SEARCH

SEARCH BY CITATION

Keywords:

  • foraging behaviour;
  • host distribution;
  • parasitoids;
  • phenotypic plasticity;
  • time allocation

Abstract

  1. Top of page
  2. Abstract
  3. Introduction
  4. 1. Do C. glomerata and C. rubecula show interspecific differences in the way they use information to make patch leaving decisions?
  5. Materials and methods
  6. Results
  7. Discussion and Conclusions
  8. Acknowledgements
  9. References

1. We analysed the foraging behaviour of two closely related parasitoid species (Cotesia rubecula and Cotesia glomerata) with respect to leaving tendencies from patches in different environments. We investigated how intrapatch experiences like contact with feeding damage and encounters with hosts influence patch leaving decisions. We also estimated the effect of experiences in previously visited patches on leaving decisions in the present patch.

2. For this analysis we applied the proportional hazards model (Cox 1972) to data collected in three versions of a multiple patch set-up. These set-ups consisted of different host species or combinations of host species: (1) Pieris rapae, (2) Pieris brassicae and (3) both P. rapae and P. brassicae. The larvae of these hosts differ in their spatial distribution on plants: P. brassicae occur in clusters and the distribution of larvae is heterogeneous; P. rapae larvae feed solitarily.

3. The specialist parasitoid C. rubecula used a simple strategy: highest leaving tendency on empty leaves, lower leaving tendency on leaves infested with the non-preferred host P. brassicae, lowest leaving tendency on leaves infested with the preferred host, P. rapae. In the environment with both host species, the leaving tendency only decreased on leaves infested with P. rapae.

4. The generalist C. glomerata used a more complex set of rules. (a) Multiple ovipositions on the present patch decreased the leaving tendency on leaves containing the gregarious host. (b) Once the parasitoid had encountered two or more hosts, it had a lower leaving tendency during subsequent patch visits. (c) The leaving tendency increased with the number of visits on infested leaves. In environments where the less preferred host P. rapae was present, C. glomerata switched to the same simple type of rule as used by C. rubecula.

5. Neither of the two Cotesia species used a count-down rule, in which ovipositions increase the leaving tendency. We discuss how patch exploitation by both Cotesia species compares to the patch exploitation mechanisms as proposed by Waage (1979) and Driessen et al. (1995).

6. We formulate an ‘adjustable termination rate’ model for patch exploitation in both Cotesia species in multi-patch environments.


Introduction

  1. Top of page
  2. Abstract
  3. Introduction
  4. 1. Do C. glomerata and C. rubecula show interspecific differences in the way they use information to make patch leaving decisions?
  5. Materials and methods
  6. Results
  7. Discussion and Conclusions
  8. Acknowledgements
  9. References

Optimal foraging theory attempts to relate variation in reproductive success of individuals to variation in their foraging behaviour. Several optimality models have focused on behavioural decision variables such as patch residence time. The marginal value theorem (Charnov 1976) considers the optimal residence time in a patch and suggests that this depends on the gain in the patch, travel times and the rate of gain in the habitat averaged over the total time spent there. Charnov's model assumes that the animal has ‘complete information’, which is unlikely to be the case in nature. His model does not suggest a mechanism or decision rule that foraging animals could apply to achieve the optimal residence time.

So, how should animals make decisions on when to leave a patch? Several authors have considered what mechanisms might be employed to achieve optimal patch residence times. It was suggested that animals use simple, rather fixed ‘rules of thumb’ (see Cowie & Krebs 1979; Stephens & Krebs 1986). A number of analyses focused on the performance of such simple patch leaving rules (Iwasa, Higashi & Yamamura 1981; McNair 1982; Green 1984). However, experimental evidence for the general use of such simple decision rules is scarce. If animals have incomplete information on patch and habitat profitability, learning through experience (sampling) may provide a mechanism for the optimization of patch leaving decisions. There is clear empirical evidence that experience affects parasitoid decisions on where to go and how long to stay (Papaj et al. 1994; Vet, Lewis & Cardé 1995). Animals can increase their foraging effort in microhabitats where they had rewarding experiences, like successful ovipositions, through associative learning (Papaj et al. 1994). However, it is not always obvious whether rewarding experiences should increase or decrease the animal's leaving tendency: in some patch types (e.g. when there is generally only a single host present) it is adaptive to leave after one oviposition (Strand & Vinson 1982), while in others it is better to stay.

This leads to the question of what kind of cues animals could use to achieve adaptive patch residence times. Waage (1979) used a parasitoid as a model system and proposed a patch exploitation mechanism for Venturia canescens: when a parasitoid enters a patch it has a tendency to stay in that patch and to turn around when encountering the edge of the patch. This responsiveness to the patch edge is initially set by the intensity of the host-associated chemicals (kairomones) in the patch. With time, this responsiveness decays at a constant rate to a specific threshold, after which the parasitoid leaves the patch. Ovipositions increase the responsiveness, thereby increasing patch residence time. In a recent study Driessen et al. (1995) proposed a decision mechanism for patch leaving in the same species that contrasted to that of Waage (1979) in one important way: After the initial assessment of host density, ovipositions decrease the responsiveness, thereby decreasing patch residence time (a count-down mechanism). The most important factor determining which mechanism is adaptive is the accuracy of patch density assessment: If this is accurate, the count-down mechanism is adaptive. The second factor differentiating between these rules is host distribution and depends on the first factor: in an environment where patches contain a uniform number of hosts (and patch assessment is accurate), a count-down rule may perform most efficiently, while in a habitat with a heterogeneous host distribution (and inaccurate patch assessment) Waage's mechanism might be adaptive.

