PLANT FINDING BEHAVIOR OF PHYTOPHAGOUS INSECTS AND BIOLOGICAL CONTROL OF AQUATIC PLANTS

Transcription

1 PLANT FINDING BEHAVIOR OF PHYTOPHAGOUS INSECTS AND BIOLOGICAL CONTROL OF AQUATIC PLANTS A dissertation submitted to Kent State University in partial fulfillment of the requirements for the degree of Doctor of Philosophy by Justin L. Reeves December 2010

2 Dissertation written by Justin L. Reeves B.A., Western State College of Colorado, 2006 Ph.D., Kent State University, 2010 Approved by Patrick D. Lorch Mark W. Kershner Ferenc A. de Szalay Marilyn A. Norconk, Chair, Doctoral Dissertation Committee, Members, Doctoral Dissertation Committee Accepted by James L. Blank John R. D. Stalvey, Chair, Department of Biological Sciences, Dean, College of Arts and Sciences ii

3 TABLE OF CONTENTS LIST OF FIGURES....vi LIST OF TABLES...viii ACKNOWLEDGMENTS...x CHAPTER I. Introduction.. 1 II. Biological control of invasive aquatic and wetland plants by arthropods: a meta-analysis of data from the last three decades Abstract...9 Introduction...10 Methods..13 Literature Search 13 Data Extraction..14 Analyses.18 Results...22 Discussion Acknowledgments.37 References..38 III. Biological control of Eurasian watermilfoil by Euhrychiopsis lecontei: assessing efficacy and timing of sampling.47 Abstract..47 Introduction...48 Methods..50 Results and Discussion.57 Acknowledgments..65 References iii

4 IV. Vision is important for plant location by the phytophagous aquatic specialist, Euhrychiopsis lecontei Dietz (Coleoptera: Curculionidae)..69 Abstract..69 Introduction Methods..73 Importance of light in plant location..77 Weevil attraction to plants in vials.78 Visual plant differentiation 79 Effect of water turbidity on plant location.80 Results 81 Importance of light in plant location..81 Weevil attraction to plants in vials.83 Visual plant differentiation 83 Effect of water turbidity on plant location.85 Discussion..85 Acknowledgments..89 References..89 V. Visual plant differentiation by the milfoil weevil, Euhrychiopsis lecontei Dietz (Coleoptera: Curcuionidae)..93 Introduction Methods..94 Results and Discussion..96 Acknowledgments..99 References 100 VI. Visual active space of the milfoil weevil, Euhrychiopsis lecontei Dietz (Coleoptera: Curcuionidae)..101 Abstract 101 Introduction..102 Methods 104 Visual active space Effect of turbidity on visual active space.108 Results..110 Visual active space Effect of turbidity on visual active space.113 Discussion 113 Acknowledgments References 120 iv

5 VII. Vision should not be overlooked as an important host-plant detection and selection mechanism for phytophagous insects Abstract Introduction..125 Assumption 1: Vision, in general, is not an important host location mechanism because insect visual acuity is poor..127 Assumption 2: Vision is only used when appropriate chemical cues are detected.132 Assumption 3: Insects cannot visually differentiate plant species.134 Assumption 4: Color is the only important visual stimulus for phytophagous insects..135 Underexplored areas of study for phytophagous insect visual ecology Conclusions..144 Acknowledgments References 145 VIII. Conclusions.163 APPENDIX I. A method for growing legumes with and without root nodules for studying nodule-attacking Rivellia (Diptera: Platystomatidae).171 Introduction..171 Methods Results and Discussion Acknowledgments 177 References 177 APPENDIX II. Reprint permissions for published chapters v

