Development Time . We measured the effect of host stage on total development time (egg-to-adult) of S. aphidivorus . Twenty 1-7 day old, mated S. aphidivorus females were introduced for 4 h in cages containing potato plants infested with either ca. 400 parasitised aphids or 400 mummies. Following parasitisation, mummies were gently removed from the foliage and individually reared in gelatine capsules whereas aphids were left on the plants until mummification after which they were put in capsules. Hyperparasitoid emergence was monitored every eight hours and the sex of each adult was determined. Temperature in cages and gelatine capsules were monitored regularly using thermocouples (OMEGA HH23). Mummies from which a hyperparasitoid had not emerged 10 days after peak emergence were not included in the analysis.

Longevity. We compared the effect of both host stages on the longevity of adult males and females at a constant temperature of 20°C. Newly emerged S. aphidivorus from the development time experiment (n > 60 per host stage and hyperparasitoid sex), were kept individually in small ventilated cylindrical cages (5 cm in diameter and 10 cm in height) with a supply of 40% sugar water, replaced every 3-4 days. Hyperparasitoids did not have access to hosts and were checked daily until death.

Fecundity, Immature Mortality and Sex Ratio . We measured total and age-specific fecundity, immature mortality, as well as the sex ratio of the progeny of S. aphidivorus females reared from parasitised aphid or aphid mummy hosts. These females were given mummies to parasitise in this experiment, since previous tests indicated that S. aphidivorus females reared on the same host stage were equally fecund when provided with either parasitised aphids or aphid mummies (ANOVA, F = 2.24, df = 1, 28; P = 0.1456; R.B., unpublished data). Females and males used in the experiment were obtained as previously described (see Development time). One newly emerged female and two males were isolated in small ventilated cylindrical cages (5 cm in diameter and 10 cm in height) with a supply of 40% sugar water, and the males removed five days later. From day of emergence to hyperparasitoid death, each female was provided 70 newly formed mummies, glued (Lepage Bondfast®) on a potato leaf held by the petiole in a glass vial filled with wet cotton wool and placed into the cage. Preliminary tests showed that S. aphidivorus reproduction measured in this way is maximal when offered 70 mummies per day. Less than 5% of the mummies were superparasitised in this set up (R.B., unpublished data). After each 24-h period, mummies were put individually in gelatine capsules, held in a growth chamber, and monitored daily for insect emergence. The secondary sex ratio (proportion of males at emergence) was determined. Mummies from which nothing had emerged were dissected (400X magnification) 10 days after hyperparasitoid peak emergence to determine if they contained a dead primary parasitoid or a dead hyperparasitoid. Fecundity of S. aphidivorus was calculated by summing the number of adults emerging and dead hyperparasitoid within the mummies. Syrphophagus aphidivorus immature mortality was expressed as the proportion of hyperparasitised mummies containing dead larva, pupa or adult hyperparasitoids. Fifteen females were tested per treatment.

Body size . We compared the effect of host stage on the body size of S. aphidivorus adult males and females. Cohorts of parasitised aphid and mummy hosts were produced and parasitised as described above (see Hosts used in the experiments). In this instance, mummies of similar weight (range 0.9-1.1 μg) were selected, so that observed differences in hyperparasitoid adult size would not be affected by mummy size. Following emergence, (maximum delay 24 h) hyperparasitoids were killed at -20ºC, dried for four days at 60ºC and individually weighed on a Mettler Toledo UMT microbalance. The head width and forewing length from humerus to apex were also used as size index and were measured to the nearest 0.01 mm using a stereomicroscope (400X magnification) equipped with an ocular micrometer.

Intrinsic rate of population increase (rm) . The rm is a demographic parameter used to estimate the population growth potential of an organism under given ecological conditions (Southwood and Henderson 2000). The rm was estimated for each host stage by repeated iteration of the Birch formula (Birch 1948):

Σ e-rmx lxmx = 1

x

where x is female age, lx is the fraction of females surviving to age x and mx is the age-specific fertility that records the number of living females born per female of age x, calculated from the daily sex ratio data measured in the fecundity experiment.

