Host Specifity Testing: Why Do We Do It and How We Can Do It Better

Proceedings: Host Specificity Testing of Exotic Arthropod Biological Control Agents: The Biological Basis for Improvement in Safety

Host Specificity Testing: Why Do We Do It and How We Can Do It Better

Rieks Dekker van Klinken CSIRO Entomology, PMB 3, Indooroopilly Queensland, Australia 4068 Email: [email protected] Fax: (07) 3214 2885

Abstract Host specificity testing is universally used in weed biological control to predict non- target effects of potential agents. Despite this, there is some confusion regarding the role of host specificity testing in making such predictions. One possible role is as an assay of field host range. In this case, the ideal host specificity test will simulate conditions encountered in the field, and the result (the estimated field host range) will be judged according to how accurately it matches the realized field host range. An alternative approach is to separate the description of innate host specificity (which includes fundamental host range, the relative acceptability and suitability of hosts, the ability to learn, and time dependent effects) from the prediction of how it will be expressed in the post-release environment (in terms of field host range and relative attack). In this case, host specificity testing is used to describe properties of the insect, which are then used in combination with ecological information to predict where, when, and to what extent non-target attack would occur. I argue that the latter approach is more powerful because non-target effects under any particular environmental conditions are predicted, rather than being estimated by attempting to experimentally simulate the release environment.

Here I discuss this more basic approach to host specificity testing in some detail in relation to the meaning of the terms host specificity and host range, and I point out the implications of this approach for the way that we conduct host specificity testing. My approach to host testing can be divided into three steps: (1) identification of aspects of life history that need to be host-specific if the insect is to be safe for release; (2) description of the fundamental host range of the organism; and (3) if non-target species are included within the fundamental host-range, prediction of whether they will be attacked under field conditions and the frequency and severity of such attacks.

Keywords: biological control, host specificity testing methodology, innate host specificity, fundamental host range, realized host range, time-dependent effects, learning.

54 Host Specificity Testing: Why We Do It and How We Can Do It Better Proceedings: Host Specificity Testing of Exotic Arthropod Biological Control Agents: The Biological Basis for Improvement in Safety

Introduction 1a). In this approach, trials conducted under field All potential weed biological control agents need to conditions are considered ideal because they are most undergo extensive host-specificity testing to ensure that realistic or “natural” (Wapshere, 1989; Cullen, 1990; their release would not result in unacceptable non-target Briese, 1999), although in practice laboratory trials are impact. The biology of each potential agent is different, often necessary. Surveys and experiments seek to estimate which means that the experimental methods used have to the likely field host range in the proposed release be modified for each species to ensure that our predictions environment. These methods are judged as successful if of non-target attack are as accurate as possible. Decisions their predictions are accurate. For example, test designs are about testing include which aspects of the insect’s life judged according to their likelihood of overestimating field history to focus on, what experimental designs to use, what host range (and thus generating “false positives”) or combinations of tests to apply, whether to apply tests to the underestimating field host range (and generating “false entire plant test list or just to a subset of it, the order in negatives”) (Marohasy, 1998; Edwards, 1999; Hill, 1999; which tests are conducted, and the balance between Heard, 2000). laboratory and field trials. Each of these decisions can This assay-based approach has a number of limitations. potentially affect the accuracy of our predictions. However, One limitation is that it is very difficult to simulate the it is the purpose of this paper to consider the more field conditions that an agent would encounter in its fundamental issues of what role host specificity testing can introduced range, particularly in laboratory trials. A and should play in the prediction of non-target attack, and second limitation is that, even if accurate simulation what that means in practice. These are important issues, were possible, the introduced range is likely to be particularly as the scientific credibility of biological control heterogeneous with respect to the relative availability of and the accuracy of its predictions, come under increasing target and non-target hosts, and this in turn can scrutiny (Thomas and Willis, 1998). significantly affect relative attack (Courtney and Kibota, There are essentially two philosophical approaches to host 1989). Estimates of relative effects on various non-target specificity testing. The first seeks to predict non-target plants that are obtained through simulation assays apply impact through experiments that attempt to simulate the to specific sets of field or experimental conditions and field conditions likely to be encountered post-release (Fig. therefore it may be difficult or impossible to generalize

A. Assay Role Prediction of Post Outcome of Host = Release Outcome Specificity Testing n Field host range n Relative attack B. Description Role Description of Innate Host Specificity Outcome of Host n Fundamental host range Specificity Testing = n Relative acceptability/suitability of hosts n Learning mechanisms Prediction of Post Release Outcome n Field host range n Relative attack Description of Release Environment n Relative availability of hosts n Host quality n Abiotic factors

Fig. 1. Diagrammatic representation of the two roles host specificity testing could play in the prediction of relative non-target attack. The first seeks to experimentally simulate field conditions. The second seeks to describe the insect’s innate host specificity, and use this description in combination with knowledge of the post-release environment to predict relative non-target attack.

Host Specificity Testing: Why We Do It and How We Can Do It Better 55 Proceedings: Host Specificity Testing of Exotic Arthropod Biological Control Agents: The Biological Basis for Improvement in Safety

such predictions to fit other conditions. A good an agent in a new environment (van Klinken, 1999a). illustration of this dilemma is the controversy In this paper I discuss the second approach. I first look surrounding the appropriate use of no-choice and at the terms “host range” and “host specificity” and choice trials (Harley, 1969; Cullen, 1990; Blossey, how they relate to the innate host finding and 1995; Harris, 1998; Edwards, 1999; Hill, 1999; accepting abilities of the insect, and to their expression Sheppard, 1999). Proponents of choice trials argue that under field conditions. I finish by examining they more realistically represent field conditions and methodological implications of this approach for the that there is a danger of no-choice trials generating “false way we go about predicting non-target attack. positives”. Proponents of no-choice trials argue that choice trials can generate “false negatives”, because the agent won’t necessarily be faced with a choice in the What is Host Specificity field. In reality, both arguments could be correct, and Host Range? sometimes. It will depend on the relative availability of The terms host specificity and host range are basic target and non-target hosts, which could vary from all to the biological control lexicon, and it is weed to all non-target species, with all possible ratios in important to understand what each means in between. relation to both the innate capabilities of the An alternative philosophical approach to host- insect and what actually happens in the field. specificity testing is to conduct experiments in order to Host specificity is used to rank insect species describe the innate host-specificity of the insect (Fig. within a continuum, from specialists to so-called 1b). To achieve this goal, we need to describe what generalists (Fig. 2). It is commonly used plant species an agent is capable of finding, accepting synonymously with host-range breadth. and using and how well it can do so, taking into However, the host-specificity of an insect can be account the plasticity of behavioral responses to further differentiated according to how deprivation and prior experience. Information thus acceptable or suitable hosts are relative to each gained can be used to predict non-target attack under other. For example, an insect that performs the full spectrum of environmental conditions the equally well on all host species would be less host insect would be likely to encounter once released (Fig. specific than an insect for which only one of the 1b). Such an approach also allows the host specificity of same range of species is an ideal host, even though insects to be compared more objectively (van Klinken, host-ranges are identical (Fig. 2). There are in press) and provides a means for assessing the therefore two dimensions to quantifying how possibility of host-specificity evolving after the release of

Fig. 2. A hypothetical gradation of host specificity from specialist to generalist. Host specificity is described with two axes, host range and relative acceptability/suitability of those hosts. Insects B and C have identical host ranges (n=5) but differ in the relative acceptability/suitability of their hosts.

