Herbivore-induced plant volatiles as cues for habitat assessment by a foraging parasitoid

Authors

  • CEDRIC TENTELIER,

    1. Institut National de la Recherche Agronomique, UMR 1112 ‘Réponse des Organismes aux Stress Environnementaux’, 400, Routes des Chappes, BP 167, 06903 Sophia Antipolis Cedex, France
    Search for more papers by this author
  • XAVIER FAUVERGUE

    1. Institut National de la Recherche Agronomique, UMR 1112 ‘Réponse des Organismes aux Stress Environnementaux’, 400, Routes des Chappes, BP 167, 06903 Sophia Antipolis Cedex, France
    Search for more papers by this author

Cédric Tentelier, Department of Biological and Environmental Sciences, University of Helsinki, PO Box 65, FIN-0014 Helsinki, Finland. Tel.: +358 9 19157697; Fax: +358 9 19157694; E-mail: cedric.tentelier@helsinki.fi

Summary

  • 1Animals usually require information about the current state of their habitat to optimize their behaviour. For this, they can use a learning process through which their estimate is continually updated according to the cues they perceive. Identifying these cues is a long-standing but still inveterate challenge for ecologists.
  • 2The use of plant cues by aphid parasitoids for the assessment of habitat profitability and the adaptation of patch exploitation was studied. Grounding on predictions from optimal foraging theory, we tested whether parasitoids exploited host patches less intensively after visiting heavily infested plants than after visiting plants bearing fewer aphids.
  • 3As predicted, after visiting heavily infested plants parasitoids reduced their residence time and attacked fewer hosts in the next patch. This was the case regardless of whether the aphids were actually present on the first plant, indicating that the cue came from the plant. Moreover, the level of infestation of a plant at some distance from the first plant visited affected parasitoid patch exploitation on the second plant in a similar manner, indicating that the cue was volatile.
  • 4These results highlight a novel role of herbivore-induced volatiles in parasitoid foraging behaviour, different from the widely studied attraction at a distance.

Introduction

Animals usually require information about the state of their environment to take adaptive decisions. For example, to optimize the exploitation of a resource patch individuals must gather information about the quality of the other patches available in the habitat (Charnov 1976). This issue of information and patch use occupies a central place in animal ecology, and the proximate cues used by individuals to gain information from patchy habitats have been studied in a wide variety of animals. Foragers assess patch quality on the basis of diverse cues, including direct information from the resource itself (e.g. ducks feeling benthic seeds with their bill: Klaassen, Nolet & de Fouw 2006; seabirds smelling krill: Nevitt 1999; hessian flies seeing host plants: Withers & Harris 1996), information from resource by-products (e.g. parasitoids using aphid honeydew or caterpillar mandibular secretion: Waage 1978; Shaltiel & Ayal 1998) and public information (e.g. birds assessing the reproductive productivity of other birds or hearing their food-related calls: Valone 1996; Doligez, Danchin & Clobert 2002). Parasitoid females foraging for patchily distributed hosts are a particularly appropriate study system, because they conform to a central assumption of most optimal foraging models: a positive correlation between the long-term rate of gain (i.e. the mean number of hosts attacked per unit of time) and fitness (Stephens & Krebs 1986).

The Marginal Value theorem is a key theoretical model combining information and patch use (Charnov 1976). According to this model, an optimal forager should leave a resource patch when the depletion of that patch is such that the instantaneous rate of gain in the patch falls to the maximal average rate of gain for the habitat as a whole. Consequently, optimal residence time on the patch increases with (1) increasing profitability of the local patch and (2) decreasing profitability of the overall habitat. This second point will be referred to here as the ‘relative profitability rule’: although the absolute profitability of a given patch is a fixed characteristic, the forager is likely to estimate the relative profitability of the patch to be higher in poor habitats than in rich habitats. Foragers are therefore likely to exploit patches of the same absolute profitability more intensively if those patches are encountered in a poor habitat than if they are encountered in a rich habitat.