Female behaviour was recorded with The Observer® (Version 3, Noldus Information Technology). Each test lasted one hour, or ended when the hyperparasitoid left the patch for more than a few seconds. The duration of the following behaviours was recorded: walking, grooming, host examination and oviposition (drilling and probing of the host). Host acceptance was defined as close examination followed by apparent oviposition and was therefore calculated by dividing the number of ovipositions by the number of examinations and converted into percentages. Multiple oviposition attempts in the same host during a bout were considered as one oviposition, because in many cases the female has to change position on the host to find the ideal angle to lay her egg. However, if a female left a host and returned later, this was counted as a new oviposition (superparasitism). Despite the fact that parasitised aphids were free to move in the patch, it was possible to follow each of them individually and determine the occurrence of superparasitism. Consequently, for each female superparasitism was calculated by dividing the number of presumably hyperparasitised hosts that was accepted for oviposition, by the total number of hosts that were accepted. Based on the duration of the different recorded behaviours, a time budget was constructed. Twenty replicates were done.

To determine the effect of behavioural defenses of the aphid on the oviposition success of S. aphidivorus , we repeated the experiment using motionless aphids. In this instance, just prior to the experiment, the aphid abdomen was glued on the leaf surface, thereby preventing the aphid from walking away, kicking or dropping from the feeding site. Ten replicates were done of this experiment.

The development time from oviposition to adult emergence was shorter by ca. two days for individuals that developed from mummies than for those that developed from parasitised aphids There was no difference between development times of males and females. (GENMOD; host: χ2 = 110.55; df = 1; P < 0.0001; sex: χ2 = 0.17, df = 1; P = 0.6818; interaction; χ2 = 3.79, df = 1; P = 0.0515).

The individuals reared from parasitised aphid hosts lived much longer than those from mummy hosts, and females lived longer than males irrespective of host (Two-way ANOVA; host: F = 16.07; df = 1, 189; P < 0.0001; sex: F = 58.44, df = 1, 189; P < 0.0001; interaction: F = 0.35; df = 1, 189; P = 0.5571).

Over their life, females from mummy hosts produced twice as much offspring as females from parasitised aphid hosts ( t -test, t = 3.90; df = 16; P = 0.0013). Figure 3-1 illustrates the mean daily fecundity of females. Patterns are similar for the two host treatments, but females from mummy hosts produced more progeny per day than females from parasitised aphid hosts. Reproduction started the first day after emergence and peaked from day 3 to 5 in both cases. Females from mummy and parasitised aphid hosts had a maximum daily fecundity of 43 and 25 progenies, respectively.

Immature hyperparasitoid mortality was slightly higher in parasitised aphid hosts, although this was only marginally significant (Table 1; t -test, t = 2.05; df = 16; P = 0.0567). The percentages of S. aphidivorus that died during larval and pupal stages ranged from 1.5% to 4.3% in mummy hosts and from 2.7% to 7.5% in parasitised aphid hosts. These values are too small to be meaningfully analysed per developmental stages.

Lifetime sex ratio differed between treatments (Table 3-1; χ2 = 39.65; df = 1; P < 0.0001); it was male-biased for hyperparasitoids developing in mummy hosts (0.7) and unbiased (0.5) for those developing in parasitised aphid hosts. However, early in reproductive life the pattern was similar for both hosts with sex ratios of 0.51 and 0.52 during the first 10 days for mummy and parasitised aphid hosts, respectively (χ2 = 0.3299; df = 1; P = 0.5657). The overall difference resulted from an increase in the proportion of males produced late in the life of females originating from the mummy host (Fig. 3-2), when reproduction was minimal in the parasitised aphid treatment.

The rm of S. aphidivorus was higher by ca. 24% on mummy host (0.2494 d-1) than on parasitised aphid host (0.1895 d-1; Table 1).