In the patch exploitation mechanisms as proposed by Waage (1979) and Driessen et al. (1995) only intrapatch factors play a role. However, the marginal value theorem assumes that the response to patches is tuned to mean host availability in the environment. Hence, it is essential to consider whether patch leaving decisions change as the parasitoid gains experience about the world in which it lives. To answer this question, it is important to observe the animal in an environment that allows it to express its entire range of natural behaviour, i.e. in a multiple patch environment in which the animal has a place to go to after leaving the patch.

Several authors have correlated patch leaving decisions to experiences on the current patch by means of the proportional hazards model (Cox 1972).

Hemerik et al. (1993) and Haccou et al. (1991) showed that the leaving tendency of two Leptopilina parasitoid species decreased with the presence of kairomone, ovipositions and high recent oviposition rates. van Roermund et al. (1994) showed that the presence of honeydew and ovipositions decreased the leaving tendency of the parasitoid Encarsia formosa. van Steenis et al. (1996) showed that the leaving tendency of Aphidius colemani only decreased at patches containing a high density of 100 hosts. On these patches A. colemani used a count-down mechanism: ovipositions increased the leaving tendency, especially after the parasitoids had encountered 100 hosts or more.

Time allocation to foraging behaviour should be adapted to the density and spatial distribution of hosts in patches. Hosts may occur singly, evenly distributed or clustered in a patch. Parasitoids that attack several host species differing in their spatial distribution may have evolved flexible patch leaving strategies. In the tritrophic system of crucifers, Pieris, and Cotesia, the host species differ markedly in their spatial distribution on the scale of plants and leaves. The small white butterfly (Pieris rapae (L.)) lays its eggs separately, often only one or a few on a leaf. This results in a rather low variance in the number of larvae per leaf. The larvae are also evenly distributed over a leaf. We will call this the ‘uniform’ distribution. The large white butterfly (Pieris brassicae (L.)) lays its eggs in clusters of 7–150 eggs, so young larvae have a more heterogeneous distribution. Cotesia glomerata (L.) is a gregarious larval endoparasitoid of several Pieris species (Brodeur, Geervliet & Vet 1996). It attacks both P. brassicae and P. rapae, with a higher acceptance of P. brassicae larvae. First and second instar P. rapae larvae are readily accepted (Brodeur et al. 1996). The solitary larval endoparasitoid Cotesia rubecula (Marshall) is a specialist on P. rapae, but it will accept P. brassicae larvae as well (Brodeur et al. 1996). Hence, C. glomerata has to deal with host species that occur in uniform as well as heterogeneous distributions, while C. rubecula is specialized on the uniform host distribution of P. rapae. To both Cotesia species there is a clear advantage in developing in the preferred host species (Geervliet & Brodeur 1992). Naive females of both C. glomerata and C. rubecula do not discriminate between infochemicals from plants infested by P. rapae and P. brassicae (Geervliet, Vet & Dicke 1994). There is evidence that both parasitoid species visit both patch types in the field (Geervliet 1997).

Using the Cotesia-Pieris system, we specifically address the following questions:

1. Do C. glomerata and C. rubecula show interspecific differences in the way they use information to make patch leaving decisions?

  1. Top of page
  2. Abstract
  3. Introduction
  4. 1. Do C. glomerata and C. rubecula show interspecific differences in the way they use information to make patch leaving decisions?
  5. Materials and methods
  6. Results
  7. Discussion and Conclusions
  8. Acknowledgements
  9. References

2. Do both Cotesia species alter their patch leaving strategies in environments that differ in the available host species?

3. Do the parasitoids use simple rules of thumb, or do they employ complex, flexible rules?

4. Do the parasitoids use Waage's (1979) patch exploitation mechanism on the host with the heterogeneous distribution and Driessen et al. 1995) count-down mechanism on the uniformly distributed host?

Materials and methods

  1. Top of page
  2. Abstract
  3. Introduction
  4. 1. Do C. glomerata and C. rubecula show interspecific differences in the way they use information to make patch leaving decisions?
  5. Materials and methods
  6. Results
  7. Discussion and Conclusions
  8. Acknowledgements
  9. References

Experiments

In our analysis we used the behavioural data collected by Wiskerke & Vet (1994). We summarize their methods below; full details of the experimental procedure are given in their paper. The foraging behaviour of C. rubecula and C. glomerata parasitoids was observed in a ‘semi-field set-up’ with Pieris-infested cabbage plants in a greenhouse compartment. The set-up consisted of eight Brussels sprouts plants placed on a table in two parallel rows. Four of these plants were clean, while the other four were infested. Two fans placed at the end of the table provided an air stream of 0·3–0·4 m s−1 at the parasitoid release site. This release site, situated at the downwind end of the table, consisted of an excised Brussels sprouts leaf with feeding damage (no hosts), inflicted by the same species of host(s) as present in the particular experiment. The leaves had a diameter of about 12 cm. Three types of experiments per parasitoid species were conducted. In these, eight plants were used in the following combinations: (i) four plants infested with P. brassicae larvae and four clean plants; (ii) four plants infested with P. rapae larvae and four clean plants; (iii) two plants infested with P. rapae larvae, two with P. brassicae larvae and four clean plants.

Butterflies of each Pieris species were allowed to oviposit on the respective experimental plants. The excess of eggs was removed to produce the desired experimental density. A plant infested with P. brassicae contained on average 20 early first instar larvae (hereafter EL1), in one or two clusters. The density in such clusters ranged from 1 to 30. A plant infested with P. rapae contained 20 larvae as well, but more evenly distributed over several leaves on the plant. Each infested leaf contained 1–8 solitarily feeding EL1 larvae (average 2·44, s.d. 1·34). These are accepted by both Cotesia species. The foraging behaviour of individual parasitoids (C. rubecula or C. glomerata) was observed and recorded continuously. Each observation started at the moment the parasitoid left the release site and flew to a plant. An observation was terminated after 1 h or when the parasitoid left the foraging arena, landed elsewhere and remained there for more than 1 min (Wiskerke & Vet 1994). For each treatment 18–24 females of each Cotesia species were tested.

