Hyperparasitoid colonies were maintained by exposing potato plants, infested with mummified aphids (for A. suspensus , D. carpenteri and S. aphidivorus ) or live parasitised aphids (for A. victrix ) to the hyperparasitoid females. Colonies were held in the laboratory at room temperature under a 16L:8D photoperiod.

For both experiments, hyperparasitised mummies were individually collected in the rearing colonies, and kept as groups of 100 mummies in a cage with a vial of sugar water as a food source at 20±1ºC, 75±10% RH, under a 16L:8D photoperiod. Males were added to ensure that at emergence females had access to potential mates. From these cages 1-6 day old females were taken for use in the bioassays. As these hyperparasitoids live more than 1 month under these experimental conditions (Christiansen-Weniger, 1992; Chow and Mackauer, 1996; R. Buitenhuis, unpublished data), females were not time-limited.

To obtain parasitised aphids and mummies for biossays, third-instar aphid nymphs were exposed to parasitism by 3-5 days old mated A. nigripes females for a 24-hr period. Presumably parasitised aphids were then reared at 20±1ºC, 75±10% RH, under a 16L:8D photoperiod. Based on embryonic and larval developmental times of A. nigripes at 20ºC (Paré et al. 1979), third instar larvae in living aphids and prepupae in mummified aphids, were obtained five and eight days following parasitisation, respectively. In the text these hosts will be referred to as parasitised aphids and mummies.

To ensure the functionality of the olfactometer, two types of pre-tests were done. When both arms carried clean air only, hyperparasitoids ( A. victrix , A. suspensus and S. aphidivorus ) chose each of them at the same frequency (χ2 test, for all species, p>0.05, n >20). In the second pre-test, males of the primary parasitoid A. nigripes chose significantly more often the arm of the olfactometer with conspecific virgin females, against clean air in the other arm (χ2=7.6190, p=0.0058, n=26 males).

Treatments . Treatments were chosen according to the quantities and concentrations that were shown to be attractive to primary parasitoids and hyperparasitoids (Read et al ., 1970; Bouchard and Cloutier, 1985; Siri, 1993).

(1) Single cues originating from all trophic levels.

From the first trophic level, we tested a clean potato seedling (Norland variety). A 15 cm high plant was washed, air dried, cut and immersed in water sealed with Parafilm© to exclude possible interference of volatiles from the cut edges. From the second trophic level, we tested potato aphids. One hundred aphids of all stages were collected in a gauze-covered container. In addition we tested honeydew that was collected as described by Bouchard and Cloutier (1984) (40 mg dried honeydew dissolved in 150 μl distilled water). Finally, from the third trophic level, we tested parasitised aphids, mummies or female A. nigripes . For these treatments either 100 4-5 days parasitised aphids, 100 newly (0-24 h) mummified aphids, or six 1-5 days old virgin A. nigripes females were collected in a gauze covered container. Odours were tested in single choice tests against air (pumped through activated charcoal and humidified). A dual-choice test was performed for S. aphidivorus to determine preference for mummies vs. parasitised aphids.

(2) Complex cues.

Aphid and possible aphid-induced plant volatiles were tested with a potato seedling infested with 50 potato aphids two days before the test. The attraction of the whole plant-host complex was tested with a potato seedling infested for two days with 25 healthy aphids, 25 parasitised aphids and 25 mummies, obtained as previously described. Mummies were glued on the leaves with non-toxic Lepage© white glue before the experiment. To exclude the possibility that hyperparasitoids were attracted to uninfested plant odours, the plant-host complex was tested in a dual choice test against a clean plant (washed and air dried potato seedling).

Bioassay . Mated 1-6 days old hyperparasitoid females were given an oviposition experience of 24 hours the day before the test with ten mummies and five live parasitised aphids on a potato leaf to standardise their searching and parasitising experience before the test. The females were individually released in the Y-tube, and used only once. After five minutes the position of the female was recorded. This duration was shown to be sufficient for the majority of the females to make a choice. If a female was found more than 15 cm into one of the arms of the olfactometer, this was recorded as a choice. Females recovered before this point, and at or before the intersection of the olfactometer arms were not considered to have made a choice. Effectively, in the experiment, females were either found at the end of the tube, or at the intersection. The Y-tube and the containers for the odour sources were washed with hot water and acetone and air-dried between each treatment. For each experiment (single and dual choice), all treatments were tested in a random order in a two-day period. In each treatment, five females per hyperparasitoid species were tested in a random order. This was repeated eight times for a total of 40 females per species per treatment.