There were significant differences between the hosts in dry weight and wing length of emerging hyperparasitoids, but not in head width. In addition there were significant differences between male and female size for all measurements. (Two-way ANOVA, dry weight: host: F = 13.94; df = 1, 205; P = 0.0002; sex: F = 11.10; df = 1, 205; P = 0.0010; interaction: F = 10.10; df = 1, 205; P = 0.0017; wing length: host: F = 6.76; df = 1, 203; P = 0.0100; sex: F = 103.82; df = 1, 203; P < 0.0001; interaction: F = 2.03; df = 1, 203; P = 0.1554; head width: host: F = 0.00; df = 1, 204; P = 0.9568; sex: F = 97.49; df = 1, 204; P < 0.0001; interaction: F = 0.73; df = 1, 204; P = 0.3930) (Table 3-2). The difference in size between hosts was only observed in females. Male size was the same for both hosts. Furthermore, females were always larger than males, except for dry weight on parasitised aphid hosts

Hosts were frequently examined a second time and superparasitism was common in our experimental set-up. With free aphids, 25.4% of all the hyperparasitised mummies were superparasitised (on average 2.2 mummies per patch). Only once a parasitised aphid was examined twice but host defence prevented an attack. With glued aphids, 33.3% of all the hyperparasitised mummies and 13.6% of all the hyperparasitised aphids were superparasitised (2.4 mummies and 0.3 parasitised aphids per patch).

When compared to the situation with free parasitised aphids, immobilising the aphids significantly increased the time spent examining (3 vs. 36 seconds) and ovipositing into (70 vs. 284 seconds) parasitised aphid hosts. ( t -tests: examination, t = 3.05; df = 28; P = 0.0133; oviposition t = 2.20; df = 28; P = 0.0365). The duration of the other behaviours were not different ( t -tests: examination mummy, t = 0.98; df = 28; P = 0.3350; oviposition mummy, t = 0.21; df = 28; P = 0.8366; walking, t = 0.53; df = 28; P = 0.6033; grooming t = 0.63; df = 28; P = 0.5312) (Fig. 3-4). In both experiments (free and glued parasitised aphid hosts), a hundredfold more time was spent examining and parasitising mummy hosts than parasitised aphid hosts, which confirms the preference for the mummy hosts.

Increased longevity is the only parameter that may provide a fitness gain to hyperparasitoids from the parasitised aphid host, over those from the mummy host. However, this may hold true only for males, as the reproductive period is similar for females developing in both hosts (Table 3-1); the extended, and unexplained, post-reproductive period of females from the parasitised host treatment apparently does not contribute to parasitoid fitness.

Hyperparasitoids developing in mummy hosts took two days less to reach the adult stage and females were larger and more fecund than those developing in parasitised aphid hosts. These differences led to a higher rm for S. aphidivorus on mummy hosts (0.25 d-1) than on parasitised aphid hosts (0.19 d-1), the rm being strongly correlated to developmental rate and early fecundity in arthropods (Roy et al. 2003). These rm values clearly indicate that mummy hosts are more suitable for S. aphidivorus than parasitised aphid hosts. Although parasitoid ecologists do not commonly use variations of the intrinsic rate of increase, they represent adequate measures of fitness differences for species with overlapping generations such as S. aphidivorus (Stearns 1992, Roitberg et al . 2001).

The observed difference in development time between hosts should be interpreted with caution. When a parasitised aphid is attacked by S. aphidivorus , hatching of the hyperparasitoid egg is delayed until aphid mummification (Kanuck and Sullivan 1992). The prolonged development of ca. two days in the parasitised aphid host corresponds to the time from oviposition to aphid mummification by the primary parasitoid, when the hyperparasitoid egg remains dormant in the parasitoid larva. We therefore suspect that actual development time from egg to adult is similar in both hosts, once embryogenesis has been initiated. Nevertheless, there are potential benefits associated with a shorter egg phase (excluding dormancy) for individuals developing in mummy hosts. First, it would reduce the time exposed to the immune system of the host, thereby lowering the potential risk of egg encapsulation. Second, it would also reduce the risk of mortality from competitors and natural enemies, as predicted by the slow-growth-high-mortality hypothesis (Clancy and Price 1987, Benrey and Denno 1997).