56 Host Specificity Testing: Why We Do It and How We Can Do It Better Proceedings: Host Specificity Testing of Exotic Arthropod Biological Control Agents: The Biological Basis for Improvement in Safety host-specific an insect is - host range breadth, and the observed in experiments is frequently broader than relative acceptability or suitability of hosts. what occurs in the field (Shepherd, 1990; Olckers, 1999). Host range can even differ across an insect’s This two-dimensional concept of host-specificity is geographic range (Hodkinson, 1997). implicit in host-specificity testing, which aims to both define the host-range and obtain comparative data One way to deal with this problem is to differentiate among hosts. However, this usage of the term differs between fundamental and realized host ranges from that in the behavioral literature, in which (Nechols et al., 1992). The fundamental host range is specificity refers specifically to differences in the the most inclusive host range because it includes all discriminatory phase, as defined by Singer et al. (1992). the plant species that an insect is capable of accepting As will be seen, the discriminatory phase (the time over and/or utilising. It therefore represents the genetically which an individual accepts one plant while the lower- determined limits to the host range of a particular ranked plant is rejected) is only one of many possible insect species or, more precisely, insect genotype. The ways in which the host-specificity of insects can be realized host range is how the fundamental host described. Limiting the description of host-specificity to range is actually expressed under particular comparisons of discriminatory phases is therefore conditions (Nechols et al., 1992). In biological unnecessarily restrictive. control we are concerned with predicting how the fundamental host-range will be realized if the agent Host Range were to be released (the field host range). In the simplest terms, the host-range of an insect is the Fundamental host range. The absolute limits to an sum of plant species (or more precisely plant insect’s host range, which circumscribe fundamental phenotypes) that are hosts. Host-range breadth will host range, are constrained by such factors as its depend on the relatedness of those hosts. For example, metabolic and sensory capabilities, physical limitations herbivores are commonly categorised as being and behavioral programming. For example, the location monophagous, oligophagous, and polyphagous, and acceptance of a host for oviposition is determined according to the degree of taxonomic relatedness of by an often complex catenary sequence of behaviours their hosts (Symons and Beccaloni, 1999). However, (Miller and Strickler, 1984; Wapshere, 1989). For some describing the host-range of an insect can be insects, this is highly constrained, with only a single complicated by the fact that host-range is sometimes plant species being accepted even when the insect is dependent on context. For example, the host-range highly deprived and is offered no alternative (Adair and

Life History Stage Fundamental Host Range

Egg Hatch All test plants (2 families)

Nymphal Development Tribe Mimoseae (Prosopis and Early (instars 1-3) Neptunia)

Late (instars 4-5) Prosopis

Adult Feeding (for oogenesis and Prosopis longevity)

Oviposition Possibly coupled with adult Host finding cues feeding (i.e., Prosopis)

Final acceptance All physically suitable test plants

Fig. 3. Fundamental host range estimated for different aspects of the life history of the psyllid, Prosopidopsylla flava Burckhardt (van Klinken, in press). P. flava inserts eggs into plant tissue and oogenesis occurs as adults. Host range breadth is represented schematically.

Host Specificity Testing: Why We Do It and How We Can Do It Better 57 Proceedings: Host Specificity Testing of Exotic Arthropod Biological Control Agents: The Biological Basis for Improvement in Safety

Scott, 1997). Constraints are likely to be primarily 1987; Slansky and Rodriquez, 1987). The behavioral, although physical factors such as ovipositor fundamental host range for larval development is length may also be constraining (Zwölfer and Harris, therefore the consequence of all of these constraints 1971). Similarly, nutritional requirements and combined. Even if desired, it is questionable whether metabolic limitations are likely to be important we could decouple physiological constraints from all constraints to larval and adult development, although other constraints. behavioral mechanisms can also play a critical role in Field host range. Field host range is what actually determining whether an insect will commence feeding happens in the field. Like fundamental host range, it on an otherwise suitable host (Scriber, 1984; Slansky can be described for different aspects of the herbivore’s and Rodriguez, 1987). life history. For example, field host range can differ In theory, fundamental host range can be described for between oviposition, breeding and adult feeding. An any aspect of the insect’s life history where interactions extreme example of differences between oviposition and between insect and host occur. For example, the breeding host range is a hepialid moth that oviposits fundamental host range could be described separately indiscriminately over pastures (Barton Browne et al., for oviposition, egg development, larval development 1969). Similarly, some psyllid species feed as adults on and adult feeding (Fig. 3). At a more refined level it more plant species in the field than they breed on could be described for each behavioral step within the (Hodkinson, 1974). catenary sequence of behaviors resulting in host Under field conditions, the realized host-range is acceptance (Wapshere, 1989; Keller, 1999). For most, frequently a subset of the fundamental host-range. if not all, insects the fundamental host range will differ That is, insects often accept or use only a proportion of dramatically for different aspects of its life history. For those that they are capable of (Harris and McEvoy, example, the pre-alighting cues of some insects such as 1995; Wapshere, 1989; Cullen, 1990). There are certain aphids can be very general, and they several possible reasons for the fundamental host range discriminate primarily on the basis of post-alighting not being fully expressed (Fig. 4). cues (Kennedy et al., 1959). Other insects appear to depend more on pre-alighting cues for host selection Clearly, for an insect to locate and use a host it must be (Barton Browne et al., 1969; Johnson and Siemens, sufficiently close to detect the necessary cues (Cullen, 1991). The fundamental host range could also be 1990). Spatial coincidence will depend on the described for the life cycle of the insect (i.e. the plant geographic distributions of the insect and host, which species that fulfill all requirements for life and in turn are determined by factors such as their reproduction), which would represent the intersection respective abiotic requirements and the absence of of fundamental host ranges for each particular aspect of geographic barriers. The distribution of the insect may life history. For the psyllid, P. flava, it would be the also depend on the availability of plant species that can genus Prosopis (Fig. 3). support a population. More locally, strong habitat preferences may prevent insect and host from co- In biological control, an insect’s maximal host range is occurring. For example, many insect species appear to frequently described as its “physiological host range” restrict their search for hosts to particular vegetation (Cullen, 1990; McEvoy 1996; Olckers, 1998). This is types (Kibota and Courtney, 1991; van Klinken, misleading as it implies that the innate capacity of an 1996). A potential host must also be available at the insect to accept and use a host is constrained only by correct time. For example, the weevil “P. cordister” physiology. This is certainly not the case, even with requires rootstock in summer, and would therefore be larval development, for which the term physiological unable to use winter annual species, even though some host range is most often applied. Larval development can support larval development (Cullen, 1990). depends on the insect having innate behavioral responses to initiate and continue feeding, having Even if a potential host is available, it may never be nutritional requirements that can be met by the plant, used because it is not included in the fundamental host having the physical ability to consume sufficient plant range of a prior step in the insect’s life history. For material to obtain necessary nutrients, and having the example, a plant species on which larvae feed in the metabolic, behavioral and other capabilities to overcome laboratory may never be used in the field because any toxic properties (Scriber, 1984; Schoonhoven, females do not recognize it as a host for oviposition