The proportional hazards model

Here we give a short explanation of the essentials of the model (for a more thorough description we refer to Kalbfleisch & Prentice (1980) and Haccou & Hemerik (1985)). The proportional hazards model is used to analyse which factors in the environment (or experiences of the parasitoid) are correlated with an increase or decrease in the tendency of a parasitoid to leave a patch. It is assumed that the parasitoids have a basic tendency to leave the patch (base line hazard), which is reset after certain events, so-called renewal points. Renewal points occur here at the moment a patch is entered and when searching resumes after an oviposition. The model is a multiple regression method, with the relative strength of each of p factors being estimated by means of partial likelihood maximization (see Kalbfleisch & Prentice 1980). The following eqn 1 describes the effect of those factors on the leaving tendency:

  • image(eqn 1)

in which h(t; z) is the probability per second to leave the patch (in our case fly away from the leaf), and h0(t) is the basic tendency to leave, i.e. when there is no effect of any factor like experience or environmental information. The zi are the factors (covariates) that might influence the leaving tendency (these are coded in advance). Each factor z can have different values, e.g. ‘no host and feeding damage present’ would be represented by 0, while the presence of the host and feeding damage would be coded by 1. The β values are the relative strengths of the effects of the covariates. These are estimated in the analysis. The influence of experiences (the covariates) on the leaving tendency is modelled as a multiplicative effect of exp(βizi) on the base-line hazard. If this term is below 1, the leaving tendency is reduced; above 1 it is increased. An increased leaving tendency implies a shorter giving up time (GUT). The expected GUT is 1/h(t; z).

Covariates

The selection of covariates is a crucial step in the analysis. Many factors could be tested for having an effect on the parasitoids’ leaving tendency. In our choice we have taken into account the biology of both Cotesia parasitoids, factors found to be important in other parasitoids and some factors generally presumed to be important in patch leaving models. From the literature (see Introduction) factors including host-induced damage, kairomone concentration and ovipositions emerge as important in the foraging process. These factors have a high informational value and are reliable cues. In our set of covariates we allow for the use of experiences acquired during previous patch visits.

It is important at which spatial scale we define the patch that is left. Depending on the behaviour of the animal, it might be the host, the spot with feeding damage and kairomones, a leaf or the plant. The parasitoids do not walk from leaf to leaf when searching for hosts on a plant. Movement from leaf to leaf is facilitated by flight. The most dramatic change in foraging behaviour occurs on leaves containing larvae and feeding damage (Wiskerke & Vet 1994). Therefore, we define the patch to be a leaf in our analysis. Hence, leaves that do not contain hosts and damage are empty patches. We define ‘leaving’ as flying away from a leaf, followed by landing somewhere else (e.g. another leaf). If the parasitoid takes off and subsequently relands on that same leaf, we consider this to be part of the same patch visit. These short excursions are rather similar to those of Venturia canescens as found by Waage (1979) and Driessen et al. (1995). We have chosen a leaving criterion that allows for a clear biological interpretation and is not as arbitrary as the ‘14 or 60 s off patch’-criteria of Waage (1979). An encounter with a host is defined by an ovipositor insertion into a host. From the behavioural records of Wiskerke & Vet (1994) it was not possible to discriminate parasitism from superparasitism for both species of Cotesia. The description of the covariates we have selected, and the way they are coded, are given in Table 1.

Table 1.  Covariates tested for having an effect on the leaving tendency
1.The present leaf contains P. brassicae and host-damage (no/yes) = (0/1)
2.The present leaf contains P. rapae and host-damage (no/yes) = (0/1)
3.The number of encounters, i.e. ovipositor insertions, during the present leaf-visit (notation for 0, 1, 2 or more insertions is (0/1/Ð2) and it is coded as (0/1/2)
4.The total number of encounters in the protocol (0/1/Ð2) = (0/1/2)
5.The number of ovipositor insertions during the previous leaf-visit (0/1) = (≤1 and Ð2)
6.P. brassicae and host damage were present during the previous leaf-visit (no/yes) = (0/1)
7.P. rapae and host damage were present during the previous leaf-visit (no/yes) = (0/1)
8.The cumulative number of host–damaged leaves visited (0/1/Ð2) = (0/1/2)
9.The number of undamaged leaves visited since the latest ovipositor insertion (0/Ð1)
10.The time since the start of the behavioural protocol

Statistical analysis

The experiments of Wiskerke & Vet (1994) resulted in six data sets. We analysed each data set by means of the proportional hazards model. After the likelihood maximization, our first step was to test whether the joint effect of all covariates was significant. Under the null-hypothesis β1 = · · · = βp = 0, none of the covariates zi has any effect on the leaving tendency. The test statistic T has asymptotically a Chi-squared distribution with p degrees of freedom (Miller 1981; Haccou & Hemerik 1985). The second step was to test for each β (and therefore each covariate) separately whether it had a significant effect. The third step was to test for pairwise effects of covariates that were not significant by themselves. When, e.g. d_f. = 8, test statistics in the interval (3·84, 15·5) are not significant by themselves, but together with another covariate they can have a significant multiplied effect (T > 15·5). We applied a multiple comparison to test for such pairwise effects. We chose to only consider single covariates and pairs of them because they allow for clear biological interpretations.