A forager may assess current habitat profitability, based on cues it has perceived in the past, through a learning process (Ollason 1980; McNamara & Houston 1985; Bernstein, Kacelnik & Krebs 1988; McNamara, Green & Olsson 2006; Valone 2006). The forager samples cues related to patch profitability, and uses the information provided by this local sample to update its estimate of overall habitat profitability. The combination of this learning process with the relative profitability rule leads to the following prediction: foragers will exploit a patch more intensively after having perceived cues indicative of a poor habitat than after having perceived cues indicative of a rich habitat. Recent experimental results for female parasitoids are consistent with the above-mentioned prediction: parasitoids remain longer on a patch if they previously visited a poor patch than if they previously visited a rich patch (Visser, van Alphen & Nell 1992; Hoffmeister et al. 2000; van Baaren, Boivin & Outreman 2005; Tentelier, Desouhant & Fauvergue 2006). Most of the information known to be used by parasitoids to adapt patch exploitation comes from hosts, parasitoid competitors or travel time between patches. Herbivore-induced plant responses, despite their widely acknowledged function in patch location (reviews in Dicke & Vet 1999; Turlings & Wäckers 2004), have been completely neglected in this context.

The hypothesis tested in this study was that parasitoids also use herbivore-induced responses to assess habitat profitability and adapt patch residence time. This hypothesis is supported by a number of independent observations. First, induced plant responses often provide reliable information about the identity of the herbivores concerned (Dicke 1999). Secondly, at least in some systems, the amounts of volatile compounds emitted by the plant are correlated with the density of phytophagous insects feeding on that plant (Du et al. 1998; Geervliet et al. 1998; Schmelz et al. 2003). These volatiles affect the flying behaviour of parasitoids, which are attracted towards the most profitable plants − those bearing the largest number of hosts (Geervliet et al. 1998). Thirdly, parasitoids are physiologically able to process the information conveyed by plant volatiles. In many parasitoid species, females can memorize plant odours associated with the presence of hosts, and are attracted preferentially to plants on which they have already found suitable hosts (reviews in Vet, Lewis & Cardé 1995; Steidle & van Loon 2003). The mechanisms involved in this associative learning probably differ from the updating of habitat assessment, but these results show none the less that parasitoids can adapt their behaviour according to past experience in the perception of plant cues. As assumed in assessment updating, information is dynamic, with the possibility of modifying initial odour preference if a new odour is found to be associated with the presence of hosts (Dejong & Kaiser 1992; Daza-Bustamante et al. 2002).

This evidence led to the hypothesis that parasitoids assess the profitability of plants by evaluating the level of the herbivore-induced response, and use this information to update their estimate of habitat profitability and to adapt patch exploitation. Again, combined with the relative profitability rule, this hypothesis leads to the prediction that a parasitoid which initially perceived high levels of herbivore-induced compounds will spend less time on a subsequent host patch than an individual which initially perceived lower levels of herbivore-induced compounds. This prediction was tested in a tritrophic system comprising cucumber plants, aphids and aphid parasitoids. The level of infestation and the presence of aphids on the first plant visited by parasitoids were manipulated independently, and the patch residence time of the parasitoid on the next plant visited was quantified. As predicted, after visiting heavily infested plants parasitoids reduced their residence time and attacked fewer hosts in the next patch. This was the case whether or not the aphids were actually present on the first plant, indicating that the cue came from the plant. Moreover, the level of infestation of a plant at some distance from the first plant visited had a similar effect on parasitoid patch exploitation on the second plant, indicating that the cue was volatile. This is the first study to show that parasitoids use herbivore-induced volatiles to adapt patch residence time from plant to plant.

Methods

The tritrophic system used in this study consisted of the cucumber plant Cucumis sativa L. (Cucurbitales: Cucurbitaceae), the aphid Aphis gossypii Glover (Homoptera: Aphididae) and the parasitoid Lysiphlebus testaceipes Cresson (Hymenoptera: Braconidae). L. testaceipes is a solitary parasitoid of aphids, including A. gossypii, on cucumber (Pike et al. 2000). There is evidence to suggest that L. testaceipes females use herbivore-induced plant responses to estimate habitat profitability. First, when foraging on a single patch, females use plant responses to assess the profitability of the patch; their residence time on an aphid colony increases with both the number of aphids in the colony and the response of the plant carrying the colony (Tentelier, Wajnberg & Fauvergue 2005). Secondly, when foraging on more than one patch, L. testaceipes females adapt their residence time in a given patch according to the profitability of previously visited patches, in a manner consistent with the relative profitability rule and updating of habitat assessment (Fauvergue et al. 2006; Tentelier et al. 2006). These two results suggest that the assessment of habitat profitability based on previously visited patches may involve the sampling of plant responses in these patches.