58 Host Specificity Testing: Why We Do It and How We Can Do It Better Proceedings: Host Specificity Testing of Exotic Arthropod Biological Control Agents: The Biological Basis for Improvement in Safety

(Wan and Harris, 1996). Similarly, host-specific pre- the most acceptable host was relatively rare, then insects alighting cues might determine that only a small subset might become sufficiently deprived to begin accepting of available plant species is actually assessed through the lower ranked hosts. If prior experience results in a contact cues (Frick and Andres, 1967; Wapshere, reversal in preference rank, then insects may in fact find 1989). Alternatively, a potential host may never be used the most abundant host more acceptable, regardless of because a more acceptable host is always available. their previous rank. However, regardless of the Behavioral reasons for this include effects of prior mechanism, relative attack rates can be profoundly experience and time dependent effects, and these are influenced by relative host availability (Thompson, discussed further in the following section. 1988; Blossey et al., 1994; Aeschlimann, 1997; Withers, 1998). Relative Acceptability or Suitability, and the Effect of Internal Status Host Specificity Testing The relative acceptability or suitability of all hosts Methodology within the host range can be compared to give a more complete picture of an insect’s host-specificity (Fig. 2). Non-target attack in the field is the consequence of the Like fundamental host range, hosts can be, and have interaction between an insect’s innate host-specificity been, compared for many aspects of the insect’s life and the environment (Fig. 1b). Host specificity testing history. For example, hosts can be compared according can describe innate host specificity in terms of to how acceptable they are for oviposition and initiation fundamental host ranges, the relative acceptability or of feeding, or for support of pre-reproductive and suitability of each host, and how that is affected by reproductive development (Marohasy, 1998). The changes in internal status. Exactly which aspects of the relative acceptability of hosts can also be compared for insect’s innate host specificity need to be described will particular behaviors within the host location and depend on what we need to know in order to be acceptance sequence (Kennedy, 1965; Courtney and confident that there will be no “undue” non-target Kibota, 1989; Keller, 1999). effects. When combined with knowledge of the release environment, these results can be used to predict field However, unlike fundamental host range, the relative host range, and if relevant, when, where and to what acceptability and suitability of hosts to the insect is a extent non-target attack will occur. dynamic property. It can be profoundly influenced by the internal status of the insect, particularly through This general approach to host specificity testing can be time-dependent processes and effects of prior summarized as a three-step process. Firstly, the aspects experience (Solarz and Newman, 1996; Newman et al., of the life history that needs to be host-specific are 1999; Heard, 2000; Withers et al., 2000). For example, identified in order to determine exactly what aspects of deprivation can result in the acceptance of previously the innate host-specificity need to be described. unacceptable hosts (Withers et al., 2000), and prior Secondly, the fundamental host range is estimated for experience can reverse preference rankings (Hanson, each such aspect. Finally, a prediction of the non-target 1976; Szentesi and Jermy, 1990). Behavioral plasticity consequences if the insect were to be released is made. is an innate property of the insect (Fig. 1b) and can be The latter step may include further description of the described experimentally. insect’s host-specificity such as the relative acceptability and suitability of hosts, and the effects of experience Behavioral plasticity means that the expression of and deprivation. behaviours under field conditions can be complex (Rausher, 1980; Prokopy et al., 1987; Courtney and In practice some steps might overlap. For example, Forsberg, 1988). For some insects, prior experience and comparative data (step 3) are often obtained as a by- time dependent effects such as deprivation are likely to product of determining the fundamental host range, be greatly influenced by encounter rates with potential and host-specificity testing often provides further hosts, which in turn will depend on the relative insights into the insect’s life history (step 1). Native availability of hosts (Prokopy and Lewis, 1993). For range studies might also be conducted before example, where the most acceptable host is relatively determining the fundamental host range. This serves abundant, insects are less likely to become deprived the dual purpose of ranking potential agents for enough to feed on a less acceptable host. Conversely, if subsequent study (according to their likely specificity,

Host Specificity Testing: Why We Do It and How We Can Do It Better 59 Proceedings: Host Specificity Testing of Exotic Arthropod Biological Control Agents: The Biological Basis for Improvement in Safety likely impact and amenability to laboratory work using, regardless of the field conditions it may [Schroeder and Goeden, 1986; Briese 1999]) and encounter. The better our estimate, the less chance we obtaining a better understanding of their life history. have of inadvertently excluding possible non-target field hosts. Carefully chosen plant test lists and Step 1. Identifying What Needs to be Host Specific experimental design will ensure our estimates are as The first step to host-specificity testing is identifying accurate, and therefore as inclusive, as possible. which aspects of the insect’s life history need to be host Theoretically, the fundamental host-range includes all specific. The requirement for host specificity will often the plant species (or more specifically, phenotypes) that depend on the life history of the insect and where it is are hosts. Given that it is not possible (or desirable) to to be released. The completion of larval development is test the total flora, it is necessary to subsample. The an essential step in the life cycle of all insects, and for centrifugal-phylogenetic method is generally applied most insects larval feeding is also the most damaging (Wapshere, 1989), although it assumes that host-range aspect. Complete larval development, and possibly will correlate with phylogenetic relatedness, which may merely larval feeding, on non-target species in the field not always hold (Weidemann, 1991). Sometimes plant is therefore of primary concern. However, other aspects traits, such as plant architecture, that are not necessarily may also have to be considered. Late instar larvae can correlated to plant phylogeny, may therefore also need sometimes feed on more plant species than neonate to be considered. Where possible, the quality of test larvae, and this may be important if there is a risk of plants should reflect what would be encountered in the them dispersing onto new plants (Cullen, 1990). field by using intact plants of the right age and Where ovipositing females damage their host (such as reproductive stage. Sometimes particular fertilizer twig-girdlers), oviposition on non-target species could regimes (Cuda et al., 1995; van Klinken, 1999b) or be a potential problem, even if larval development prior exposure to sun (Cullen, 1990) may be necessary. cannot occur. Similarly, adult feeding on non-target species may be a concern, even if it does not result in Both time dependent effects (Papaj and Rausher, 1983) oogenesis. Even exploratory feeding on non-target and effects of prior experience (Szentesi and Jermy, species could be a problem where adults are known 1990) can limit the full expression of the fundamental virus vectors (Briese, 1988). Thus, for some insects we host-range (Marohasy, 1998; Heard, 2000; Withers et have to ensure that more than one aspect of their life al., 2000). “Maximum likelihood” tests must therefore history is sufficiently host-specific. be designed to exclude this possibility. Time dependent effects can be excluded by conducting no-choice trials The potentially damaging aspects of an insect’s life for the duration of the insect’s life. A no-choice design history need not, however, be studied directly. ensures that there is no alternative, more acceptable Sometimes the potentially damaging aspects of an host, to confound the insect’s response, and conducting insect’s life history are preceded by prerequisite trials for the duration of the insect’s life ensures that the behavioral or developmental steps (Wapshere, 1989), insect will be sufficiently deprived to accept a poorer and these steps could be studied instead. For example, host. Possible effects of prior experience can be if larvae depend on their mother to select the right host, it might be sufficient just to study oviposition (Heard Table 1. Possible methods for obtaining rapid, et al., 1997). In some cases it may even be necessary to yet accurate, estimates of host rangesa study the pre-requisite step, such as when the damaging step is not easily studied (for example when Single test conducted on complete plant list culturing is difficult) or is not sufficiently host-specific • Examine limited part of life history (e.g., first feeding (Wapshere, 1989; Harris and McEvoy, 1995). instar) • Choice minus target trials Step 2. Estimating Fundamental Host Range Compromised test on complete plant list, test Given that we have identified the parts of the life assumptions on a subset history for which we want to determine host-specificity, • Use experienced individuals the fundamental host range can then be described for each. Since the fundamental host range represents the • Limit the duration of tests limits of an insect’s ability, estimating it will identify all *See text for benefits and drawbacks of each test. the plant species an insect is capable of accepting or