In addition to the proportional hazards analysis, we analysed part of the data with tests as used by Waage (1979) and Driessen et al. (1995). Waage's mechanism (1979) and Driessen et al's count-down mechanism (1995) clearly contrast in the way ovipositions affect the residence time in a patch. However, the two mechanisms do not contrast as clearly with respect to the effect of an oviposition on the leaving tendency (see Fig. 1). Hence, separate tests were conducted for the effect of encounters on patch residence times. This makes comparisons more straightforward and serves as an extra check on the results of the model analysis. In these analyses we used all visits to leaves that the parasitoid had not visited before (the first visits to patches).

image

Figure 1. Simplified graphical representation of Waage's (1979) patch time model (a) and the count-down mechanism (Driessen et al. 1995) (b). The arrow marks the occurrence of an oviposition. If no oviposition occurs, the patch should have been left at the time indicated by the asterisk. Underneath the graphs the residence time and GUT of a patch visit in which one oviposition occurred are compared to that of a patch visit in which no oviposition occurred. It should be noted that in both models the intercept of the line increases with host infestation.

Download figure to PowerPoint

Results

  1. Top of page
  2. Abstract
  3. Introduction
  4. 1. Do C. glomerata and C. rubecula show interspecific differences in the way they use information to make patch leaving decisions?
  5. Materials and methods
  6. Results
  7. Discussion and Conclusions
  8. Acknowledgements
  9. References

Foraging behaviour

Wiskerke & Vet (1994) give an elaborate description of the foraging behaviour of both Cotesia species in the different environments. After taking off from the release site the parasitoids often flew to an infested plant and hovered over the leaves at a distance of a few centimetres. Most parasitoids visited several leaves on different plants. Both Cotesia species frequently made short excursions. Either they hovered over other areas of the same leaf, or they even flew over other leaves before relanding.

Baseline hazards

The estimated baseline hazards of C. rubecula and C. glomerata in the three different environments are given in Table 2. The basic leaving tendencies of C. rubecula and C. glomerata are high in the set-up where only their less preferred host was present, P. brassicae and P. rapae, respectively. When we compare the estimated baseline hazards for both Cotesia species, it is clear that C. rubecula tends to have a higher leaving tendency than C. glomerata in all of the three environments.

Table 2.  Estimated baseline hazards for C. rubecula and C. glomerata (probability per second to fly away, when there is no effect (yet) of the covariates) in the different environments
Set-upC. rubeculaC. glomerata
P. rapae0·0097490·004175
P. brassicae0·016780·002410
Both Pieris spp.0·0082610·004501

Leaving tendency of c. rubecula

The combined effect of all covariates was significant in each of the three experiments (P < 0·05, Table 3). C. rubecula used the same cue in all three environments: on leaves containing hosts and feeding damage the leaving tendency decreased (Table 4). In the environment with P. rapae this resulted in an increase in the expected GUT [1/h(t; z)] from 103 s on an empty leaf, to 237 s on an infested leaf. In the environment with P. brassicae the expected GUT increased from 60 to 102 s. In the environment with both Pieris species the leaving tendency only decreased on leaves containing the preferred host, P. rapae. The expected GUT increased from 121 s on a leaf that was clean or infested with the less preferred host P. brassicae, to 401 s on a patch with P. rapae. C. rubecula rarely stayed the complete 1-h observation period in the set-up with P. brassicae. The parasitoids often flew to the roof of the glasshouse after about 30 min. C. rubecula employed a simple rule in all three environments, using information on the current patch to make patch leaving decisions.

Table 3.  The value of the test statistic T (d_f.) for the combined effects of all covariates on the leaving tendency in experiments with C. rubecula
Set-upT(d_f.)
  • *

    P Š 0·001.

C. rubecula on P. rapae71·06(8)*
C. rubecula on P. brassicae39·35(8)*
C. rubecula on both Pieris spp.124·94(10)*
Table 4.  Covariates for C. rubecula (numbers as in Table 1). Test statistic T is marked with an asterisk if a covariate (or pair of covariates) has a significant effect. Downward arrows indicate a decreasing effect, upward arrows an increasing effect of a covariate on the leaving tendency
  • *

    P < 0·01;

  • **

    P < 0·001.

Host(s)CovariatesT (d_f.)leaving tendencyβ
P. rapae(2) present leaf contains P. rapae and damage43·77 (8)**[DOWNWARDS ARROW]−0·8385
P. brassicae(1) present leaf contains P. brassicae and damage21·03 (8)*[DOWNWARDS ARROW]−0·5384
Both host spp.(2) present leaf contains P. rapae and damage85·26 (10)**[DOWNWARDS ARROW]−1·198

Leaving tendency of c. glomerata

The combined effect of all covariates was significant in all three experiments (Table 5). C. glomerata used information on the presence of hosts and feeding damage in all three environments: On infested leaves the leaving tendency decreased (Table 6). In the environment with P. rapae this resulted in an increase in the expected GUT from 240 s on an empty leaf, to 484 s on an infested leaf. In the environment with P. brassicae the expected GUT increased from 415 to 1021 s. In the environment with both Pieris species the increase was from 222 s on an empty patch to 499 s on a patch with P. rapae and 779 s on a patch with P. brassicae.

Table 5.  The value of the test statistic T (d_f.) for the combined effects of all covariates on the leaving tendency in experiments with C. glomerata
Set-upT(d_f.)
  • *

    P < 0·001.

C. glomerata on P. brassicae140·94(8)*
C. glomerata on P. rapae27·22(8)*
C. glomerata on both Pieris spp.82·10(10)*
Table 6.  Covariates for C. glomerata (numbers as in Table 1). Test statistic T is marked with an asterisk if a covariate (or pair of covariates) has a significant effect. Downward arrows indicate a decreasing effect, upward arrows an increasing effect of a covariate on the leaving tendency
Host(s)CovariatesT (d_f.)Effect on leaving tendencyβ
  • *

    P < 0·05;

  • **

    P < 0·01;

  • ***

    P < 0·001; NS, not significant.