6 LIST OF FIGURES CHAPTER II. Biological control of invasive aquatic and wetland plants by arthropods: a meta-analysis of data from the last three decades. Fig Categories used to organize papers from literature search and their overall proportion of the entire set of papers...15 Fig Log response ratio effect size estimates for biological control agent analysis. 24 Fig Log response ratio effect size estimates for one vs. two agent use analysis...26 Fig Log response ratio effect size estimates for observational (no control group; initial value used as control) vs. experimental (true control group) study analysis Fig Log response ratio effect size estimates for lab vs. field study analysis.. 28 Fig Log response ratio effect size estimates for plant variable measured analysis.. 29 Fig Log response ratio effect size estimates for replicate type analysis...30 CHAPTER III. Biological control of Eurasian watermilfoil by Euhrychiopsis lecontei: assessing efficacy and timing of sampling. Fig Geographic distribution of the Michigan and Wisconsin lakes included in this study Fig Changes in proportional plant density [(final plant density initial plant density) / initial plant density] at lake-specific control sites vs. treatment sites Fig The relationship between final EWM densities (stems/m 2 ) and the number of days passed between initial and final surveys for control and treatment sites...61 vi

7 Fig Average proportional EWM density change by lake and year at control vs. treatment sites.. 63 CHAPTER IV. Vision is important for plant location by the phytophagous aquatic specialist, Euhrychiopsis lecontei Dietz (Coleoptera: Curculionidae). Fig Photograph of an adult Euhrychiopsis lecontei head Fig Grid used to track weevil movement in experimental arenas Fig Results of light importance experiment Fig Results of weevil attraction to plants in vials experiment CHAPTER VI. Visual active space of the milfoil weevil, Euhrychiopsis lecontei Dietz (Coleoptera: Curculoionidae). Fig Cross section and dimensions of trough used for these experiments Fig Results of pooled active space replicates Fig Results of turbidity experiment Fig Conceptual model of plant finding by E. lecontei. 118 vii

8 LIST OF TABLES CHAPTER II. Biological control of invasive aquatic and wetland plants by arthropods: a meta-analysis of data from the last three decades. Table 2.1. Results of meta-analyses performed in this study...23 Appendix 2.A. Biological control agents and their target plants used in the literature search for this study...41 Appendix 2.B. Papers used in the above analyses. 44 CHAPTER III. Biological control of Eurasian watermilfoil by Euhrychiopsis lecontei: assessing efficacy and timing of sampling. Table 3.1. Lakes used in this analysis, their locations (county, state), surface area, average and maximum depths, and mean/range of final weevil densities (# / stem) at treatment sites CHAPTER VI. Visual active space of the milfoil weevil, Euhrychiopsis lecontei Dietz (Coleoptera: Curculionidae). Table 6.1. Pooled visual active space replicates showing how many weevils swam toward the vial with a plant stem (out of 40), how many of these contacted the vial first, and whether they contacted the vial or trough wall first significantly more often..111 Table 6.2. Turbidity results showing how many weevils swam toward the vial with a plant stem (out of 25), how many of these contacted the vial first, and whether they contacted the vial or trough wall first significantly more often viii

9 CHAPTER VII. Vision should not be overlooked as an important host-plant detection and selection mechanism for phytophagous insects. Appendix 7.A. Bibliographic information for references on phytophagous insect visual ecology..158 ix

10 ACKNOWLEDGMENTS This dissertation is dedicated to my family: Pat, Andrea, and Candice Reeves. Their love and support have allowed me to successfully pursue my interest in biology since I was a child. I would never have gotten this far if I did not have such great parents. Special thanks are extended to my advisor, Pat Lorch, whose advice, guidance, and willingness to let me pursue my own interests allowed me to keep motivated until the end. I also thank Mark Kershner for, amongst many things, fostering me in his lab early in my graduate career when things got interesting. My other committee members, Ferenc de Szalay and Marilyn Norconk, are thanked for their guidance as well. Much gratitude is extended to Marty Hilovsky and EnviroScience, Inc., whose donations of data, supplies, and weevils were instrumental to successfully completing my dissertation work (and even getting me to Kent State in the first place). Next, I would of course like to thank my many graduate student colleagues and friends for their help and fun times along the way. There are too many to list here, and I do not want to miss anyone, so you are collectively thanked. Ben Foote and Joe Keiper are also thanked for their advice along the way and for all the bug lunches at Bob Evans. I am especially grateful to Ben Foote for always wanting to talk about the West and complain about Ohio. Finally, Julie Proell is thanked for all the love, support, help, and fun she has provided since the start of my graduate career. x