The parasitoids used in the experiments were young (< 12 h) mated L. testaceipes females. They originated from a strain founded with 10 individuals captured at emergence from mummified A. gossypii collected on hibiscus in Lisbon, Portugal. Before this study, this parasitoid strain was mass-reared for about 200 generations on A. gossypii as a host insect and on cucumber (C. sativa var. Carmen) as the host plant. Individuals may have been exposed to plant odours during immature stages (e.g. Gandolfi, Mattiacci & Dorn 2003), but the tested adults were kept naive by isolation before emergence (at the mummy stage of the parasitized host), with one parasitoid per glass vial. After emergence, adults were sexed and each female was housed with a male for 1 h for mating. Females were then kept alone in their vials and fed with honey.

Aphid colonies were reared in a greenhouse on 4-week-old cucumber plants (C. sativa var. Serit F1). At this age, the plants had two fully developed leaves. The level of herbivore-induced volatiles emitted by plants was manipulated by varying the duration of plant infestation with aphids. For this, a young (< 48 h) adult A. gossypii (strain ‘Antibes SF2’) female was placed on one randomly chosen leaf and allowed to feed and reproduce for the required length of time. The duration of infestation therefore covaried with variables other than plant response, such as the number of aphid larvae or the amount of honeydew. In our experiments, infestations of 1, 3 and 7 days resulted in 10·1 (9·6–10·7), 20·1 (18·8–21·5) and 49·0 (42·5–55·5) aphids on the plant, respectively (mean and 95% confidence intervals). Infestations were planned so that all plants were of identical age at the time of testing.

Two experiments were carried out to test whether female L. testaceipes adapted their patch residence time on a plant according to the level of herbivore-induced volatiles perceived on a previously visited plant. The aim of the first experiment was to test if the parasitoids need direct contact with aphids or aphid products to assess the profitability of the first-visited plant. The aim of the second experiment was to test if the parasitoids need direct contact with the plant to assess its profitability. The first plant visited is referred to as the ‘sample plant’ and the second plant as the ‘test plant’. Each experiment consisted of two successive steps (Fig. 1). In the first step − the sample step − parasitoids were allowed to gain experience on a leaf of the sample plant on which the level of herbivore-induced volatiles and the presence of aphids were manipulated independently. In the second step − the test step − the patch use behaviour of the parasitoids on the aphid patch supported by the test plant was quantified. For the sample step, parasitoids were exposed individually to a leaf of the sample plant for 1 h. The tested parasitoid was forced to remain on the chosen leaf by enclosing the wasp and the leaf between two sheets of Plexiglas sealed with a 1-cm-thick rubber joint; 1 h later the parasitoid was recaptured in a glass vial and released on the patch of the test plant, where its total number of attacks on aphids and patch residence time were recorded as in Tentelier et al. (2005, 2006). This test patch was constant in all respects. The duration of infestation of the test plant was always 3 days, so that the number of aphids on the patch (i.e. the absolute profitability of the patch) and the presumed amounts of herbivore-induced volatiles emitted by the plant were similar for all levels of treatment.

Figure 1.

Experimental set-up used in the two experiments. Experiments were organized in two successive steps: the sample step and the test step. In the sample step, parasitoids were restricted to one leaf of the sample plant. In the test step, the foraging behaviour of the parasitoids was quantified on the patch carried by the test plant. The level of emission of herbivore-induced volatiles of sample plants was manipulated by changing the duration of aphid infestation, which is correlated with the number of aphids on the aphid-infested leaf. (a) In the first experiment, the two factors manipulated were (1) the duration of infestation of the sample plant and (2) restriction of the parasitoids to an aphid-infested or an aphid-free leaf of the sample plant. (b) In the second experiment, the three factors manipulated were (1) the duration of infestation of the sample plant, (2) the presence/absence of aphids on the aphid-infested leaf of the sample plant (droplets indicate that aphid products were washed from the aphid-infested leaf), and (3) the duration of infestation of another plant in the vicinity of the sample plant.

