At Barrier Lake, the phorid parasitoid A. borealis probably affected bumble-bees more than the conopid parasitoid P. texana. Phorids parasitised bees during a longer period (Fig. 1a), at higher frequencies (Figs 1b and 2), and imposed greater host mortality (Fig. 4) than did conopids. The average phorid prevalence at Barrier Lake was twice the 5–10% range reported previously (Brown, 1993; Disney, 1994), peaking at nearly 50% in mid summer (Fig. 1b).
Phorid parasitism of bumble-bees
At Barrier Lake, phorid prevalence differed among bumble-bee species (Fig. 3a). Because bee-species abundance did not correlate positively with phorid prevalence, it seems unlikely that differences in the seasonal abundance of host species can explain the observed patterns in phorid prevalence (e.g. more abundant bee species experiencing higher prevalence). In addition, there was little evidence that host size differed consistently between phorid parasitised and unparasitised bumble-bees. Further investigation of host selection by A. borealis is in progress.
Phorid prevalence differed among pollen-collecting but not among nectar-collecting workers (Fig. 3b); however the interpretation of differences in prevalence among pollen collecting workers is hampered by a lack of information about the cause and effect relationship between parasitism and pollen collection. On the one hand, variation in parasitism between pollen and nectar collectors may arise because the risk of infection differs according to host foraging behaviour. For example, differences in pollen and nectar collecting between bumble-bee species, as well as differences in plant species preference among bumble-bee species (e.g. Brian, 1957; Liu et al., 1975; Heinrich, 1979), may contribute to these trends if A. borealis locates hosts non-randomly with respect to plant species. Alternatively, parasitism may alter the propensity of hosts to collect pollen. Schmid-Hempel and Schmid-Hempel (1991) found that fewer B. pascuorum workers harbouring late-instar conopid larvae collected pollen than did unparasitised conspecifics. If parasitism alters host behaviour such that infested bumble-bees are more or less likely to collect pollen than uninfested bees, this result should be generally consistent between host species. In this study, however, the relative incidence of parasitism for pollen and nectar collectors differed among host species (Fig. 3b), making this explanation less feasible.
Phorid parasitism reduced the survival of bumble-bees severely after capture. Bumble-bees containing phorid larvae had considerably shorter residual lifespans than unparasitised bees and, at best, survived less than half as long as worker bumble-bees survive in the field (for lifespan of field bumble-bees, see Rodd et al., 1980). Further, the residual lifespans of bees parasitised by phorids varied little within and among species and between sexes, suggesting that the rate of development of A. borealis larvae and their impact on bee lifespan is very consistent. Consequently, the observed mortality effects of phorid parasitism (Fig. 4) probably constitute real costs for bees. Based on Rodd et al.'s (1980) estimates of bumble-bee worker mortality, the results presented here suggest that phorids may reduce the average lifespan of workers by up to 70%. Although rates of male bumble-bee mortality are not known, it seems reasonable to suggest that phorids reduce the lifespan of males significantly. Phorid-induced mortality of workers may diminish the resources (pollen and nectar) gathered for colony growth. In turn, reduced colony size could ultimately decrease the number and/or quality of sexuals (males and queens) produced (Owen et al., 1980; Müller & Schmid-Hempel, 1992a,b).
The residual lifespan of parasitised males, but not workers, varied negatively with the number of A. borealis larvae. Further, males parasitised by phorids contained significantly more phorid larvae than did workers. Such a negative relationship between residual lifespan and parasitoid load may represent an increasing mortality effect at high levels of infestation (i.e. many larvae per bee) or higher levels of superparasitism (hosts containing larvae from multiple phorid females) in older bees.