11 CHAPTER I INTRODUCTION The field of biological control of invasive plants has seen many successes since its inception over 100 years ago, with some of the most striking examples coming from control of aquatic plants (reviewed by Andres and Bennett 1975; McFadyen 1998). Because biological control may provide the most ecologically and economically sustainable approach to controlling invasive plants, much research is warranted and has been done in this area (McFadyen 1998). Though a good deal of research has been performed on biological control of aquatic plants, more basic research is still needed in many areas, especially on the biological control agents themselves (Cuda et al. 2008). The overarching theme to my dissertation, then, was to explore the overall success of aquatic plant biological control at many scales. The work presented within covers the field of aquatic plant biological control as a whole, to plant finding behaviors of the native milfoil weevil (Euhrychiopsis lecontei Dietz; Coleoptera: Curculionidae), a biological control agent for Eurasian watermilfoil (Myriophyllum spicatum L.). I have focused on plant finding behaviors of E. lecontei for closer study because plant finding mechanisms are one of the most understudied and perhaps most important aspects for understanding the efficacy of aquatic weed biological control (Cuda et al. 2008). 1

12 2 Though most of the successful aquatic weed biological control programs have involved classical biological control (whereby plant pests are collected in the invasive plant s native range and released into the plant s invasive range to ideally permanently control the plant; McFadyen 1998), one interesting biological control system uses the native milfoil weevil, E. lecontei, to control M. spicatum. Since its introduction into the U.S. in the 1940 s, M. spicatum has spread to at least 45 states and three Canadian provinces (Newman 2004), causing much ecological and economic damage (Grace and Wetzel 1978; Smith and Barko 1990; Boylen et al. 1999). Because E. lecontei is native to the U.S. and is released into lakes that may already have populations of the weevil, many of the safety concerns related to releasing novel, non-native insects are reduced. Euhrychiopsis lecontei is a specialist on plants in the genus Myriophyllum (Newman 2004), so these weevils do not harm other macrophyte species, even including many native Myriophyllum spp. (Sheldon and Creed 2003). This specificity is a plus because host-range is an extremely important concern when releasing insect biological control agents, as non-target damage can be a large problem (McFadyen 1998). Both laboratory and field testing have shown E. lecontei to be the most promising biological control agent for M. spicatum (Sheldon and Creed 1995; Newman 2004). However, much variability can still be seen in weevil efficacy both within and across lakes (Reeves et al. 2008). Because of this variability, and because plant finding mechanisms have been recently thought to be important in explaining efficacy of biological control agents for aquatic weeds (Cuda et al. 2008), I have explicitly examined plant finding by E. lecontei. More specifically, as a complement to Marko et al. (2005)

13 3 who studied chemosensory abilities for plant location, I studied the use of visual cues by E. lecontei for host-plant detection. Because E. lecontei overwinters on land as adults in leaf litter (Newman et al. 2001), the plant finding mechanisms of E. lecontei become especially interesting because the weevils have the challenge of finding plants in an entirely different habitat than that in which they overwintered. Since biological control ideally provides long term or permanent control of invasive plants, an understanding of how E. lecontei finds plants will need to be developed before we can understand when E. lecontei can find plants, which is directly related to predicting efficacy. That is, if adults cannot find plants after overwintering on land, they cannot damage or control the plants, so understanding plant finding mechanisms is clearly important. In order to contribute to our understanding of plant finding mechanisms used by aquatic biological control agents (Cuda et al. 2008), my dissertation includes work at many scales. Work is presented at scales ranging from the E. lecontei plant finding behavior studies noted above, to a meta-analysis of biological control efficacy of aquatic and wetland plants as a whole. At the largest scale, Chapter II is an overarching treatment of biological control of aquatic and wetland weeds by arthropods. Meta-analytic techniques were used to categorize 541 papers that have been published on biological control agents, extract and combine data when appropriate, and perform cumulative analyses across studies to find any significant effects of introducing insects into lakes or wetlands with invasive plant problems. I looked for differences in control efficacy by plant species, biological control agent species, and experimental design to provide an understanding of biological control