The development of an aphid colony may induce a plant response, as hypothesized, but it is also associated with other potential cues for parasitoids, such as the presence of aphids and their by-products. To demonstrate that parasitoids use the plant response to estimate habitat profitability, it is necessary to separate the effects of plant response and the effects of direct contact with hosts on parasitoid behaviour. In the first experiment, the herbivore-induced plant response and the opportunity for parasitoids to come into contact with aphids and aphid products were manipulated independently in a 4 × 2 factorial design (Fig. 1a). Twenty replicates were carried out for each treatment level. The duration of infestation, used to manipulate the level of herbivore-induced response, was zero (no aphid placed on the plant), 1, 3 or 7 days. The opportunity for parasitoids to come into contact with aphids and aphid products was manipulated by individually exposing parasitoids to aphid-infested leaves or aphid-free leaves from the sample plant. Assuming that the plant response is systemic (as in Guerrieri et al. 1999), parasitoids in both levels of treatment were exposed to the plant response, but only those exposed to infested leaves came into direct contact with aphids and aphid products. If the plant response is the cue sampled by parasitoids, then patch residence time during the test step should decrease with increasing duration of infestation of the sample plant, regardless of whether parasitoids were exposed to aphid-infested leaves or aphid-free leaves. One potential bias in such an experiment relates to the possibility that parasitoids exposed to the aphid-infested leaf during the sample step had lower egg loads, resulting in less intensive patch exploitation. Egg load was therefore investigated to determine whether it differed according to treatment and whether it affected the behaviour of parasitoids on the test plant. Egg load on arrival on the test plant was estimated by dissecting the parasitoids at the end of the test step and adding the number of mature eggs remaining in the ovaries to the number of attacks observed on the test plant. Mature eggs were identified by their typical lemon-like shape (Le Ralec 1991).

Another possible bias with such an experiment is that parasitoids experienced to the aphid-free leaves might have been exposed to herbivore-induced volatiles as well as volatiles from aphids and their by-product from the infested leaf. A second experiment was designed to untangle this uncertainty. In this experiment, parasitoids were exposed systematically to the aphid-free leaf of the sample plant, to prevent any contact with aphids and aphid products. Three factors were tested: (1) the duration of infestation of the sample plant; (2) the presence/absence of aphids and aphid products on the aphid-infested leaf of the sample plant; and (3) the duration of infestation of a neighbouring plant in the vicinity of the sample plant (Fig. 1b). As in the first experiment, the duration of infestation of the sample plant was used to manipulate the level of herbivore-induced plant response. Durations of 1 or 7 days were used. The presence/absence of aphids and aphid products on the aphid-infested leaf of the sample plant was manipulated to determine whether the differences in behaviour observed in the first experiment were due to plant cues or volatile cues from aphids. If parasitoids use the herbivore-induced plant response as the cue for assessing habitat profitability, then patch residence time on the test plant should decrease with increasing duration of infestation of the sample plant, regardless of the presence or absence of aphids on that plant. If parasitoids use volatile aphid products as a cue, then this effect should occur only when aphids are present on the sample plant. The presence/absence of aphids and their products on the sample plant was manipulated by either leaving the aphids on the infested leaf or gently wiping them off the leaf with forceps and rinsing the leaf for 30 s with distilled water, as described by Ohara et al. (2003; see also Henneberry et al. 2000). In addition, a plant infested for 1 or 7 days was placed 20 cm away from the sample plant. If the cue sampled by parasitoids is volatile, then patch residence time on the test plant should also decrease with increasing duration of infestation of the neighbouring plant. A 2 × 2 factorial design was used for the duration of infestation and the presence/absence of aphids on the sample plant. Twenty-two replicates were carried out for each level of these factors. For each replicate, the duration of infestation of the neighbouring plant was completely randomized. In this experiment parasitoids did not lay eggs during the sample step. Egg load was therefore not affected by treatment level and was not measured.

Patch residence times on the test plant were analysed with generalized linear models, based on a log-link function and a gamma distribution. The validity of such models was checked by plotting residuals against predicted values. The significance of main effects and interactions was tested using log-likelihood ratio tests. For the first experiment, egg load was analysed with a generalized linear model based on a normal distribution and was also introduced, as a continuous covariate, in the models analysing patch residence time. Analyses were carried out with the genmod procedure in sas version 8·0 (SAS Institute Inc. 1999). The correspondence between the patch residence times obtained and patch exploitation was verified by assessing the correlation between log-transformed data for patch residence time and the total number of attacks on the patch. This was achieved with the corr procedure in sas.