Conopid parasitism of bumble-bees
Conopid prevalence among bumble-bees at Barrier Lake (Figs 2 and 3a) was less than half that reported from Europe (Schmid-Hempel & Schmid-Hempel, 1988; Schmid-Hempel et al., 1990; Shykoff & Schmid-Hempel, 1991), where conopids parasitised 5–10% of males and 20–30% of workers. Although the results presented here agree quite closely with other North American data (MacFarlane & Pengelly, 1974), more work is necessary to determine whether conopid parasitism is typically less common in North America than in Europe. At least in Europe (Schmid-Hempel & Schmid-Hempel, 1988, 1996a; Schmid-Hempel et al., 1990) and Japan (Maeta & MacFarlane, 1993), the presence of both Physocephala and Sicus conopids contributes to the higher incidence of bumble-bee parasitism than at Barrier Lake, where only Physocephala occurred. For example, Schmid-Hempel and Schmid-Hempel (1996a) found that bumble-bees in northern Switzerland were parasitised most often by Sicus ferrugineus Latreille, whereas parasitism by Physocephala rufipes Fabricius accounted for less than half of the total conopid parasitism. Indeed, greater conopid diversity in the Palaearctic than in the Nearctic has been suggested to pose a substantially greater risk to certain bee species (e.g. Megachile rotundata Fabricius) (Doroshina, 1991).
In contrast to B. bifarius, B. occidentalis, and B. flavifrons, which experienced a similar incidence of conopid parasitism, no B. californicus workers collected during 1998–1999 contained conopid larvae or eggs. Because B. californicus is a late-season bee at Barrier Lake, their relative scarcity during the period of conopid activity (Fig. 1a) may explain this result. Similarly, Schmid-Hempel and Schmid-Hempel (1988) and Schmid-Hempel et al. (1990) found particularly low incidences of conopid parasitism among B. pratorum and B. hortorum, species that are active primarily during early summer (for bumble-bee phenologies, see Prys-Jones & Corbet, 1987), prior to conopid activity; however the occurrence of conopid parasitism in B. californicus workers was still significantly lower than that among other species when considering only late summer sampling dates (1998, G = 22.84, d.f. = 1, P < 0.001; 1999, G = 8.15, d.f. = 1, P < 0.01), suggesting that host species phenology alone cannot explain this pattern.
Another explanation for the low incidence of conopid parasitism among B. californicus workers is that P. texana does not seek hosts at the plant species favoured by B. californicus. At Barrier Lake, B. californicus is a relatively long-tongued bumble-bee and foraged almost exclusively from flowers with relatively deep corollas, particularly Castilleja spp. and Trifolium pratense. In contrast, P. texana is a more catholic flower visitor, frequenting the relatively shallow flowers of composites for nectar (Freeman, 1966; Smith & Peterson, 1987). Such differences in flower-visiting habits may result in a low risk of conopid parasitism for B. californicus workers if conopids use flowers as both nectar sources and host-location sites. Indeed, Maeta and MacFarlane (1993) observed that European bumble-bee species with more specialised flower-visiting habits experience lower levels of conopid parasitism. Further, in some eastern Canadian populations, workers of the relatively long-tongued bumble-bee species, B. fervidus, experience a substantially lower incidence of conopid parasitism than other bumble-bee species (C. Cormier and D. B. McCorquodale, pers. comm.).
Surprisingly, at Barrier Lake, workers parasitised by conopids died at a similar rate to unparasitised workers (Fig. 2a). In contrast, studies in Europe have found conopid parasitism to increase bumble-bee mortality (Schmid-Hempel & Schmid-Hempel, 1988, 1989, 1991). The effect of conopid parasitism on worker survival found in this study probably differs from the results of Schmid-Hempel and Schmid-Hempel (1988, 1989, 1991) because unparasitised bees in this study had considerably shorter lifespans after capture than those in European studies. For example, the mean lifespan for unparasitised workers reported here (11.22 days) is less than half of that found by Schmid-Hempel and Schmid-Hempel (1991) (23.2 days). Despite this difference, the average residual lifespan of workers parasitised by Physocephala conopids at Barrier Lake (9.66 days) is very similar to that estimated from European studies (8.3 days; Schmid-Hempel & Schmid-Hempel, 1989) and to the estimated development time of a Physocephala conopid larva (11.4 days; Schmid-Hempel & Schmid-Hempel, 1996b). Based on these results, it appears that the mean lifespan of workers parasitised by Physocephala conopids is consistently close to previous estimates of the mean lifespan of workers in the field (13.2 days; Rodd et al., 1980). Consequently, significant worker mortality arising from conopid parasitism may not be apparent in all bumble-bee populations and is unlikely to be a general influence on bumble-bee life-history traits, such as colony investment into sexuals and population sex ratio, as suggested previously (e.g. Schmid-Hempel & Schmid-Hempel, 1988; Müller & Schmid- Hempel, 1992a,b).