14 4 efficacy on a finer, more quantitative and objective scale than can be provided by the available narrative reviews (e.g., Andres and Bennett 1975; McFadyen 1998; Cuda et al. 2008). Though no heterogeneity was seen between agents, many significant effects were seen when considering experimental design. Thus, it was shown that biological control of aquatic and wetland plants has been successful, but study design may impact results. Chapter III (Reeves et al. 2008) is a more specific examination of biological control efficacy, in which I characterized the efficacy of E. lecontei at controlling M. spicatum. I analyzed data from 30 lakes over six years of introductions of E. lecontei by EnviroScience, Inc. (an environmental firm that rears and stocks E. lecontei for controlling M. spicatum infestations). Much variability was seen in weevil efficacy, with at least some evidence that the weevils are working. Because timing of data collection seemed to be important in determining final plant densities (i.e., plant senescence may be skewing results) recommendations are made for better sampling designs for collecting data in this system. Because plant finding mechanisms are understudied and may be important for understanding efficacy in aquatic biological control systems (Cuda et al. 2008), Chapters IV-VI focus on visual plant finding behavior of E. lecontei. I demonstrated in Chapter IV (Reeves et al. 2009b) that vision is important for plant location by E. lecontei, while in Chapter V (Reeves and Lorch 2009) I extended this notion to show that E. lecontei can visually differentiate between host-plants and at least one non-host-plant species (Ceratophyllum demersum L.) using vision alone. Finally, Chapter VI extends the work on E. lecontei visual abilities by elucidating the visual active space of these weevils (i.e.,

15 5 the distance at which weevils are visually attracted to plants), and how turbidity may affect the visual active space. In general, active spaces are understudied for phytophagous insects (Schoonhoven et al. 2005), so Chapter VI also serves to add to the small number of examples of visual active spaces that have been elucidated (see Table 6.2 in Schoonhoven et al. 2005). An understanding of visual plant finding cues/abilities of E. lecontei not only can lead to a better understanding of when weevils can be expected to find and damage target host-plants, but may also lead to the development of traps to assess weevil presence/absence and abundance. To aid in these goals, Chapter VI provides a conceptual model for plant finding by E. lecontei that incorporates my research with other published studies on weevil behavior and life history. This model may also be applied to aquatic phytophagous insects in general. I found through my literature searches for Chapters IV-VI that relatively few people have studied the use of vision for host-plant location by phytophagous insects, so Chapter VII is a forum/discussion type article addressing this lack of research. Vision has historically been ignored because of the assumption that all plants look green while having different chemistries, thereby making chemicals the obvious (and only) important cue (Schoonhoven et al. 2005). Thus, chemicals have always dominated research as the major host-locating cues for phytophagous insects (Prokopy and Owens 1983). My weevil work, along with some other striking examples in the recent literature about the use of vision for plant location (i.e., since the review by Prokopy and Owens 1983) are discussed to frame my argument that more research should be performed in this area, not just for biological control agents. Many assumptions leading to discounting the