Results

The results of the first experiment show clearly that parasitoids were able to assess the profitability of the sample plant without coming into contact with aphids or aphid products. Consistent with the hypothesis that parasitoids use host-induced plant responses to assess habitat profitability, patch residence time on the test plant decreased with increasing duration of infestation of the sample plant (d.f. = 3; χ2 = 26·56; P < 0·0001; Fig. 2a). This was true whether parasitoids were exposed to aphid-infested or aphid-free leaves from the sample plant (main effect of presence of aphids on the leaf: d.f. = 1; χ2 = 0·76; P = 0·3842; in interaction with plant response: d.f. = 3; χ2 = 1·45; P < 0·693). The effect of the duration of infestation could not be accounted for by differences in egg load following exposure to colonies of different sizes, because egg load on arrival on the test plant was similar for all treatment levels (Table 1). Egg load was not affected by the duration of infestation of the sample plant (d.f. = 1; χ2 = 1·19; P = 0·2760), contact with aphids on the sample plant (d.f. = 1; χ2 = 0·0; P = 0·9958) or by the interaction between these two variables (d.f. = 1; χ2 = 0·14; P = 0·7066). This absence of effect may be due to the relatively small number of eggs laid in the sample step with respect to total egg load. When included as a covariate in the generalized linear model, larger egg load did not affect patch residence time (d.f. = 1; χ2 = 0·8; P = 0·7822). As assumed, the total number of attacks was correlated positively with patch residence time (Fig. 3a; Pearson's correlation coefficient r = 0·73; P < 0·0001), indicating that patch residence time was a good indicator of patch exploitation.

Figure 2.

Mean patch residence time (± standard error) on the test plant as a function of duration of infestation of the sample plant. (a) First experiment: closed circles correspond to parasitoids exposed to aphid-infested leaves, and open circles correspond to parasitoids exposed to aphid-free leaves. Different letters indicate significant differences in pairwise comparisons with sequential Bonferroni correction. (b) Second experiment: closed circles correspond to parasitoids visiting a sample plant with aphids, and open circles correspond to parasitoids visiting a sample plant without aphids. Dashed lines indicate a neighbouring plant infested for 1 day. Solid lines indicate a neighbouring plant infested for 7 days. See text for the significance of the various factors.

Table 1.  Estimated number of mature eggs remaining in parasitoid ovaries at the end of the sample step for each treatment level in the first experiment. The first row indicates the duration of infestation of the sample plant; the first column indicates whether parasitoids were placed on the aphid-infested or aphid-free leaf of the sample plant. Numbers indicate the mean number of eggs (± standard error). The number of eggs did not differ across treatments
 0 day1 day3 days7 days
Aphid-infested187 ± 19209 ± 22198 ± 25220 ± 21
Aphid-free178 ± 18221 ± 18194 ± 8205 ± 19
Figure 3.

Total number of attacks vs. patch residence time on the test plant for (a) the first experiment and (b) the second experiment.

The results of the second experiment indicate that parasitoids assessed the profitability of the sample plant on the basis of a volatile response of the plant to aphid infestation. Consistent with the results of the first experiment and our predictions, longer durations of infestation of the sample plant resulted in shorter patch residence times on the test plant (d.f. = 1; χ2 = 39·77; P < 0·0001; Fig. 2b). This effect did not depend on the presence of aphids on the sample plant (main effect: d.f. = 1; χ2 = 0·27; P = 0·6022; in interaction with the duration of infestation: d.f. = 1; χ2 = 0·24; P = 0·6222). Thus, the cue sampled by parasitoids originated from the plant only. Furthermore, the duration of infestation of the neighbouring plant also affected the parasitoids’ patch residence time on the test plant (d.f. = 1; χ2 = 6·26; P = 0·0123; Fig. 2b). Thus, the cue sampled by parasitoids is volatile, because parasitoids had no contact with the neighbouring plant, and could certainly not see the aphids on it. It could be argued that the effect of the neighbouring plant was due to direct cues from the aphids on that plant. However, the effect of the sample plant demonstrated clearly that the cue perceived by the parasitoids came from the plant rather than from the aphids. It would therefore seem unlikely that the effect of one plant was due to herbivore-induced plant volatiles whereas that of the other was due to the direct products of aphids. There was no interaction between the duration of infestation of the sample plant and that of the neighbouring plant (d.f. = 1; χ2 = 0·04; P = 0·8382), indicating that the responses of the two plants had additive effects on patch residence time. This suggests that parasitoids summed all the herbivore-induced volatile compounds perceived during the sample step to assess habitat profitability. As in the first experiment, the total number of attacks was positively correlated with patch residence time (Fig. 3b; Pearson's correlation coefficient r = 0·80; P < 0·0001), again indicating that patch residence time is a good indicator of patch exploitation. The results from this second experiment clearly demonstrate that the cue used by parasitoids to adapt their patch residence time on the test plant was the herbivore-induced volatile response of the sample plant, and not volatile cues emanating from aphids.