P. brassicae(1) present leaf contains P. brassicae and damage16·34 (8)*[DOWNWARDS ARROW]−0·9003
(3) encounters on the present leaf14·86 (8) NS[DOWNWARDS ARROW]−0·5573
(4) total encounters with hosts5·96 (8) NS[DOWNWARDS ARROW]−0·3644
(8) total infested leafs visited7·80 (8) NS[UPWARDS ARROW]  0·4640
(10) the time in the protocol 3 & 4 3 & 8 3 & 10 4 & 106·66 (8) NS 28·67 (8)*** 27·79 (8)*** 23·83 (8)** 20·43 (8)**[DOWNWARDS ARROW]−0·0003096
P. rapae(2) present leaf contains P. rapae and damage15·56 (8)*[DOWNWARDS ARROW]−0·7038
Both host(1) present leaf contains P. brassicae and damage21·67 (10)*[DOWNWARDS ARROW]−1·254
spp.(2) present leaf contains P. rapae and damage27·30 (10)**[DOWNWARDS ARROW]−0·8082

In the set-up where P. brassicae was present, encounters with hosts on the present patch decreased the leaving tendency even further although its effect was only significant in combination with other covariates (Table 6). After multiple encounters the parasitoids stayed on average 2550 s. In the P. brassicae environment experience acquired during previous patch visits influenced the leaving tendency. Three factors had a pairwise effect in combination with the number of encounters on the present patch: (i) the total number of encounters with hosts; (ii) the total number of infested leaves already visited; and (iii) the time in the protocol. The first decreased the leaving tendency, the second increased it, while the third had a decreasing effect. The total number of encounters with hosts and the time in the protocol had a pairwise decreasing effect on the leaving tendency. C. glomerata used a complex rule in an environment where only the gregarious host P. brassicae was present. In environments with the solitary host it switched to a simple rule in which the leaving tendency was mainly affected by the presence and type of infestation.

The effect of host distribution and density

Leaves infested with one cluster of P. brassicae larvae usually contained one large feeding damage site. In contrast, leaves infested with low densities of P. rapae larvae usually contained several little feeding damage sites. Both Cotesia species clearly distinguished the type of damage: C. glomerata had a high leaving tendency on the low density P. rapae patches and a lower leaving tendency on the high density P. brassicae patches. C. rubecula had a high leaving tendency on the high density P. brassicae patches and a low leaving tendency on the comparatively low density P. rapae patches. Apparently, for C. rubecula a high density does not automatically translate into a low leaving tendency. This excited our interest in the effect of larval density (and the corresponding level of feeding damage) on the leaving tendencies of both Cotesia species on patches with P. rapae. Analysis of leaving tendencies on patches containing one, two, three and four or more P. rapae larvae for visits in which no host encounters occurred, revealed that there was no significant effect of density on the leaving tendency, neither for C. rubecula (P = 0·13, logrank test for survival curves), nor for C. glomerata (P = 0·34, logrank test). Hence, larval density on P. rapae-infested leaves was not found to influence the leaving tendency in the parasitoids. However, the parasitoids do respond differently to clean and host-infested leaves and significantly stay longer on the latter (C. rubecula: P = 5·3×10−5, logrank test, Nempty = 50, Ninfested = 176); C. glomerata: P = 0·029, logrank test, Nempty = 48, Ninfested = 108).

For C. rubecula there was a significant effect of encounters (0, 1, 2) on the residence time in patches infested with P. rapae (PŠ 10−6, Kruskal–Wallis test; all pairs of groups are different, non-parametric multiple comparisons (Siegel & Castellan 1988), all P < 0·025). For C. glomerata on P. rapae there was a significant effect as well of encounters on the residence time ((0, 1, 2 encounters), PŠ 10−3, Kruskal–Wallis test; all pairs of groups are different, multiple comparisons, all P < 0·01). In both Cotesia species encounters with the solitarily feeding P. rapae larvae increased patch residence time.

Discussion and Conclusions

  1. Top of page
  2. Abstract
  3. Introduction
  4. 1. Do C. glomerata and C. rubecula show interspecific differences in the way they use information to make patch leaving decisions?
  5. Materials and methods
  6. Results
  7. Discussion and Conclusions
  8. Acknowledgements
  9. References

Initial search

When an inexperienced C. rubecula or C. glomerata female takes off for the first time, her informational state is incomplete. She has some clue of the level of infestation in her microhabitat from a distance (Geervliet et al. 1998a), but seems incapable of long-distance discrimination between patches containing preferred and less preferred hosts (Geervliet et al. 1994). The present study aimed to address how parasitoids make patch leaving decisions on different types of patches and how experience modifies decision making.

C. rubecula

C. rubecula had the lowest leaving tendency on leaves containing the preferred host P. rapae. In the set-up with P. brassicae only, the leaving tendency was highest, both on empty and infested leaves. Thus, C. rubecula tends to spend a short time in unfavourable parts of its environment and will forage extensively in the good parts. In all three set-ups previous experiences (i.e. the number of infested leaves visited and the total number of hosts encountered) did not affect the leaving tendency in the current patch.