16 6 importance of vision are debunked as well. This chapter extends the reach of my dissertation beyond aquatic biological control systems to insect-plant systems as a whole. Finally, Appendix I (Reeves et al. 2009a) also extends the reach of my dissertation beyond aquatic biological control systems by providing a method for growing legumes both with and without root nodules for studying insects such as Rivellia (Diptera: Platystomatidae) that have subterranean, legume root nodule-feeing larvae. The methods in Appendix I will allow for testing many hypotheses about host finding and selection in legume nodule associated insects. For instance, since subterranean larvae such as Rivellia have limited mobility relative to terrestrial adults for finding suitable host-plants, adult Rivellia have a potentially large selective pressure to oviposit on appropriate plants. It can then be hypothesized that adults can somehow tell from above ground the below ground condition of the plant (i.e., healthy, active nodules) to select an appropriate host. Overall, Appendix I and Chapter VII extend the reach of my dissertation from aquatic plant biological control to host-plant finding and selection in general, making this dissertation what I hope to be a valuable addition to the literature. References Andres, L. A. and F. D. Bennett Biological control of aquatic weeds. Annual Review of Entomology 20:31-46.

17 7 Boylen, C. W., L. W. Eichler and J. D. Madsen Loss of native aquatic plant species in a community dominated by Eurasian watermilfoil. Hydrobiologia 415: Cuda, J. P., R. Charudattan, M. J. Grodowitz, R. M. Newman, J. F. Shearer, M. L. Tamayo and B. Villegas Recent advances in biological control of submersed aquatic weeds. Journal of Aquatic Plant Management 46: Grace, J. B. and R. G. Wetzel The production biology of Eurasian watermilfoil (Myriophyllum spicatum L.): a review. Journal of Aquatic Plant Management 16:1-11. Marko, M. D., R. M. Newman and F. K. Gleason Chemically mediated host-plant selection by the milfoil weevil: a freshwater insect-plant interaction. Journal of Chemical Ecology 31: McFadyen, R.E.C Biological control of weeds. Annual Review of Entomology 43: Newman, R. M Biological control of Eurasian watermilfoil by aquatic insects: basic insights from an applied problem. Archiv fur Hydrobiologie 159: Newman, R. M., D. W. Ragsdale, A. Milles and C. Oien Overwinter habitat and the relationship of overwinter to in-lake densities of the milfoil weevil, Euhrychiopsis lecontei, a Eurasian watermilfoil biological control agent. Journal of Aquatic Plant Management 39: Prokopy, R. J. and E. D. Owens Visual detection of plants by herbivorous insects. Annual Review of Entomology 28:

18 8 Reeves, J. L., P. D. Lorch, M. W. Kershner and M. A. Hilovsky Biological control of Eurasian watermilfoil by Euhrychiopsis lecontei: assessing efficacy and timing of sampling. Journal of Aquatic Plant Management 46: Reeves, J. L., B. A. Foote and P. D. Lorch. 2009a. A method for growing legumes with and without root nodules for studying nodule-attacking Rivellia (Diptera: Platystomatidae). Proceedings of the Entomological Society of Washington 111: Reeves, J. L. and P. D. Lorch Visual plant differentiation by the milfoil weevil, Euhrychiopsis lecontei Dietz (Coleoptera: Curculionidae). Journal of Insect Behavior 22: Reeves, J. L., P. D. Lorch and M. W. Kershner. 2009b. Vision is important for plant location by the phytophagous aquatic specialist Euhrychiopsis lecontei Dietz (Coleoptera: Curculionidae). Journal of Insect Behavior 22: Schoonhoven, L. M., J. J. A. van Loon and M. Dicke Insect-Plant Biology, Oxford University Press, New York. Sheldon, S. P. and R. P. Creed, Jr Use of a native insect as a biological control for an introduced weed. Ecological Applications 5: Sheldon, S. P. and R. P. Creed, Jr The effect of a native biological control agent for Eurasian watermilfoil on six North American watermilfoils. Aquatic Botany 76: Smith, G. S. and J. W. Barko Ecology of Eurasian watermilfoil. Journal of Aquatic Plant Management 28:55-64.