60 Host Specificity Testing: Why We Do It and How We Can Do It Better Proceedings: Host Specificity Testing of Exotic Arthropod Biological Control Agents: The Biological Basis for Improvement in Safety excluded to some extent by using neonate larvae and assumption that prior experience will not limit the newly emerged adults. Further precautions against insects’ ability to feed on a test plant can be tested experience effects can be taken by washing eggs or separately using naive insects on a subset of plant dissecting out pupae. However, examples where prior species. Similarly, for insect species that are particularly experience has resulted in an irreversible change in long-lived, or are in short supply, conducting trials until fundamental host-range are rare (Ma, 1972; Renwick all individuals are dead may not be cost-effective. In and Lopez, 1999). this case compromises can be made. For example, no- choice oviposition or feeding trials can be conducted Conducting experiments to determine necessary sequentially (Heard and van Klinken, 1998), so that fundamental host ranges is relatively simple and rapid each plant is exposed for much less than the duration of for some insect species, particularly for those which the insect’s life. It is however possible that these tests express their most discriminating host finding and will result in an underestimation of the fundamental acceptance behaviours under laboratory conditions host range because insects never become sufficiently (e.g., Gassmann and Tosevski, 1994; Adair and Scott, deprived to accept the non-target, or because of effects 1997; van Klinken, in press). However, for other of prior experience. Once again, both assumptions insects, the required experimental design may not be could be tested on a subset of plant species. possible in practice, or at the very least it might be unnecessarily strict and time-consuming. There are Fundamental host ranges are already being estimated in several ways of streamlining the estimation of most host-specificity studies under the guise of fundamental host range (Table 1). “laboratory host-range”, “physiological host-range” or “experimental host-range” (Zwölfer and Harris, 1971; One approach is to describe the fundamental host Gassmann and Tosevski, 1994; Olckers, 1998; Purcell range for a limited part of the insect’s life history. For et al., 1998; Hill et al., 1999). However, the aspect of example, the fundamental host range for larval the insect’s life history for which the fundamental host development could be estimated for just the first range is being described is often not stated, and factors feeding instar, rather than for complete development. If that could potentially limit the full expression of host larvae do develop through to the next instar a separate range are rarely, if ever, explicitly excluded. Improved trial could be run to determine if complete estimates of fundamental host ranges can generally be development would occur on those species (e.g., van made by stating what it is that the fundamental host Klinken, 1999b). Another potential approach when range is being described for, and by explicitly removing studying host-acceptance is to conduct choice minus factors that could result in it not being fully expressed. target trials (Heard and van Klinken, 1998; Edwards, Although it will not always be possible to entirely 1999). If no plants are accepted, then the trial exclude such factors, what would result is the best effectively becomes a no-choice trial. However, if at possible estimate of what plant species the insect is least one plant species is attacked then the trial would capable of accepting and/or using. have to be repeated without that species to ensure lower ranked hosts have not been missed (Peschken and Step 3. Extrapolating to the Field Derby, 1988; Marohasy, 1998). The compromise in Some insects have fundamental host-ranges which are this trial design is that deterrents from one plant species essentially restricted to the target, and no further work could completely mask an otherwise acceptable host is therefore required (e.g., Heard et al., 1997; van (Marohasy, 1998), although I am not aware of any Klinken, in press; van Klinken and Heard, in press). cases where this has been documented. The possibility However, if the insect is capable of attacking non-target could be minimised by randomising the combination species in a way that is considered potentially of test plants presented in any one test. detrimental, then several avenues exist in order to Another approach to obtaining more flexibility in predict what will happen after release. Predictions can experimental design is to apply a less strict design on be made as to whether those non-target species will the entire plant test list, but then to test any resulting actually be attacked in the field, the relative and assumptions on a subset of plant species (Table 1). For absolute level of such attack, and its consequences. example, if newly emerged adults are difficult to obtain Only the prediction of field host range and relative it will sometimes be easier to use experienced adults non-target attack are considered below. The population when conducting feeding trials for life. The additional dynamics of the insect would need to be predicted in

Host Specificity Testing: Why We Do It and How We Can Do It Better 61 Proceedings: Host Specificity Testing of Exotic Arthropod Biological Control Agents: The Biological Basis for Improvement in Safety

larval hosts would not be expected if oviposition were restricted to the target and there were no possibility that larvae would disperse onto other larval hosts. Similarly, it is sometimes argued that species in which there is oviposition on non-target species in cages will still be host-specific in the field because distance cues that are effectively bypassed in cages are specific to the target. However, in each case the host-specific step must indeed be prerequisite, which may be difficult to prove conclusively in some cases. Fig. 4. Possible reasons for the fundamental host range (e.g., for complete A potentially confounding factor in larval development or adult feeding) not being fully realized in the field. using estimated fundamental host range to predict field host range is any difference between test plants and order to translate predictions of relative attack to one of field plants that is biologically important. For example, absolute attack. Predicting the consequence of any non- Monochoria species supported complete development target attack in terms of ecological, economic, social and/ of Xubida infusellus (Walker) in the laboratory, but it or political impact are considered elsewhere (McFadyen, was predicted that complete development would not 1998; Waage and Kirk, 1999). occur in the field because plants are shorter-lived than Predicting field host range. The fundamental host those grown in the laboratory (Stanley and Julien, range will not necessarily be fully expressed in the 1998). This represents a limitation imposed on the field (Fig. 4). In practice, the main arguments for fundamental host-range estimate (because not all plant potential non-target hosts not being attacked after phenotypes were tested), rather than a discrepancy release are lack of coincidence with the potential between fundamental and field host ranges. distribution of the agent, and host-specificity of Predicting relative non-target attack. If non-target prerequisite steps. Lack of coincidence between insect species are likely to be attacked, then we can predict and potential host can be demonstrated by predicting where, when and to what extent this may occur by the potential distribution of the insect and comparing considering the relative acceptability and suitability of it with existing non-target distributions (Heard and hosts, and how that might be expressed in the release Forno, 1996). However, potential changes in the environment (Fig. 1b). If non-target attack is distribution of both the target (if agent survival sufficiently “minor,” the agent could still be considered depends on it) and the non-target hosts (e.g., through for use in biological control (Day, 1999; Hill, 1999). changing land use) may also need to be considered. The theory and methodology necessary for describing At a more local scale, vegetation associations of the relative acceptability and suitability of hosts and how it agent and non-target species, and the phenology of can be influenced by the internal status of the insect the agent in relation to the non-target species, could (through time dependent effects and prior experience) be used to argue that non-target species will not be has received much attention in the general literature attacked (Frick and Andres, 1967; Harris and (Bernays and Chapman, 1994; Eigenbrode and McEvoy, 1995). Bernays, 1997), and in the biological control literature Host-specificity of prior steps in an insect’s life history or (Zwölfer and Harris, 1971; Marohasy, 1998; Heard, host selection behaviour might limit or prevent an 2000; Withers et al., 2000). otherwise suitable non-target host from being attacked Acceptability and suitability can be compared in in the field (Harris and McEvoy, 1995; Wan and numerous ways, including acceptability for oviposition Harris, 1996). For example, larval attack on non-target (during the discriminatory phase) (Singer et al., 1992;