C. glomerata

C. glomerata females had a lower leaving tendency on leaves containing the preferred host than on leaves infested with P. rapae. Patch leaving decisions in C. glomerata changed with experience. When a naive C. glomerata parasitoid foraged in an environment with P. brassicae, it had a lower leaving tendency on leaves containing feeding damage than on clean leaves. This tendency decreased after the parasitoid's first few ovipositions. During subsequent visits to infested leaves the leaving tendency increased. In the field, larval density on patches varies considerably: P. brassicae clusters can consist of anything between 7 and 150 larvae. In such a variable environment it can be adaptive to use information on the quality of previously visited patches to adapt foraging decisions during subsequent patch visits. The increase in leaving tendency with the number of infested leaves visited in the P. brassicae set-up was strongest early in the protocol. This is in line with the predictions of the marginal value theorem. When more good patches are visited with relatively short travel times, the estimated value of the environment as a whole increases and consequently the leaving tendency should increase. C. glomerata switched to a simple rule in environments containing P. rapae: information on previous patch visits did not affect the leaving tendency in P. rapae environments.

Interspecific differences

The two parasitoid species showed interspecific variation in the way they made patch leaving decisions. C. rubecula had a higher leaving tendency than C. glomerata in all environments. The high leaving tendencies of C. rubecula fit well into a foraging strategy focusing on a solitary host. The generalist C. glomerata was more plastic in its decisions than the specialist C. rubecula.

Results compared to the models of waage and driessen et al.

In the patch exploitation mechanisms proposed by Waage (1979) and Driessen et al. (1995) the concentration of kairomone on a patch determines the initial responsiveness of the parasitoid. In environments where (i) patches contain reliable and detectable information on host availability or (ii) contain a uniform number of hosts, a count-down mechanism (Driessen et al. 1995) would perform best (factor (i) being the key factor). When initial patch density assessment is inaccurate, it would be adaptive for the parasitoids to use an incremental mechanism, i.e. to prolong the time in the patch with ovipositions, as in Waage's (1979) model.

Previous experiments with C. glomerata, using individual females on single leaves and methods similar to Driessen et al. (1995), showed that the parasitoids did not have different leaving tendencies on leaves containing feeding damage from 1 or 8 P. brassicae larvae (P = 0·77, nd1 = 15, nd8 = 15, logrank test, Vos, unpublished data). This suggests that an incremental mechanism should be used. Indeed, ovipositions did prolong the time C. glomerata spent on a patch with P. brassicae. Waage's (1979) patch time model did not address the effects of factors as experience in previously visited patches or olfactory cues from the environment on patch time in the current patch. For C. glomerata these factors proved to be important. First of all patch leaving decisions changed as a function of experience acquired during successive patch visits. Secondly, patch leaving decisions were environment-dependent.

Neither of the two Cotesia species used a count-down rule in the environment with the uniform distribution of P. rapae. Several factors may inhibit the use of a count-down rule.

1. Kairomone concentration or feeding damage may be an unreliable source of information on host presence. Young P. rapae larvae, for example, face a high risk of mortality due to predation (Jones et al. 1987). A count-down rule would perform badly in a patch that was recently depleted by a predator.

2. Detectability may be problematic. Driessen et al. ’s patches were small (< 3·4 cm). A Brussels sprouts leaf is large (< up to 40 cm). The parasitoids may not be able to detect the concentration of leaf damage and kairomone on the entire leaf, when searching at a specific site. The results indeed show that although both Cotesia species stay longer on infested than on uninfested leaves, they do not significantly increase their residence time with larval density.

3. The distribution of hosts across patches in the P. rapae environment (low density and variance) is not strictly uniform. Patches mostly contained one, two or three hosts. This may be too variable for the count-down rule to perform well.

Interestingly, Driessen et al. (1995) showed that V. canescens needed less time to get multiple ovipositions on a patch containing four hosts, than to get one oviposition on a patch containing a single host. Hence, for V. canescens, efficiency increased with the host density in the patch. This is not the case for either Cotesia species on P. rapae. The time until first oviposition (TUFO) did not differ significantly on patches containing different host densities. Therefore, time until first oviposition is not a reliable source of information on host density. There was no significant difference in the efficiency (number of encounters/patch residence time) on different host densities. On patches containing two or more hosts, the TUFO and time between first and second oviposition did not differ significantly either. An increased leaving tendency following an oviposition (as in a count-down rule) would only diminish the probability of finding another host on that same leaf. When the estimation of the damage level at low densities of P. rapae is inaccurate, and the time when ovipositions occur reveals little information on density or depletion, neither Driessen et al. 1995), nor Waage's (1979) patch exploitation mechanisms should be used. It may be better to adjust the leaving tendency to a cue like the presence or absence of feeding damage, and reset the leaving tendency after each oviposition to the initial value as it was set upon entering the patch. This is exactly what both Cotesia species seem to do: on P. rapae there is no effect of density on the leaving tendency; encounters reset the leaving tendency, thereby prolonging the residence time with another 1/h(t; z) seconds (i.e. the new expected GUT).

If no good intrapatch information on host density is available, this will constrain the use of interpatch information as well. Even if the animal has information on the value of patches in its environment through oviposition experiences, it may be impossible to use that information, because the value of the current patch relative to that of the ‘average patch’ is unknown. This may be the reason that neither C. glomerata, nor C. rubecula showed use of information on previous patch visits in the environment with P. rapae (with our covariates).

Adjustable termination rates

Based on the proportional hazards analysis, we propose the following patch exploitation mechanism for the description of the behaviour of C. rubecula.

1. Set a ‘basic leaving tendency’ to the available olfactory information in the environment while still in flight.

2. Land on a plant that releases olfactory and/or visual information on feeding damage.

3. Have a relatively high leaving tendency if the leaf is clean.

4. Have a decreased leaving tendency if feeding damage is present (this decrement may also depend on the environmental olfactory information).

5. Have a strong decrease in leaving tendency on a leaf containing the type of damage caused by the preferred host species.