19 CHAPTER II BIOLOGICAL CONTROL OF INVASIVE AQUATIC AND WETLAND PLANTS BY ARTHROPODS: A META-ANALYSIS OF DATA FROM THE LAST THREE DECADES Abstract Results from the last three decades covering many arthropod biological agents and their target invasive aquatic and wetland plants were combined using meta-analytical techniques to provide a more objective, quantitative understanding of biological control efficacy than can be provided by available narrative reviews. More specifically, analyses were performed to determine if there are differences in how well diverse biological agents perform, and if different experimental designs (i.e., observational vs. experimental studies, lab vs. field studies, variable measured, replicate type used, and use of one vs. two agents) impact the results of biological control studies. Across all analyses, biological control of aquatic and wetland plants was generally successful (or at least the agents are significantly damaging their target plants). Though little heterogeneity in efficacy was seen between the biological control agents used, all experimental design analyses showed significant heterogeneity between the study types in each respective 9

20 10 analysis, implying that study design can significantly impact the results of biological control studies. Observational studies reported significantly higher effects than experimental studies. We tested the hypothesis that this may be due to the lack of controls in observational studies. Based on these and other results, we suggest that field studies (with control groups) be performed using subsamples of an area (quadrats, transects, etc.), with biomass or density being the plant variable measured. On a more qualitative basis, we suggest that more long-term (2+ years) studies and more non-target effect studies be performed in the future. Introduction Classical biological control can provide the most ecologically and economically sound methodology for controlling invasive plants and their negative impacts (McFadyen 1998). Recent evaluations of biological control of plants (both terrestrial and aquatic) have identified a range of factors correlated with effectiveness, but these studies have either been qualitative (e.g., Andres and Bennett 1975; McFadyen 1998; others listed in Cuda et al. 2008) or quantitative but did not compare data within or between individual invasive plant and control agent species (Stiling and Cornelissen 2005). Particularly for biological control of invasive submersed aquatic plants, many general factors have been shown to influence biological control of a wide range of plant species. Abiotic factors such as climate, habitat conditions, and concurrent management tactics such as mechanical harvesting or pesticide application have been shown to influence biological

21 11 control efficacy in multiple systems. Similarly, biotic factors such as host-plant quality or genotype, biological control agent density, and agent mortality factors such as predation, parasitism, or disease can also influence control efficacy. A qualitative review of these factors, along with an overview of the field of biological control of submersed aquatic plants, is provided by Cuda et al. (2008). Successful biological control of at least one floating plant, water hyacinth (Eichhornia crasspies), may be dependent on control agent (larval) density as well, while larval density is in turn impacted by plant status (Wilson et al. 2006). Thus, many of the factors that influence control of submersed plants are likely applicable to aquatic and wetland plants in general. Though the many factors above have been found to potentially impact biological control efficacy, much work still needs to be done, particularly for aquatic plants (Cuda et al. 2008). Quantitative reviews of the literature may help to more fully understand control efficacy and go beyond the general claims of successful biological control of aquatic plants that have come from narrative reviews such as McFadyen (1998). This may be especially true for understanding how differences in control efficacy depend on which agent is used, or on experimental design, both of which are factors on which little or no work, especially quantitative work, has been done on aquatic plants. Meta-analysis provides a method for quantitatively synthesizing the results of independent experiments to draw general conclusions (see Cooper and Hedges 1994; Gurevitch and Hedges 2001). Because meta-analysis can be a useful tool for understanding patterns across multiple studies, it has been used in many ecological fields, including biological control (Stiling and Cornelissen 2005). Though the meta-analysis