62 Host Specificity Testing: Why We Do It and How We Can Do It Better Proceedings: Host Specificity Testing of Exotic Arthropod Biological Control Agents: The Biological Basis for Improvement in Safety

Eigenbrode and Bernays, 1997; Marohasy, 1998; acceptance and use of different hosts (Wiklund, 1981; Withers, 1998), total fecundity (where oogenesis results Papaj and Rausher, 1983; Singer et al., 1993). from larval feeding) (van Klinken and Heard, in press) Although rarely documented, such variation can have and preference rank (Wiklund, 1975, 1981; Marohasy implications both in terms of immediate post-release 1998; Stanley and Julien, 1999). Similarly, the attack and the rapid evolution of insect preferences and suitability of plant species for complete larval performances (Thompson, 1998). development can be compared by determining the When predicting relative attack in the field, relative proportion of neonates which develop to adults, acceptability and suitability must be considered in comparing growth parameters such as relative growth terms of the relative availability of target and non-target rate and developmental times, or comparing measures of hosts. The simplest case is if non-target and target “fitness” such as total fecundity, size and weight of populations are far enough apart such that the non- emerging adults (e.g., Tabashnik, 1983; Wan et al., target has to be a “sufficiently good host” to sustain a 1996). More detailed studies can be conducted to viable population (Heard and Forno, 1996). Prediction compare relative growth rates within instars, efficiency is not so straightforward if target and non-target species of conversion of ingested and digested food, and overlap. Plant (and insect) populations are typically approximate digestibility (Berenbaum and Zangerl, heterogeneous and dynamic, and this needs to be 1991). understood in order to predict what environments the The way hosts are compared will depend to a large insect is likely to encounter post-release. Heterogeneity degree on experimental design. No-choice trials can be is particularly obvious for annuals, or in cases in which used to compare various traits such as larval survival and the weed is eventually brought under effective control development rates, the amount of adult feeding and in parts of its range. Under these circumstances, the resulting fecundity, and rates of natural increase (e.g., challenge is to predict how availability of hosts will Blossey et al., 1994; Wan and Harris, 1997). Often affect their relative acceptability and suitability, and these comparative data can be obtained when thus relative attack. estimating the fundamental host range. Direct In practice, concluding that a potential host is safe to (continuous) observation or temporal sampling can release, despite the inclusion of non-target species in the often provide additional information such as duration field host range, will be easiest where differences of the discriminatory phases and feeding bouts (Solarz between relative acceptability and suitability are great. and Newman, 1996; Eigenbrode and Bernays, 1997; Where non-target and target are likely to be sympatric, Withers, 1998; Singer et al., 1993). Choice trials can be cases in which behavioral plasticity is limited would be used to rank hosts at particular relative densities easiest to interpret. (Wiklund, 1981; Marohasy, 1998; Briese, 1999; Stanley and Julien, 1999). In many experiments the internal status of the test insect can be manipulated in Concluding Remarks order to see how it influences relative acceptability and Although host specificity testing is central to the suitability. For example naive and experienced insects prediction of non-target attack, confusion remains can be compared (van Klinken, in press), as can insects regarding its precise role. One approach is to view it as a at different levels of deprivation (Withers et al., 2000). direct estimate of field host range and relative attack. Two potential qualifiers to the description of relative The primary limitation of this approach is that acceptability and suitability are effects of plant quality potentially profound effects of environmental variation and intra-specific variation. Plant quality can differ (such as changes in relative host availability) on relative between and among test plants and field plants in ways attack can only be determined by estimation, not that affect their relative acceptability and suitability as prediction. That is, experiments need to realistically hosts (Lowman and Box, 1983; Leather, 1989; Waring simulate each of the possible environments that an and Cobb, 1989; Cullen, 1990; Price et al., 1990; van insect is likely to encounter post-release. The alternative Dam and Hare, 1998; Baars and Neser, 1999; van approach is to use host specificity testing to describe the Klinken, 1999b). This can make the interpretation of insect’s innate host-specificity, which might include its experimental results difficult. Similarly intra-specific fundamental host range, and how the relative genetic variation among herbivores can result in acceptability and suitability of hosts are influenced by dramatic differences among individuals in their changing internal status. The strength of this approach

Host Specificity Testing: Why We Do It and How We Can Do It Better 63 Proceedings: Host Specificity Testing of Exotic Arthropod Biological Control Agents: The Biological Basis for Improvement in Safety

is that it concentrates on describing properties of the References insect, which can in turn be used to accurately predict Adair, R.J. and J.K. Scott. 1997. Distribution, life history relative attack under any likely post-release conditions. and host specificity of Chrysolina picturata and This second approach can be translated into a Chrysolina sp. B (Coleoptera: Chrysomelidae), two methodology for host specificity testing which biological control agents for Chrysanthemoides produces generalizable results with which to make monilifera (Compositae). Bulletin of Entomological accurate predictions of non-target attack in the field. Research 87:331-341. Fundamental host range, which represents the absolute Aeschlimann, J.P. 1997. Reappraising the potential of limits of the insect’s innate host specificity, is described biological control against the weed Carthamus lanatus. first for aspects of the insect’s life history that need to be Entomophaga 42:559-568. host specific. If non-target species are included, Baars, J.R. and S. Neser. 1999. Past and present initiatives on predictions can be made as to whether non-target the biological control of Lantana camara (Verbenaceae) species within the fundamental host range will indeed in South Africa. In Olckers, T. and M. P. Hill (eds.) be attacked in the field. If they will be, further host Biological Control of Weeds in South Africa (1990-1999). specificity testing can be conducted in order to describe African Entomology Memoir 1:21-33. relative acceptability and suitability of the different Barton Browne, L., C.F. Soo Hoo, A.C.M. van Gerwen hosts and how possible learning or time dependent and I. R. Sherwell. 1969. Mating flight behaviour in mechanisms modify them. These results can be used, three species of Oncopera moths (Lepidoptera: together with a detailed knowledge of the release Hepialidae). Journal of the Australian Entomological environment, to predict when, where and to what Society 8: 168-172. extent non-target attack is likely to occur. Berenbaum, M.R. and A.R. Zangerl. 1991. Acquisition This approach differs from existing experimental of a native hostplant by an introduced oligophagous approaches in one or more of the following ways. It herbivore. Oikos 62:153-159. distinguishes between the innate capacity of an insect Bernays, E.A. and R.F. Chapman. 1994. Host-Plant to interact with plants, and how that innate capacity is Selection by Phytophagous Insects. Chapman and Hall, expressed under particular field conditions in terms of New York. field host range and relative attack (Figs. 1b, 4). It Blossey, B. 1995. Host-specificity screening of insect describes host specificity as having two dimensions, biological weed control agents as part of an host range and the relative acceptability and/or environmental risk assessment, pp. 84-89. In suitability of hosts (Fig. 2). It acknowledges that Hokkanen, H. M. T. and J. M. Lynch (eds.) Biological fundamental host range can be described for any aspect Control: Benefits and Risks. Cambridge University of an insect’s life history where the insect interacts with Press, Cambridge, United Kingdom. plants (Fig. 3). It accounts for possible behavioral plasticity resulting from prior experience or time Blossey, B., D. Schroeder, S.D. Hight and R.A. Malecki. dependent changes in internal status. Finally, it views 1994. Host specificity and environmental impact of the role of host specificity testing as describing innate two leaf beetles (Galerucella calmariensis and G. pusilla) host specificity (including fundamental host range, for biological control of purple loosestrife (Lythrum relative acceptability and suitability of hosts, and salicaria). Weed Science 42:134-140. behavioral plasticity), rather than predicting field host Briese, D.T. 1988. Host-specificity and virus-vector range (Fig. 1). potential of Aphis chloris Koch (Hemiptera: Aphidae), a biological control agent for St. John’s wort in Australia. Entomophaga 34: 247-264. Acknowledgments Briese, D.T. 1999. Open field host-specificity tests: is I thank Lindsay Barton Browne, Tim Heard, John “natural” good enough for risk assessment? pp. 44- Stanley and Gimme Walter for many stimulating 59. In Withers, T. M., L. Barton Browne, and J. discussions; and Lindsay Barton Browne, Owain Stanley (eds.) Host Specificity Testing in Australasia: Edwards, Bill Palmer, Urs Schaffner, Marijke van Towards Improved Assays for Biological Control. Klinken and Gimme Walter for their critical feedback Department of Natural Resources, Coorparoo, on earlier drafts. Queensland, Australia.