Treatments . Each plant was randomly allocated to one of the following treatments: control (uncontaminated plant), aphids (plant infested with 100 aphids for two days), honeydew (plant infested with 100 aphids for two days after which aphids and exuviae were removed with a paintbrush before the experiment), aphids + parasitised aphids (PA) (plant infested for two days with 50 aphids and 50 parasitised aphids) and plant-host complex (PHC) (aphids + parasitised aphids, and two mummies glued on the underside of leaves 4, 6 and 8). Parasitised aphids were marked on the abdomen with a non-toxic marker (Sharpie©) to distinguish them from unparasitised aphids during observations. This did not seem to disturb the aphids or to change their behaviour.

Bioassay . Mated 1-6 days old females were given an oviposition experience of 24 hours the day before the test, individually in cages with a potato leaf with hosts (for A. suspensus and D. carpenteri two mummies; S. aphidivorus two mummies and two parasitised aphids; A. victrix two parasitised aphids), before being used in the experiments.

At the beginning of a test, one female was released from a gelatine capsule on the upper side of leaf number 4. Her behaviour was observed with the Observer© (Noldus, 1997, version 3 for Macintosh) for one hour, or until the female left the plant for more than 5 seconds. One plant was used for one female of each hyperparasitoid species. Hosts that were parasitised by a hyperparasitoid female were replaced after each observation.

The duration of the following behaviours was recorded: walking, resting, grooming, feeding, flying, examining (aphid, parasitised aphid or mummy) and ovipositing (aphid, parasitised aphid or mummy). Furthermore the position of the female was recorded continuously by noting the leaf number and plant part (upper- or under side of the leaf, petiole or stem).

The order in which the hyperparasitoid species were tested was randomised within treatments. Ten females of each species were tested per treatment. Because the treatment with the parasitised aphids and the plant-host complex were the same for A. victrix , this species was not tested on the plant with parasitised aphids but only on the plant-host complex.

From the timetable that was created by the Observer© the following parameters were calculated: The total visit time was defined as the time spent on the plant from release to departure. The search time was defined as the time spent walking. The search time was subdivided between time spent on the upper and lower surface of the leaves. Aphid hyperparasitoids have long handling times of several minutes per host (Sullivan, 1987). The handling time was defined as the total time that a female spent examining and parasitising hosts during the visit. Finally, the number of different leaves that were visited was calculated.

Only females that had come in contact with the offered stimuli were used in the analysis. Also, observations where females immediately left the plant after aphid defence were discarded from the analysis because these did not represent a comparable visit (max. 2 cases out of 10 for S. aphidivorus where the aphid kicked and caused the female to fly up).

Because of the presence of censored data in the foraging behaviour, the visit time and search time, these data were analysed per species using the LIFEREG procedure, using a log-normal error function. The number of leaves visited was analysed with General Linear Models (GENMOD) using a Poisson error function. The time spent on the upper- and under side of the leaves was compared with a paired t-test.

For all analyses, the level of significance was α=0.05 and all data were analysed using SAS (SAS, 1999).

The total visit time (Fig. 4-2a) was divided in three categories of behaviours (Fig. 4-2b-d): search time, time spent with hosts (examining and ovipositing into parasitised aphids or mummies), and other behaviours (resting, grooming, flying, feeding and examining healthy aphids).

The search time was influenced by different stimuli for all hyperparasitoid species (LIFEREG A. victrix : χ2=16.0711, df=3, p=0.0011; A. suspensus : χ2=16.0553, df=4, p=0.0029; D. carpenteri : χ2=38.9377, df=4, p<0.0001; S. aphidivorus : χ2=36.8214, df=4, p<0.0001) (Fig. 4-2b). Female A. victrix searched longer on plants with aphids and on the plant-host complex than on the other plant treatments. The other three species searched longer on all treatments as compared to the control. Female D. carpenteri searched the longest time on plants with honeydew and the plant–host complex. Search time of S. aphidivorus females was significantly longer on plants with parasitised aphids as compared to the other treatments.