In theory, because hyperparasitoid egg development is arrested until aphid mummification, S. aphidivorus larvae should have access to the same resources for development regardless of the host stage in which the egg was laid. Under experimental conditions, after oviposition either in parasitised aphids of mummies, immature S. aphidivorus exploited the same primary parasitoid stage. Therefore, why are mummy hosts more suitable than parasitised aphid hosts? Differences in hyperparasitoid fitness likely originate from factors associated with the pre-mummification period. For instance, parasitism might affect growth of the primary parasitoid larva, thereby the overall quality of the subsequent pupa. At oviposition parasitoid females typically inject virus-like particles and venom into the host, which are important in disarming host defences and disrupting host physiology (Piek 1986, Stoltz 1993). We do not know if this is the case in S. aphidivorus . Preliminary data indicate that parasitism of hosts within live aphids by S. aphidivorus does not affect the pre-pupal weight of the primary parasitoid within the mummified aphid (R.B., unpublished data). More information is needed to assess the quality of parasitoid hosts originating from parasitised aphids vs. aphid mummies. There might also be a cost associated with arrested development for eggs laid in host larvae in live aphids. Syrphophagus species produce relatively large, nutrient rich eggs (Griswold 1929) that could be partially depleted during the resting period, thereby lowering the fitness of the resulting offspring.

Besides the transition hypothesis, the dual oviposition behaviour of S. aphidivorus might also be an evolved strategy to host distribution. Aphid mummies and parasitised aphids can either be found within or near the aphid colony (Brodeu and McNeil 1989, Müller et al . 1997) and both hosts could be simultaneously encountered on plants. The ability to attack parasitised aphid and mummy hosts may therefore provide a larger range of potential hosts to foraging S. aphidivorus females. As with all other parasitoids, host selection would be determined by the physiological state of the S. aphidivorus female and the ecological differences that exist between host stages: suitability, nutritional value, vulnerability to natural enemies, and abundance (Vinson 1984, Bell 1990).

Finally, the dual oviposition behaviour may contribute to reduce competition for hosts between S. aphidivorus and other aphid hyperparasitoids, mainly those that attack mummified aphids. Several studies examined aspects of interspecific competition between endo- and ectohyperparasitoid species (Sullivan 1972, Matejko and Sullivan 1984) and between two ectohyperparasitoids (Carew and Sullivan 1993). Competition may occur between wasps during the host selection process or through interactions between immature parasitoids. Ovipositing females can reduce the chance that other females later attack the host by marking it (Roitberg and Mangel 1988), while a parasitoid larva can eliminate a competitor by way of physical attack, chemical suppression or resource competition (Mackauer 1990). Whatever the mechanisms, the outcome frequently depends on the sequence of events: the first female to oviposit, or the first larva to emerge usually wins the competition. For example, Sullivan (1972) and Matejko and Sullivan (1984) examined interspecific larval competition between different associations of two Alloxysta species, which attack parasitised aphids, and two mummy attacking hyperparasitoids, Asaphes californicus and Dendrocerus carpenteri . They concluded that the Alloxysta species usually win competition with hyperparasitoids that attack the aphid mummywhen the latter oviposit in older mummies, containing an Alloxysta (pre-) pupa. Similarly, ovipositing in parasitised aphids would partially secure the host and could therefore provide a competitive advantage to S. aphidivorus over species attacking aphid mummies.

It is unclear whether the atypical dual oviposition behaviour of S. aphidivorus , as well as its fitness consequences, as observed here can be extrapolated to predict patterns of host use and interactions with competitors and natural enemies under field conditions. Syrphophagous aphidivorus is ubiquitous in many agricultural and natural systems (Sullivan and van den Bosch 1971, Mertins 1985, Völkl and Barczak 1990). This ubiquity might partly be due to its capacity to attack both parasitised aphid and mummy hosts.