64 Host Specificity Testing: Why We Do It and How We Can Do It Better Proceedings: Host Specificity Testing of Exotic Arthropod Biological Control Agents: The Biological Basis for Improvement in Safety

Courtney, S.L. and J. Forsberg 1988. Host use by pierid Hanson, F.E. 1976. Comparative studies on induction of butterflies varies with host density. Functional Ecology food choice preferences in lepidopterous larvae. 2:67-75. Symposia Biologica Hungarica 16: 71-77. Courtney, S.P. and T.T. Kibota. 1989. Mother doesn’t Harley, K.L.S. 1969. The suitability of Octotoma scabripennis know best: selection of hosts by ovipositing insects, Guer. and Uroplata girardi Pic (Col., Chrysomelidae) pp. 161-188. In Bernays, E. A. (ed.) Insect-Plant for the control of Lantana (Verbenaceae) in Australia. Interactions, Volume II. CRC Press, Boca Raton, Bulletin of Entomological Research 58: 835-843. Florida, USA. Harris, P. 1998. Evolution of classical weed biocontrol: Cuda, J.P., H.B. Johnson and C.R. Tischler. 1995. Trophic meeting survival challenges. Entomological Society of interactions between Mozena, mesquite and a microbe: Canada Bulletin 30: 134-143. implications for host-specificity testing of insects of Harris, P. and P. McEvoy. 1995. The predictability of insect leguminous weeds, p. 575. In Delfosse, E. S. and R. R. host plant utilisation from feeding tests and suggested th Scott (eds.) Proceedings of the VIII International improvements for screening weed biological control Symposium on Biological Control of Weeds, 2-7 February agents, pp. 125-131. In Delfosse, E. S. and R. R. 1992. DSIR/CSIRO, Melbourne, Australia. Scott (eds.) Proceedings of the VIIIth International Cullen, J.M. 1990. Current problems in host-specificity Symposium on Biological Control of Weeds, 2-7 Feb. screening, pp. 27-36. In Delfosse, E. S. (ed.) 1992. DSIR/CSIRO, Melbourne, Australia. Proceedings of the Seventh International Symposium on Heard, T.A. 2000. Concepts in insect host-plant selection the Biological Control of Weeds, 6-11 March 1988, behavior and their application to host specificity Rome, Italy. MAF, Rome. testing, pp 1-10. In Van Driesche, R. G., T. Heard, Day, M.D. 1999. Continuation trials: their use in assessing A. S. McClay and Richard Reardon (eds.) Host the host range of a potential biological control agent, Specificity Testing of Exotic Arthropod Biological Agents: pp. 11-19. In Withers, T. M., L. Barton Browne, The Biological Basis for Improvement in Safety. X and J. Stanley (eds.) Host Specificity Testing in International Symposium on Biological Control of Australasia: Towards Improved Assays for Biological Weeds, July 4-14, 1999, Bozeman, Montana. USDA Control. Department of Natural Resources, Forest Service Bulletin, FHTET-99-1, Morgantown, Coorparoo, Queensland, Australia. West Virginia, USA. 95pp. Edwards, P.B. 1999. The use of choice tests in host- Heard, T.A. and I.W. Forno. 1996. Host selection and specificity testing of herbivorous insects, pp. 35-43. host range of the flower-feeding weevil, In Withers, T. M., L. Barton Browne, and J. Stanley Coelocephalapion pigrae, a potential biological control (eds.) Host Specificity Testing in Australasia: Towards agent of Mimosa pigra. Biological Control 6:83-95. Improved Assays for Biological Control. Department of Heard, T.A. and R.D. van Klinken. 1998. An analysis of Natural Resources, Coorparoo, Queensland, designs for host range tests of insects for biological Australia. control of weeds, pp. 539-546. In Zalucki, M. P., R. Eigenbrode, S.D. and E.A. Bernays. 1997. Evaluation of A. I. Drew, and G. G. White (eds.) Pest Management factors affecting host plant selection, with an emphasis - Future Challenges, Sixth Australasian Applied on studying behaviour, pp. 147-169. In Dent, D. R. Entomology Research Conference, Brisbane. The and M. P. Walton (eds.) Methods in Ecological and University of Queensland Printery, Brisbane, Agricultural Entomology. CAB International. Australia. Wallingford, Oxon, United Kingdom. Heard, T.A., R. Segura, M. Martinez, and I.W. Forno. Frick, K.L. and L.A. Andres. 1967. Host specificity of the 1997. Biology and host range of the green-seed weevil, ragwort seed fly. Journal of Economic Entomology Sibinia fastigiata, for biological control of Mimosa 60:457-463. pigra. Biocontrol Science and Technology 7:631-644. Gassman, A. and I. Tosevski. 1994. Biology and host Hill, M.P., C.J. Cilliers, and S. Neser 1999. Life history specificity of Chamaeshpecia hungarica and Ch. and laboratory host range of Eccritotarsus catarinensis astatiformis (Lep.: Sesiidae), two candidates for the (Carvalho) (Heteroptera: Miridae), a new natural biological control of leafy spurge, Euphorbia esula enemy released on water hyacinth (Eichhornia crassipes (Euphorbiaceae) in North America. Entomophaga (Mart.) Solms-Laub.) (Pontederiaceae) in South 39:237-254. Africa. Biological Control 14:127-133.