The long total visit times of females of the three mummy-attacking hyperparasitoids on plants with their hosts were actually caused by the time spent with mummies (Fig. 4-2c). Asaphes suspensus spent 60 ± 4% (mean ± SE) of the total visit time examining and parasitising mummies, D. carpenteri 31 ± 21%, and S. aphidivorus 42 ± 18%. The time required to parasitise a mummy was very long ( A. suspensus 1888 ± 204 s, D. carpenteri 462 ± 312 s and S. aphidivorus 390 ± 138 s). In contrast, time spent with parasitised aphids did not take such a substantial proportion of total visit time (Fig. 4-2c). Alloxysta victrix spent time parasitising its hosts (larvae within live aphids) for only 7 ± 9% of the total visit time, and S. aphidivorus, 10 ± 15% (PA) and 2 ± 2% (PHC). The time required to parasitise a host within a parasitised aphid was only 96 ± 18 s for A. victrix, and 102 ± 48 s for S. aphidivorus .

When hosts were present, a significant proportion of females stayed on the plant for the whole duration of the experiment (1 hour; A. suspensus 100%, A. victrix 20%, D. carpenteri 70%, S. aphidivorus (PA) 71%, S. aphidivorus (PHC) 43%). It is likely that in these cases visit and search time would have been longer if the experiment would have been permitted to last longer.

The number of leaves that was visited was generally small as compared to the number of leaves available (10). There were differences in the number of leaves visited between species, and between some treatments (2-way GENMOD, treatment χ2 = 11.43, df = 4, p=0.0221, species, χ2 = 41.48, df = 3, p<0.0001, treatment*species χ2 = 30.87, df = 11, p=0.0012) (Table 4-1). In general, D. carpenteri visited more leaves than the other three species. For each species, the differences in number of leaves visited between the treatments are similar to the results for search time. Asaphes suspensus visited an equal number of leaves in each treatment, A. victrix visited more leaves on the plant with aphids, D. carpenteri visited more leaves on the honeydew, parasitised aphid and plant-host complex treatments and S. aphidivorus visited more leaves on the parasitised aphid and plant-host complex treatments as compared to the control.

After release, female hyperparasitoids explored the plant mainly by walking. Only occasionally were females observed to use short flights to move between the leaves (1.1 ± 0.2 SE flights·female-1·observation-1). Females searched both sides of leaves, often alternating rapidly between the upper and under sides. The time allocated to searching on the upper and lower surfaces of leaves did not differ significantly for any species or treatment, except for A. victrix (Table 4-1), which searched longer on the upper than the lower surface of leaves on honeydew-contaminated plants (paired t-test t8=4.59, p=0.0018). When visiting different leaves, A. suspensus moved slightly upward on the plant in all treatments. Alloxysta victrix always moved to the highest leaves before taking off. Dendrocerus carpenteri and S. aphidivorus moved up and down on the plant without a clear pattern.

Although we cannot exclude the doubt that our olfactometer set-up was not functional for aphid hyperparasitoids, several arguments imply that our results are valid. First, a similar set-up has been used successfully for aphid hyperparasitoids before (Read et al ., 1970; Singh and Shrivastava, 1987a/b; Siri, 1993). Second, the pre-tests showed that the set-up was functional for primary parasitoids. Male A. nigripes was attracted to the odour of conspecific virgin females. Third, we obtained one positive response of S. aphidivorus , that was attracted to the odour of live parasitised aphids vs. aphid mummies. However, we cannot explain why S. aphidivorus preferred the odour of live parasitised aphids to that of aphid mummies in the dual choice test, while in the single choice test it was neither attracted nor repelled by any of these odour sources. Finally, attempts to test the hyperparasitoid species in a windtunnel with the same odour sources did not succeed, because females would not fly, even at low windspeeds.

Other studies, with similar set-ups, report varying results. Alloxysta fuscicornis (= Charips brassicae ) was attracted to female primary parasitoids, but not to plant or aphid odours (Read et al ., 1970). On the other hand, Alloxysta pleuralis is attracted to volatiles from various plants (Singh and Shrivastava, 1987a/b). Furthermore, A. victrix was attracted to herbivore-induced volatiles and a synthetic aphid alarm pheromone, and D. carpenteri was attracted to herbivore-induced volatiles, conspecific females and mummies, but neither species reacted to aphids, plants or primary parasitoid females (Siri, 1993).

The differences between these studies and our results might be explained by differences in the hyperparasitoid species that were tested, or may be due to differences in plant – aphid – primary parasitoid systems (oat – Sitobion avenae - Aphidius uzbekistanicus ) (Siri, 1993). Also, in the above-mentioned studies, a confounding effect of attraction between females that were tested simultaneously in groups of 10-30 in the olfactometer cannot be excluded. While volatiles might play a role in host searching in some hyperparasitoid species, our results suggest that they are not strong cues in aphid hyperparasitoids.