6. Reset the leaving tendency after each oviposition, resulting in an expected GUT similar to what it was upon entering the patch (note that this does increase the residence time).

Go through the same cycle after taking off from the leaf. The decrement in the leaving tendency in steps 4 and 5, and the reset in step 6 should increase the average time spent in the patch such that it equals the average time needed to find a host in an infested patch. The steps involved in the leaving tendency (1, 3, 4, 5 and 6) can be represented mathematically by an equation highly similar to the proportional hazards model (1). Factors involved in a tendency to arrive (like step 2) can be formulated by essentially the same type of equation (Ormel, Gort & van Alebeek 1995). In adjustable termination rate models the animal is considered as continually adjusting its probability per unit time to leave (termination rate), according to ‘good’ and ‘bad’ experiences. The model can be used in simulations aimed at comparing the performance of complex and simple decision rules in different environments, taking into account the constraints on the animal's informational state. Step 2 of the model is supported by the analysis of Wiskerke & Vet (1994): first landings occur mostly on infested plants, and by extensive work on olfactory responses in these parasitoids (Steinberg et al. 1992; Geervliet, Vet & Dicke 1994; Geervliet et al. 1996). The proportional hazards analysis provided the evidence for steps 1, 3, 4, 5 and 6 (these steps are based on those covariates that had a significant effect on the leaving tendency).

C. glomerata will use a patch-exploitation mechanism rather similar to the one used by C. rubecula in environments containing only P. rapae. However, for C. glomerata it would be adaptive to learn to discriminate between the olfactory information released from P. brassicae- and P. rapae-infested plants in habitats containing both host species, and land on leaves with the preferred host (in step 2). Recently it was shown that C. glomerata's response to P. brassicae-infested leaves does increase with oviposition experience in P. brassicae (Geervliet et al. 1998b). As this parasitoid learns to prefer to land on P. brassicae patches, it may start to specialize on this host and treat environments containing both hosts like an environment containing solely P. brassicae. In such an environment C. glomerata differs from C. rubecula in a number of ways: (i) in step 6, the leaving tendency decreases with ovipositions; (ii) the parasitoids use the more complex mechanism in which the use of information on previously visited patches is incorporated as well. For C. glomerata we propose steps 7, 8 and 9: Adjust the leaving tendency according to (i) the cumulative number of ovipositions experienced, (ii) the number of infested patches visited and (iii) time. The exact balance of these factors will depend on the amount of heterogeneity in the environment and may change as the parasitoid moves into richer or poorer parts of its environment.

Arrival and leaving tendencies

The analysis presented here focused on patch leaving decisions. Although a proportional hazards analysis of arrival tendencies is beyond the scope of this paper, it is clear that the efficiency of a patch leaving mechanism depends heavily on the tendency to arrive in certain patch types. Patch exploitation strategies can be thought of as balanced sets of effects of experiences on the tendencies to arrive in and leave certain microhabitats.

Termination rates of foragers are local in time and space: they depend on the animal's environment and may change as it gains experience and moves into other parts of its habitat. This is a source of behavioural variation that should not be ignored when considering the ‘optimality’ of animal decisions.

Acknowledgements

  1. Top of page
  2. Abstract
  3. Introduction
  4. 1. Do C. glomerata and C. rubecula show interspecific differences in the way they use information to make patch leaving decisions?
  5. Materials and methods
  6. Results
  7. Discussion and Conclusions
  8. Acknowledgements
  9. References

We would like to thank Carlos Bernstein, Gerard Driessen, Kate Flanagan, Jeffrey A. Harvey and an anonymous referee for their valuable suggestions and comments on previous versions of the manuscript. Nico J. de Boer kindly provided separate data sets, similar to those of Wiskerke & Vet (1994), that he collected in the same glasshouse compartment. These were helpful in the process of selecting covariates.