Host Specificity Testing: Why We Do It and How We Can Do It Better 65 Proceedings: Host Specificity Testing of Exotic Arthropod Biological Control Agents: The Biological Basis for Improvement in Safety

Hill, R.L. 1999. Minimising uncertainty – in support of no- McEvoy, P.B. 1996. Host specificity and biological pest choice tests, pp.1-10. In Withers, T. M., L. Barton control: how well is research on host specificity Browne, and J. Stanley (eds.) Host Specificity Testing addressing the potential risk of biological control? in Australasia: Towards Improved Assays for Biological Bioscience 46:401-405. Control. Department of Natural Resources, McFadyen, R.E.C. 1998. Biological control of weeds. Coorparoo, Queensland, Australia. Annual Review of Entomology 43:369-393. Hodkinson, I.D. 1974. The biology of Psylloidea Marohasy, J. 1998. The design and interpretation of host- (Homoptera): a review. Bulletin of Entomological specificity tests for weed biological control with Research 64:325-339. particular reference to insect behaviour. Biocontrol Hodkinson, I. D. 1997. Progressive restriction of host plant News and Information 19: 13N-20N. exploitation along a climatic gradient: the willow Miller, J. and K. L. Strickler. 1984. Finding and accepting psyllid Cacopsylla groenlandica in Greenland. Ecological host plants, 127-157. In Bell, W. J. and R. T. Carde Entomology 22:47-54. (eds.) Chemical Ecology of Insects. Chapman and Hall, Johnson, C.D. and D.H. Siemens. 1991. Expanded London. oviposition range by a seed beetle (Coleoptera: Nechols, J.R., W. C. Kauffman, and P. W. Schaefer. 1992. Bruchidae) in proximity to a normal host. Significance of host specificity in classical biological Environmental Entomology 20:1577-1582. control, pp. 41-52. In Kauffman, W. C. and J. R. Keller, M.A. 1999. Understanding host selection Nechols (eds.) Selection Criteria and Ecological behaviour: the key to more effective host specificity Consequences of Importing Natural Enemies. testing, pp. 84-92. In Withers, T. M., L. Barton Entomological Society of America, Lanham, Browne, and J. Stanley (eds.) Host Specificity Testing Maryland, USA. in Australasia: Towards Improved Assays for Biological Newman, R.M., G. Cronin, S.L. Solarz, and D.M. Lodge. Control. Department of Natural Resources, 1999. The importance of prior experience and Coorparoo, Queensland, Australia. population source in the determination of host range. Kennedy, J.S. 1965. Mechanisms of host plant selection. Evaluating Indirect Ecological Effects of Biological Annals of Applied Biology 56:317-322. Control, Global IOBC International Symposium, Kennedy, J.S., C.O. Booth, and W.J.S. Kershaw. 1959. Montpellier, France, 17-20 October 1999. Bulletin Host finding by aphids in the field II. Aphis fabae of IOBC/WPRS 22(2):52. Scop. (gynoparae) and Brevicoryne brassicae L.; with Olckers, T. 1998. Biology and host range of Platyphora a re-appraisal of the role of host-finding behavior in semiviridis, a leaf beetle evaluated as a potential virus spread. Annals of Applied Biology 47:424-444. biological control agent for Solanum mauritianum in Kibota, T. and S. P. Courtney. 1991. Jack of one trade, South Africa. Biocontrol 43: 225-239. master of one: host choice by Drosophila Olckers, T. 1999. Biological control of Solanum magnaquinaria. Oecologia 86:251-260. mauritianum Scopoli (Solanaceae) in South Africa: Leather, S.R. 1989. Life history traits of insect herbivores a review of candidate agents, progress and future in relation to host quality, pp. 175. In Bernays, E. A. prospects. In Olckers, T. and M. P. Hill (eds.) (ed.) Insect Plant Interactions, Vol. V. CRC Press, Boca Biological Control of Weeds in South Africa (1990- Raton, Florida, USA. 1999). African Entomology Memoir 1:65-73. Entomological Society of Southern Africa. Lowman, M.D. and J.D. Box. 1983. Variation in leaf toughness and phenolic content among five species Papaj, D.R. and M.D. Rausher. 1983. Individual variation of Australian rain forest trees. Australian Journal of in host location by phytophagous insects, 77-124. In Ecology 8:17-25. Ahmad S. (ed.) Herbivorous Insects: Host Seeking Behaviour and Mechanisms. Academic Press, New Ma, W-C. 1972. Dynamics of feeding responses in Pieris York. brassicae Linn as a function of chemosensory input: a behavior, ultrastructural and electrophysiological Peschken, D.P. and J.L. Derby. 1988. Host specificity of study. Mededelingen Landbouwhogeschool Wageningen Liriomyza sonchi Hendel (Diptera: Agromyzidae), a 72-11:1-162. potential biological control agent for the control of weedy sow-thistles, Sonchus spp., in Canada. The Canadian Entomologist 120:593-600.

66 Host Specificity Testing: Why We Do It and How We Can Do It Better Proceedings: Host Specificity Testing of Exotic Arthropod Biological Control Agents: The Biological Basis for Improvement in Safety