22 12 performed by Stiling and Cornelissen (2005) included a categorical analysis of weed biological control that showed general control success, it was not specific to aquatic or wetland invasive plants (the plant species included were not listed), nor were invasive plants the main focus of the study (the study covered the entire field of biological control of all types of organisms). There were also no analyses performed to specifically compare individual biological control agent species, which can be important in deciding which agents to use for which plants, or at least in examining which agents have worked and which have not. The purpose of the work presented here was to broadly quantify the efficacy of biological control agents specifically of invasive aquatic and wetland plants using metaanalytical techniques. This effort was designed to improve on previous narrative reviews of aquatic plant biological control (e.g., Andres and Bennett 1975; Cuda et al. 2008). Both aquatic and wetland plants were included here because it can be difficult to differentiate between the two plant types (Cuda et al. 2008). Heterogeneity between arthropod agents was tested to determine if there are any differences in how well the various biological control agents have performed when compared to one another (which may help decide which agents to use for which plants), or at least which agents have had a significant effect on their target plant. Further, it was hypothesized that experimental design may influence biological control efficacy, so we quantified differences in effect across agents based on experimental design factors such as use of 1 vs. 2 biological control agents, experimental (with controls) vs. observational (without controls) studies, lab vs. field studies, the response variable measured (e.g., plant density, biomass, percent

23 13 cover, weight change, leaf area grazed) the type of replicate used (whole lake, subsample of an area, aquaria, or individual plants), and finally the study duration. Data from the last three decades of biological control work published in a wide variety of journals were used to empirically test these hypotheses and provide a more objective exploration of aquatic plant biological control efficacy than can be provided by narrative reviews. Methods Standardized meta-analysis techniques were used for this study following Gurevitch and Hedges (2001). Literature search methods, data inclusion methods, and measures to avoid non-independence of data are often underreported in meta-analysis papers, so included below are specific sections to detail these aspects of this study. Literature Search An extensive list of specialist arthropod biological control agents (both candidate agents and currently employed agents) and their corresponding target invasive aquatic and wetland plants was built through intensive internet and literature searches and cross referenced with Julien and Griffiths (1998) to make sure no agents (or at least agents with published articles containing biological control data) were missed. Once the list of agents and plants was compiled (Appendix 2.A), the literature search for studies to be used for the meta-analyses was performed in two well known, comprehensive databases: ISI Web of Science and Biological Abstracts. Within each of these databases, the scientific names

24 14 of all of the control agents in Appendix 2.A were used as search terms. References for each article containing the names of the biological control agents (n = 541) were used and categorized by reading the title and abstract (and the actual paper when necessary) into the categories listed in Fig. 2.1 to provide a broad overview of the type of research that had been performed using each of the biological control agents. The categories were not mutually exclusive, and studies were placed into each appropriate category. Overall, the literature search covered studies performed between 1980 and September Data Extraction Of the 541 articles found through the literature search (that included all articles on each agent and their life history even if no plant data were reported), 126 articles reported some sort of agent effects on plants (any plant effect measured was included at this stage; some papers in Fig. 2.1 were included in both lab and field categories when necessary). Of these 126 articles, 53 articles provided 62 data points appropriate for meta-analysis (Appendix 2.B) using the data inclusion and analysis methods described below. The large discrepancy between the total number of papers and papers with appropriate data for analysis was a result of searching by agent names rather than control-type terms like biological control. This was done to make sure papers were not excluded due to the choice of search terms (Cooper and Hedges 1994), giving as broad an overview of the literature as possible (as in Stiling and Cornelissen 2005). For the articles with appropriate data, the models used here tested for differences in control efficacy based on

25 15 Fig Categories used to organize papers from literature search and their overall proportion of the entire set of papers (n = 541). Categories were not mutually exclusive, and studies were placed in all appropriate categories. The numbers above each bar represent the number of studies associated with that respective category Proportion of Studies dist. and life history (incl. host range) short-term effects: field behavior (incl. oviposition) fecundity and development biotic/abiotic influences population establishment Study Catetory short term effects: lab long term effects (2+ yrs) non-target effects