References

  1. Top of page
  2. Abstract
  3. Introduction
  4. 1. Do C. glomerata and C. rubecula show interspecific differences in the way they use information to make patch leaving decisions?
  5. Materials and methods
  6. Results
  7. Discussion and Conclusions
  8. Acknowledgements
  9. References
  • Brodeur, J., Geervliet, J.B_f., Vet, L.E.M. 1996 The role of host species, age and defensive behaviour on ovipositional decisions in a solitary specialist and gregarious generalist parasitoid (Cotesia species). Entomologia Experimentalis et Applicata, 81, 125 132.
  • Charnov, E.L. 1976 Optimal foraging, the marginal value theorem. Theoretical Population Biology, 9, 129 136.
  • Cowie, R.J. & Krebs, J.R. 1979 Optimal foraging in patchy environments. The British Ecological Society Symposium, Vol. 20. Population Dynamics(eds R. M.Anderson, B. D.Turner & L. R.Taylor), pp. 183 205. Blackwell Scientific Publications, Oxford.
  • Cox, D.R. 1972 Regression models and life tables. Biometrics, 38, 67 77.
  • Driessen, G., Bernstein, C., Van Alphen, J.J.M., Kacelnik, A. 1995 A count-down mechanism for host search in the parasitoid Venturia canescens. Journal of Animal Ecology, 64, 117 125.
  • Geervliet, J.B_f. 1997 Infochemical use by insect parasitoids in a tritrophic context: comparison of a generalist and a specialist. PhD-dissertation, Wageningen.
  • Geervliet, J.B_f., Ariëns, S., Dicke, M., Vet, L.E.M. 1998a Long distance assessment of patch profitability through volatile infochemicals by the parasitoids Cotesia glomerata and C. rubecula (Hymenoptera: Braconidae). Biological Control, 11, 113 121.
  • Geervliet, J.B_f. & Brodeur, J. 1992 Host species affecting the performance of the larval parasitoids Cotesia glomerata and Cotesia Rubecula (Hymenoptera: Braconidae). II. Comparative suitability of three Pieris species (Lepidoptera: Pieridae). Mededelingen Faculteit Landbouwwetenschappen Universiteit Gent, 57/2b. 547–550.
  • Geervliet, J.B_f., Vet, L.E.M., Dicke, M. 1994 Volatiles from damaged plants as major cues in long-range host-searching by the specialist parasitoid Cotesia rubecula (Hymenoptera: Braconidae). Entomologia Experimentalis et Applicata, 73, 289 297.
  • Geervliet, J.B_f., Vet, L.E.M., Dicke, M. 1996 Innate responses of the parasitoids Cotesia glomerata and C. rubecula (Hymenoptera: Braconidae) to volatiles from different plant–herbivore complexes. Journal of Insect Behaviour, 9, 525 538.
  • Geervliet, J.B_f., Vreugdenhil, A.I., Dicke, M., Vet, L.E.M. 1998b Learning to discriminate between infochemicals from different plant-host complexes by the parasitoids Cotesia glomerata and C. Rubecula. Entomologia Experimentalis et Applicata, 86, 241 252.
  • Green, R_f. 1984 Stopping rules for optimal foragers. The American Naturalist, 123, 30 43.
  • Haccou, P., De Vlas, S.J., Van Alphen, J.J.M., Visser, M.E. 1991 Information processing by foragers: effects of intra-patch experience on the leaving tendency of Leptopilina heterotoma. Journal of Animal Ecology, 60, 93 106.
  • Haccou, P. & Hemerik, L. 1985 The influence of larval dispersal in the cinnabar moth (Tyria jacobaeae) on predation by the red wood ant (Formica polyctena): an analysis based on the proportional hazards model. Journal of Animal Ecology, 54, 755 770.
  • Hemerik, L., Driessen, G., Haccou, P. 1993 Effects of intra-patch experiences on patch time, search time and searching efficiency of the parasitoid Leptopilina clavipes. Journal of Animal Ecology, 62, 33 44.
  • Iwasa, Y., Higashi, M., Yamamura, N. 1981 Prey distribution as a factor determining the choice of optimal foraging strategy. The American Naturalist, 117, 710 723.
  • Jones, R.E., Nealis, V.G., Ives, P.M., Scheermeyer, E. 1987 Seasonal and spatial variation in juvenile survival of the cabbage butterfly Pieris rapae: Evidence for patchy density dependence. Journal of Animal Ecology, 56, 723 737.
  • Kalbfleisch, J.D. & Prentice, R.L. 1980The Statistical Analysis of Failure Time Data. Wiley and Sons, New York.
  • McNair, J.N. 1982 Optimal giving-up times and the marginal value theorem. The American Naturalist, 119, 511 529.
  • Miller, R.G. Jr 1981Simultaneous Statistical Inference. Springer, New York.
  • Ormel, G.J., Gort, G., Van Alebeek, F.A.N. 1995 Analyzing host-location in Uscana lariophaga Steffan (Hymenoptera; Trichogrammatidae), an egg parasitoid of bruchids (Coleoptera; Bruchidae), using Cox's proportional hazards model. Bulletin of Entomological Research, 85, 113 123.
  • Papaj, D.R., Snellen, H., Swaans, K., Vet, L.E.M. 1994 Unrewarding experiences and their effect on foraging in the parasitic wasp Leptopilina heterotoma (Hymenoptera: Eucoilidae). Journal of Insect Behaviour, 7, 465 481.
  • Van Roermund, J.W., Hemerik, L., Van Lenteren, J.C. 1994 Influence of intrapatch experience and temperature on the time allocation of the whitefly parasitoid Encarsia formosa (Hymenoptera: aphelinidae). Journal of Insect Behavior, 7, 483 499.
  • Siegel, S. & Castellan, N.J. Jr 1988Nonparametric Statistics for the Behavioral Sciences. McGraw-Hill, New York.
  • Van Steenis, M.J., El-Khawass, K.A.M.H., Hemerik, L., Van Lenteren, J.C. 1996 Time allocation of the parasitoid Aphidius colemani foraging for Aphis gossypii on cucumber leaves. Journal of Insect Behavior, 9, 283 295.
  • Steinberg, S., Dicke, M., Vet, L.E.M., Wanningen, R. 1992 Response of the braconid parasitoid Cotesia glomerata to volatile infochemicals: effects of bioassay set-up, parasitoid age and experience and barometric flux. Entomologia Experimentalis et Applicata, 63, 163 175.
  • Stephens, D.W. & Krebs, J.R. 1986Foraging Theory. Princeton University Press, Princeton, New Jersey.
  • Strand, M.R. & Vinson, S.B. 1982 Behavioral response of the parasitoid Cardiochiles nigriceps to a kairomone. Entomologia Experimentalis et Applicata, 31, 308 315.
  • Vet, L.E.M., Lewis, W.J., Cardé, R.T. 1995 Parasitoid foraging and learning. Chemical Ecology of Insects, 2nd edn.(eds W.Bell & R. T.Cardé), pp. 65 101. Chapman & Hall, London.
  • Waage, J.K. 1979 Foraging for patchily distributed hosts by the parasitoid, Nemeritis canescens. Journal of Animal Ecology, 48, 353 371.
  • Wiskerke, J.S.C. & Vet, L.E.M. 1994 Foraging for solitarily and gregariously feeding caterpillars: a comparison of two related parasitoid species (Hymenoptera: Braconidae). Journal of Insect Behavior, 7, 585 603.

Received 6 January 1997;revision received 17 December 1997