Price, P. W., N. Cobb, T.P. Craig, G.W. Fernandes, J.K. T. M., L. Barton Browne, and J. Stanley (eds.) Host Itami, S. Mopper, and R.W. Preszler. 1990. Insect Specificity Testing in Australasia: Towards Improved herbivore population dynamics on trees and shrubs: Assays for Biological Control. Department of Natural new approaches relevant to latent and eruptive species Resources, Coorparoo, Queensland, Australia. and life table development, pp. 1-38. In Bernays, E. Singer, M.C., D. Vasco, C. Parmesan, C.D. Thomas, and A. (ed.) Insect-Plant Interactions, Vol. II. CRC Press, D. Ng. 1992. Distinguishing between ‘preference’ Boca Raton, Florida, USA. and ‘motivation’ in food choice: an example from Prokopy, R.J. and W.J. Lewis. 1993. Application of insect oviposition. Animal Behavior 44:463-471. learning to pest management, pp. 308-342. In Papaj, Singer, M. C., C. D. Thomas, and C. Parmesan, C. 1993. D. R. and A. C. Lewis (ed.) Insect Learning: Ecological Rapid human-induced evolution of insect-host and Evolutionary Perspectives. Chapman and Hall, associations. Nature 366:681-683. New York. Slansky, F. and J.G. Rodriguez. 1987. Nutritional ecology Prokopy, R.J., D.R. Papaj, S.B. Opp, and T.T.Y. Wong. of insects, mites, spiders and related invertebrates: an 1987. Intra-tree foraging behaviour of Ceratitis overview, pp. 1-69. In Slansky, F. and J. G. Rodriguez capitata flies in relation to host fruit density and (eds.) Nutritional Ecology of Insects, Mites, Spiders and quality. Entomologia Experimentalis et Applicata Related Invertebrates. John Wiley and Sons, New York. 45:251-258. Solarz, S.L. and R.M. Newman. 1996. Oviposition Purcell, M.F., J.K. Balciunas, and P. Jones. 1997. Biology specificity and behavior of the watermilfoil specialist and host-range of Boreioglycaspis melaleucae Euhrychiopsis lecontei. Oecologia 106:337-344. (Hemiptera: Psyllidae), a potential biological control Stanley, J.N. and M.H. Julien. 1998. The need for post- agent for Melaleuca quinquenervia (Myrtaceae). release studies to improve risk assessment and decision Environmental Entomology 26:366-372. making in classical biological control, pp. 561-564. Rausher, M.D. 1980. Host abundance, juvenile survival, In Zalucki, M., Drew, R., and G. White (eds.) Sixth and oviposition preference in Battus philenor. Evolution Australasian Applied Entomological Research Conference 34:342-355. Addendum. The University of Queensland Printery, Renwick, J.A.A. and K. Lopez. 1999. Experience-based Brisbane, Australia. food consumption by larvae of Pieris rapae: addiction Stanley, J.N. and M.H. Julien. 1999. The host range of to glucosinolates? Entomologia Experimentalis et Eccritotarsus catarinensis (Heteroptera: Miridae), a Applicata 91:51-58. potential agent for the biological control of water Schoonhoven, L.M. 1987. What makes a caterpillar eat? hyacinth (Eichhornia crassipes). Biological Control The sensory code underlying feeding behavior, pp. 14:134-140. 69-97. In Chapman, R. F., E. A. Bernays, and J. G. Symons, F.B. and G.W. Beccaloni. 1999. Phylogenetic Stoffolano Jr (eds.) Perspectives in Chemoreception and indices for measuring the diet breadths of Behaviour. Springer-Verlag, New York. phytophagous insects. Oecologia 119:427-434. Schroeder, D. and R.D. Goeden, 1986. The search for Szentesi, A. and T. Jermy. 1990. The role of experience in arthropod natural enemies of introduced weeds for host choice by phytophagous insects, pp. 39-74. In biological control - in theory and practise. Biocontrol Bernays, E. A. (ed.) Insect-Plant Interactions, Vol. II. News and Information 7:147-154. CRC Press, Boca Raton, Florida, USA. Scriber, J.M. 1984. Host-plant suitability, pp. 159-202. Tabashnik, B. E. 1983. Host range evolution: the shift In Bell, W. J. and R. T. Carde (eds.) Chemical Ecology from native legume hosts to alfalfa by the butterfly, of Insects. Chapman and Hall, New York. Colias philodice eriphyle. Evolution 17:150-162. Shepherd, R.C.H. 1990. Problems which arise with host Thomas, M.B. and A.J. Willis. 1998. Biocontrol – risky specificity testing of insects, pp. 85-92. In Delfosse, but necessary? Trends in Ecology and Evolution th E. S. (ed.) Proceedings of the VII International 13:325-329. Symposium on the Biological Control of Weeds, 6-11 Thompson, J.N. 1988. Variation in preference and March 1988, Rome Italy. MAF, Rome. specificity in monophagous and oligophagous Sheppard, A.W. 1999. Which test? A mini review of test swallowtails. Evolution 42:118-128. usage in host specificity testing, pp. 60-69. In Withers,

Host Specificity Testing: Why We Do It and How We Can Do It Better 67 Proceedings: Host Specificity Testing of Exotic Arthropod Biological Control Agents: The Biological Basis for Improvement in Safety

Thompson, J.N. 1998. Rapid evolution as an ecological Wan, F. H., P. Harris, L.M. Cai, and M.X. Zhang. 1996. process. Trends in Ecology and Evolution 13:329-332. Host specificity of Altica carduorum Guer. van Dam, N.M. and D. Hare. 1998. Differences in (Chrysomelidae: Coleoptera), a defoliator of Cirsium distribution and performance of two sap-sucking arvense (L.) Scop. (Asteraceae) from north-western herbivores on glandular and non-glandular Datura China. Biocontrol Science and Technology 6:521-530. wrightii. Ecological Entomology 23:22-32. Wapshere, A.J. 1989. A testing sequence for reducing van Klinken, R.D. 1996. Understanding diversity in rejection of potential biological control agents for ecological communities: a study of Australian weeds. Annals of Applied Biology 114:515-526. Drosophilidae, with emphasis on the coracina group Waring, G.L. and N.S. Cobb. 1989. The impact of plant of Scaptodrosophila Species. PhD Thesis, University stress on herbivore populations dynamics, pp. 167- of Queensland, Brisbane, Australia, 271 pp. 226. In Bernays, E. A. (ed.) Insect-Plant Interactions, van Klinken, R.D. 1999a. Is the host-specificity of Vol. IV. CRC Press, Boca Raton, Florida, USA. biological control agents likely to evolve once released? Weidemann, G.J. 1991. Host-range testing: safety and Evaluating Indirect Ecological Effects of Biological science, pp. 83-96. In TeBeest, D. O. (ed.) Microbial Control, IOBC International Symposium, 17-20 Control of Weeds. Chapman and Hall, London. October 1999, Conference Proceedings. Bulletin of Wiklund, C. 1975. The evolutionary relationships between IOBC/WPRS 22(2):67. adult oviposition preference and larval host range. van Klinken, R.D. 1999b. Developmental host-range of Oecologia 18:185-197. Mozena obtusa (Heteroptera: Coreidae), a potential Wiklund, C. 1981. Generalist versus specialist oviposition biocontrol agent for mesquite (Prosopis species). behaviour in Papilio machaon (Lepidoptera) and Biological Control 16:283-290. functional aspects on the hierarchy of oviposition van Klinken, R.D. Host-specificity constrains evolutionary preferences. Oikos 36:163-170. host change in the psyllid Prosopidopsylla flava. Withers, T. M. 1998. Influence of plant species on host Ecological Entomology. (In press) acceptance behaviour of the biocontrol agent van Klinken, R.D. and T.A. Heard. Estimating the Zygogramma bicolorata (Col.: Chrysomelidae). fundamental host range of Evippe sp. #1 Biological Control 13:55-62. (Gelechiidae), a potential biocontrol agent for Withers, T. M., L. Barton Browne and J.N. Stanley. 2000. Prosopis species (Leguminosae). Biocontrol Science How time-dependent processes can affect the and Technology. (In press) outcomes of assays used in host specificity testing, Waage, J. and A. Kirk. 1999. Foreword. Evaluating Indirect pp. 27-41. In Van Driesche, R.G., T. A. Heard, A. S. Ecological Effects of Biological Control, Global IOBC McClay and R. Reardon (eds.) Proceedings: Host International Symposium, Montpellier, France, 17- Specificity Testing of Exotic Arthropod Biological Agents: 20 October 1999. Bulletin of IOBC/WPRS 22(2):1. The Biological Basis for Improvement in Safety. X International Symposium on Biological Control of Wan, F.H. and P. Harris. 1996. Host finding and Weeds, July 4-14, 1999, Bozeman, Montana. USDA. recognition by Altica carduorum Guer. Forest Service Bulletin, FHTET-99-1, Morgantown, (Chrysomelidae: Coleoptera), a defoliator of Cirsium West Virginia, USA. 95 pp. arvense, (L.) Scop. Entomologia Experimentalis et Applicata 80:491-496. Zwölfer, H. and Harris, P. 1971. Host specificity determination of insects for biological control of Wan, F. H. and P. Harris. 1997. Use of risk analysis for weeds. Annual Review of Entomology 16:159-178. screening weed biocontrol agents: Altica carduorum Guer. (Chrysomelidae: Coleoptera) from China as a biocontrol agent of Cirsium arvense (L.) Scop. in North America. Biocontrol Science and Technology 7:299-308.

68 Host Specificity Testing: Why We Do It and How We Can Do It Better