Discussion

This study reports the use of herbivore-induced volatile compounds by parasitoids for the adaptation of foraging behaviour. So far, the use of herbivore-induced volatile compounds by parasitoids has been studied almost exclusively in the context of patch location (reviews in Dicke & Vet 1999; Turlings & Wäckers 2004). The findings presented here show that wasps adjust their patch exploitation behaviour on a plant according to the herbivore-induced volatiles emitted by the plants they have visited previously, or by other plants in the vicinity of those plants. For the parasitoids, this behaviour is adaptive because it enables them to exploit the host patches encountered according to the profitability of the overall habitat, as assumed by optimal foraging models (Stephens & Krebs 1986). For the plants, this parasitoid behaviour is a source of additional selection pressure, favouring the emission of large amounts of herbivore-induced volatiles: a plant emitting more synomones than its counterparts is likely to attract more parasitoids and to retain them for relatively longer, thereby increasing the number of phytophagous insects attacked on that plant.

The mechanism by which parasitoids adjust current patch use based on previous experience in terms of levels of herbivore-induced volatiles may involve habituation to chemical compounds, as suggested by Shettleworth (1998) and Thiel & Hoffmeister (2004). After exploiting a patch a parasitoid is habituated to host-related cues, the level of habituation increasing with the intensity of the cues. The motivation driving the parasitoid to exploit a subsequent patch then depends on resensitization to these cues. If a parasitoid has most recently visited a rich patch, its level of habituation is high on leaving this patch, resulting in it spending less time in the next patch. This process may account for our observations. Moreover, if the travel time between patches is long, parasitoids may become resensitized to the cues, resulting in more thorough exploitation of the next patch, as suggested by Thiel & Hoffmeister and observed in several parasitoids (Thiel & Hoffmeister 2004; Tentelier et al. 2006; Thiel, Driessen & Hoffmeister 2006). This mechanism is very different from associative learning, which is involved in the use of plant volatiles for host location (Turlings et al. 1993). In our study, associative learning by adults was probably not possible, because the parasitoids were isolated in glass vials at the mummy stage. Only individuals exposed to aphid-infested leaves in the first experiment had the opportunity to associate plant odour with oviposition during the sample step, but these individuals behaved similarly to individuals exposed to aphid-free leaves. This suggests that the wasps associated plant responses with the presence of aphids either innately or during immature stages (as in van Emden et al. 1996 or Gandolfi et al. 2003).

Why do parasitoids use herbivore-induced volatile compounds rather than direct encounters with aphids to assess habitat profitability? Indeed, the volatile compounds emitted by whole plants are indirect cues and are probably less reliable than direct host sampling. For example, the quantitative relationship between the number of hosts and the amounts of volatiles may be affected by physiological constraints on volatile compound production metabolism, or the presence of other biotic or abiotic stresses (Takabayashi, Dicke & Posthumus 1994; Takabayashi et al. 2006). However, as natural selection favours the discretion of herbivores and the conspicuousness of plant signals to parasitoids, plant signals can be detected much more easily by foraging parasitoids (Vet 1999). This high detectability allows parasitoids to limit the costs of information acquisition, providing a trade-off with low levels of reliability (Koops 2004). Cue detectability has not been considered in most theoretical models of habitat assessment updating, which assume that information is acquired by direct sampling in patches (Krebs, Kacelnik & Taylor 1978; Ollason 1980; McNamara & Houston 1985; Bernstein et al. 1988). However, many animals clearly use indirect cues to assess habitat quality. The best-documented examples, other than the use of volatile plant compounds by parasitoids, relate to the use of public information by animals as a means of assessing the quality of nesting sites (Doligez et al. 2002) and of the use of environmental variables by prey species to assess predation risk (Kortet & Hedrick 2004; Orrock, Danielson & Brinkerhoff 2004; Watson, Mathis & Thompson 2004). In this second example, the cost of acquiring direct information is very high, as it implies encounters with predators and, possibly, death. Foragers that spend time sampling patches may incur less extreme costs, but these costs may none the less be high enough to select for the use of indirect cues.

Acknowledgements

We thank Karine Malvaux for technical assistance, Denis Bourguet, Thomas Hoffmeister, Corinne Vacher, Fabrice Vavre and an anonymous referee for their comments on the manuscript, Nathan Fauvergue for his innovative view of tritrophic relations, and the Département Santé des Plantes et Environnement, Institut National de la Recherche Agronomique and Région Provence Alpes Côte d’Azur for financial assistance (grant no. 2002–12324 allocated to X. F. and C. T. and grant 2004-1112-01 allocated to X. F.).

Ancillary