Contrary to expectation, at Barrier Lake, workers parasitised by conopids displayed a lower mortality rate than unparasitised bees during early residual lifespan classes (1–9 days). This pattern may indicate that female conopids tend to parasitise relatively young bees. Schmid-Hempel and Schmid-Hempel (1996b) estimated the developmental period of conopid larvae to be 11 days, which is very close to the average residual lifespan reported here for unparasitised bees. Consequently, only young bees are likely to live long enough for successful development of conopid larvae, favouring active selection of young bees by female conopids. Indeed, Schmid-Hempel and Schmid-Hempel (1991) estimated (via fat body condition, a correlate of bee age) that B. pascuorum workers parasitised by conopids were significantly younger than unparasitised conspecifics. Consequently, young bees parasitised by conopids may survive better shortly after capture than unparasitised bees, which on average are older.
Differences in phorid and conopid prevalence among bumble-bee workers and males
Male bumble-bees experienced consistently less parasitism by conopids than workers, both in this study (Fig. 2) and in other populations (MacFarlane & Pengelly, 1974; Schmid-Hempel & Schmid-Hempel, 1988; Maeta & MacFarlane, 1993). Interestingly, male bumble-bees at Barrier Lake experienced significantly higher rates of phorid parasitism than workers (Fig. 2); however because phorid and conopid parasitism were found to occur independently of one another, it is unlikely that such contrasting incidences of parasitism are the result of phorid larvae outcompeting conopid larvae in cases of multiparasitism. Further, similarly low rates of conopid parasitism in males have been observed at sites with comparable incidences of conopid parasitism, yet where phorids are absent (T. L. Whidden and M. C. Otterstatter, unpublished).
Alternatively, intersexual differences in the incidence of phorid and conopid parasitism may reflect differences in the ecology of worker and male bumble-bees. After leaving their parent colony, males remain in the field during both day and night, whereas workers return frequently throughout the day and typically return for the duration of the night. Because environmental hazards away from the nest probably cause most bee mortality (Brian, 1952; Sakagami & Fukuda, 1968), males should be particularly prone not only to weather (Brian, 1952, 1965) but also to predators such as crab spiders (Araneae: Thomisidae) (Morse, 1979, 1986). Consequently, males may represent low-quality hosts for conopid larvae, which require a relatively long host lifespan for development, but not so for the rapidly developing phorid larvae. Furthermore, exposure of field-dwelling males to cool night-time temperatures (compared with the warmer temperatures workers would experience inside their parent colony) may prolong the development of parasitoid larvae. If this were the case, conopid larvae might not be able to complete development in a typical male bumble-bee before it succumbed to extrinsic sources of mortality.
Prevalence of the phorid parasitoid Apocephalus borealis differed among bumble-bee species, sexes, and between pollen and nectar collectors. Male bees experienced a higher incidence of phorid parasitism than workers in all host species. Prevalence of the conopid parasitoid Physocephala texana differed by host species and sex, with male bees experiencing a significantly lower incidence of parasitism than workers. The effect of host size on the incidence of phorid and conopid parasitism was typically weak and varied between host species and years. Surprisingly, conopid-imposed host mortality was non-significant, while that of phorids was especially severe. These results suggest that while conopid parasitism may be less important locally than elsewhere (e.g. Müller & Schmid-Hempel, 1992a; Maeta & MacFarlane, 1993), phorid parasitism may place significant stress on bumble-bee populations.