Comparative Studies. New York: John Wiley. Begon, M., Harper, J. L. & Townsend, C. R. 1990. Ecology: Individuals, Populations and Communities, 2nd edn. Oxford: Blackwell. Boorstein, S. M. & Ewald, P. W. 1987. Costs and benefits of behavioral fever in Melanoplus sanguinipes infected by Nosema acridophagus. Physiol. Zool., 60, 586–595. Cartar, R. V. 1991. Colony energy requirements affect response to predation risk in foraging bumble bees. Ethology, 87, 90–96. Cartar, R. V. & Dill, L. M. 1990. Why are bumble bees risk-sensitive foragers?. Behav. Ecol. Sociobiol., 26, 121–127. Dobson, A. P. 1988. The population dynamics of parasite-induced changes in host behavior. Q. Rev. Biol., 63, 139–165. Durrer, S. & Schmid-Hempel, P. 1994. Shared use of flowers leads to horizontal pathogen transmission. Proc. R. Soc. Lond. Ser. B, 258, 299–302. Durrer, S. & Schmid-Hempel, P. 1995. Parasites and the regional distribution of bumble bee species. Ecography, 18, 114–122. Ellington, C. P., Machin, K. E. & Casey, T. M. 1990. Oxygen consumption of bumblebees in forward flight. Nature, Lond., 347, 472–473. Heinrich, B. 1976. The foraging specializations of individual bumblebees. Ecol. Monogr., 46, 105–128. Heinrich, B. 1979a. Bumblebee Economics. Cambridge, Massachusetts: Harvard University Press. Heinrich, B. 1979b. ‘Majoring’ and ‘minoring’ by foraging bumblebees, Bombus vagans: an experimental analysis. Ecology, 60, 245–255. Heinrich, B. & Heinrich, M. J. E. 1983. Heterothermia in foraging workers and drones of the bumblebee Bombus terricola. Physiol. Zool., 56, 563–567. Heinrich, B., Mudge, P. R. & Deringis, P. G. 1977. Laboratory analysis of flower constancy in foraging bumblebees: Bombus ternarius and B. terricola. Behav. Ecol. Sociobiol., 2, 247–265. Hohorst, W. & Graefe, G. 1961. Ameisen: obligate Zwischenwirte des Lanzettegels (Dicrocoelium dendriticum). Naturwissenschaften, 48, 229–230. Holmes, J. C. & Bethel, W. M. 1972. Modification of intermediate host behavior by parasites. In: Behavioral Aspects of Parasitic Transmission (Ed. by
Schmid-Hempel & Stauffer: Parasites and flower choice E. U. Canning & C. A. Wright), pp. 123–149. New York: Academic Press. Inouye, D. W. 1977. Species structure of bumblebee communities in North America and Europe. In: The Role of Arthropods in Forest Ecosystems (Ed. by W. J. Mattson), pp. 35–40. New York: Springer. Inouye, D. W. 1978. Resource partitioning in bumblebees: experimental studies of foraging behavior. Ecology, 59, 672–678. Inouye, D. W. 1980. The effect of proboscis and corolla tube lengths on patterns and rates of flower visitation by bumblebees. Oecologia (Berl.), 45, 197–201. Keymer, A. E. & Read, A. F. 1991. Behavioural ecology: the impact of parasitism. In: Parasite–Host Associations: Coexistence or Conflict? (Ed. by C. A. Toft, A. Aeschlimann & L. Bolis), pp. 37–61. Oxford: Oxford University Press. Laverty, T. M. 1994. Bumblebee learning and flower morphology. Anim. Behav., 47, 531–545. Lipa, J. J. & Triggiani, O. 1980. Crithidia bombi sp.n. a flagellated parasite of a bumble-bee Bombus terrestris L. (Hymenoptera, Apidae). Acta Protozool., 27, 287– 290. Milinski, M. 1990. Parasites and host decision-making. In: Parasitism and Host Behaviour (Ed. by C. J. Barnard & J. M. Behnke), pp. 95–116. London: Taylor & Francis. Moore, J. 1984. Altered behavioral responses in intermediate hosts: an acantocephalan parasite strategy. Am. Nat., 123, 572–577. Mu¨ller, C. B. 1994. Parasite-induced digging behaviour in bumblebee workers. Anim. Behav., 48, 961–966. Mu¨ller, C. B. & Schmid-Hempel, P. 1993. Exploitation of cold temperature as defence against parasitoids in bumblebees. Nature, Lond., 363, 65–67. Mu¨ller, C. B., Blackburn, T. M. & Schmid-Hempel, P. 1996. Field evidence that host selection by conopid parasitoids is related to host body size. Insectes soc., 43, 227–233.
825
Pekkarinen, A. 1984. Resource partitioning and coexistence in bumblebees (Hymenoptera, Bombinae). Annls Entomol. Fenn., 50, 97–107. Pyke, G. H. 1978. Optimal foraging in bumblebees and coevolution with their plants. Oecologia (Berl.), 36, 281–293. Pyke, G. H. 1982. Local geographic distributions of bumble bees near Crested Butte, Colorado: competition and community structure. Ecology, 63, 555–573. Ranta, E. & Tiainen, M. 1982. Structure in seven bumblebee communities in eastern Finland in relation to resource availability. Holarct. Ecol., 5, 48–54. SAS Institute 1990. SAS User’s Guide: Statistics, Version 6. Cary, North Carolina: SAS Institute. Schmid-Hempel, P. & Durrer, S. 1991. Parasites, floral resources and reproduction in natural populations of bumblebees. Oikos, 62, 342–350. Schmid-Hempel, P. & Schmid-Hempel, R. 1990. Endoparasitic larvae of conopid flies alter pollination behaviour of bumblebees. Naturwissenschaften, 27, 50–52. Schmid-Hempel, P., Mu¨ller, C., Schmid-Hempel, R. & Shykoff, J. A. 1990. Frequency and ecological correlates of parasitism by conopid flies (Conopidae, Diptera) in populations of bumblebees. Insectes soc, 37, 14–30. Schmid-Hempel, R. & Schmid-Hempel, P. 1991. Endoparasitic flies, pollen collection by bumblebees and a potential host-parasite conflict. Oecologia (Berl.), 87, 227–232. Schmid-Hempel, R. & Schmid-Hempel, P. 1996. Larval development of two parasitic flies (Conopidae) in the common host Bombus pascuorum. Ecol. Entomol., 21, 63–70. Schneider, G. & Hohorst, W. 1971. Wanderungen der Metacercarien des Lanzett-Egels in Ameisen. Naturwissenschaften, 58, 327–328. Trexler, J. C. & Travis, J. 1993. Nontraditional regression analyses. Ecology, 74, 1629–1637.