As could be expected, S. aphidivorus females spent more time searching on plants with parasitised aphids, than on plants with unparasitised aphids. However, this was not observed on plants with mummies (plant-host complex). This is curious because, of the two hosts, mummies are reported to be the preferred and most suitable one (Kanuck and Sullivan, 1992; Buitenhuis et al ., submitted). Perhaps the different proportions of parasitised aphids and mummies in the plant treatments had an influence on the females’ perception of the patch. Further study will have to point out how oviposition in one of the two hosts influences searching time in this species.

The presence of hosts did prolong visit time in most cases, an effect that would probably be even stronger if females would be permitted to stay longer than 60 minutes on the plant. This increase in visit time was due to the long handling times of mummies (6 ± 2 to 32 ± 3 minutes) for A. suspensus and D. carpenteri . Longer search following successful oviposition (success-motivated search) could not be demonstrated in this experiment, but might be if females would be observed until they left the plant.

Another potential determinant of searching behaviour is the host range. Generally, the use of cues can be transposed on a specialist-generalist continuum: from intense and specific through weak and non-specific, to the absence of cue use (Vet and Dicke, 1992). Similarly, the four tested hyperparasitoids ranged from one relatively host specific species ( A. victrix , attacking parasitoids of only one genus) to three species with a very large host range ( D. carpenteri , A. suspensus and S. aphidivorus , attacking a wide variety of genera) (van den Bosch, 1981; Höller et al ., 1993; Brodeur, 2000). However, in this study, no differences between hyperparasitoid species could be observed.

We designed this study for realistic comparison between the behaviour of aphid hyperparasitoids and the host search behaviour of the primary parasitoid A. nigripes (Bouchard and Cloutier, 1985; Cloutier and Bauduin, 1990) on the same potato – Macrosiphum euphorbiae system. Contrary to the behaviour of A. nigripes which was attracted to the odours of several aphid species and to aphid honeydew (Bouchard and Cloutier, 1985), none of the four hyperparasitoids was attracted to olfactory cues. On the other hand, there were similarities in the search behaviour on a plant of the primary parasitoid and hyperparasitoids. Aphidius nigripes showed longer residence and searching times, visited more leaves and spent more time per leaf in response to honeydew and aphids (Cloutier and Bauduin, 1990). This arrestment and search stimulation was also found in the hyperparasitoid species. Not all hyperparasitoids were arrested by honeydew, in contrast to what was found for A. nigripes . Both upper and lower leaf surfaces were searched equally by most hyperparasitoids, contrary to A. nigripes that searched more on the lower leaf surface, where it is more likely to find M. euphorbiae aphids.

In summary, our study suggests that aphid hyperparasitoids may not resemble primary parasitoids in attraction to olfactory stimuli, but it demonstrates that their behaviour on a plant shows several similarities, although this depends on the hyperparasitoid species. There are two non-exclusive explanations for differences between primary parasitoids and hyperparasitoids. First, many of the cues that are direct and reliable for primary parasitoids, are indirect cues for hyperparasitoids and therefore less reliable. First, the presence of aphids on a plant, a reliable cue for primary parasitoids, does not guarantee the presence of suitable parasitised aphids to hyperparasitoids. Secondly, compared to primary parasitoids, hyperparasitoids generally have a broader host range (Gordh, 1981; Sullivan, 1987; Sullivan and Völkl, 1999; but see van den Bosch, 1981 and Brodeur, 2000). Vet and Dicke (1992) hypothesised that contrary to specialists, the use of kairomones by generalists should be weak and non-specific, or could even be impossible because the great diversity of potentially useful chemical information would generate a physiological constraint on sensory processing, and common chemical components would be very limited. The hyperparasitoids tested here have been reported on many different plants and aphids (e.g. Gutierrez and van den Bosch, 1970; Sullivan and van den Bosch, 1971; Johnson et al ., 1979; Thiboldeaux et al ., 1987; Mertins, 1985; Höller et al ., 1993, Müller et al ., 1999). In the absence of common, detectable cues it is therefore likely that aphid hyperparasitoids search mainly in the habitat where they are born, or select a habitat at random and that search is induced by contact stimuli on the plant.