26 16 the biological control agent used (Appendix 2.A). Similarly, to examine the impact of experimental design on biological control efficacy across agents, the models used here quantified and compared the effects of whether one or two biological control agents were used concurrently in the study, the type of control used [initial value (observational studies) or real control (experimental studies)], whether the study was performed in the lab or field, the response variable measured (biomass, density, percent cover, or other related measures), the replicate type used (individual plant, aquarium, sub-samples of an area, or entire lake/pond), and finally study duration. Heterogeneity between agents or study designs was examined to determine if any significant differences in control efficacy by any of these factors exists. Similarly, heterogeneity within agents and study designs was examined. For all appropriate papers, sample sizes, means and standard deviations (sometimes calculated from other variance measures) were collected for the effect size calculations (see below) for both treatment and control groups. Some of the studies in the dataset included only observational data in which only initial and final plant densities, biomasses, etc. were reported. In these cases, the initial value was used as the control value for the analyses. When necessary, graphs were digitized and data were extracted using the program Data Thief III (version 1.5). Because non-independence of data is a frequent problem in meta-analysis (Gurevitch and Hedges 1999), six steps were taken to help avoid this potential confound. First, because long-term control is one of the primary goals of biological control (McFadyen 1998), when data were presented for multiple time steps in a given study, the

27 17 longest time span presented was used in the analysis (a method also used by Shurin et al. 2002; McCarthy et al. 2006). Second, when multiple densities of control agents were used, the highest treatment density was selected for the analyses (as in McCarthy et al. 2006). Third, when more than one variable was measured to estimate control efficacy, only one variable was selected for inclusion in the analyses based on what was thought best represented plant control (e.g., biomass, density, percent cover reduction). Fourth, for studies that reported many different sites (lakes or wetlands), means and standard deviations were calculated across all sites. Fifth, when more than one biological control agent was independently used in a given study (see Appendix 2.B) all agents were included in the analyses if the two (or more) agents used were treated as independent experiments. Finally, when more than one agent was used at once in one location and in a single experiment, all agents were counted together as one data point (one agent with two names) for the analyses. Since standard deviation data are required for modern, weighted meta-analyses (Rosenberg et al. 2000), studies that did not report standard deviation (or equivalent) data (n = 14) were excluded from the analyses. Similarly, seven studies were removed because they were simply short narrative reviews of projects without providing usable (or any) data. Incomplete data reporting is a common problem for meta-analyses (Gurevitch and Hedges 1999). Next, 33 studies were excluded for various other problems such as reporting data that were not directly related to plant control (such as leaf hardness, nutrient levels, etc.) or reporting unusable non-mean/standard deviation type data. Finally, it should be noted that studies that reported percent or number of leaves (or other

28 18 plant parts) damaged (n = 12) were not included in the analyses, as it was felt feeding scars may only be indicative of agent population establishment and not actual control. Because one preferred definition of biological control involves decreasing vigor, density, or reproductive capacity of weeds (DeLoach 1997 as cited in Cuda et al. 2008), studies reporting percent of leaves with scars or damage were not included since there is no way of knowing if those scars are indicative of the plant actually being controlled. For this reason, the studies that were included in the analyses were those that reported biomass decreases, density decreases, percent cover decreases, or similar values (such as percent total defoliation), which all are clearly related to control. Finally, the statistical software used (see below) rejected some studies (n = 7) that were initially included for computational conflicts such as having a mean control or treatment value of zero, or a sample size of one (Rosenberg et al. 2000). Overall, as suggested by Englund et al. (1999), no inclusion/exclusion decisions were made based on the quality of data (by study or journal). Data were included or excluded based on the objective criteria noted above, and all data that could be included were used. Analyses All analyses were performed using MetaWin 2.1 (Rosenberg et al. 2000), as in Stiling and Cornelissen (2005). The effect size metric used was the log response ratio: ln(r) = ln(x T / X C ); where X T is the mean for the treatment group and X C is the mean for the control group. This metric results in meaningful values (i.e., value is natural log of

29 19 proportion difference between treatment and control groups) with known statistical properties indicating how the treatment group responded relative to the control group within a study (Hedges et al. 1999). Here, it indicates the effect of the addition of biological control agents. Negative values of the log response ratio indicate that, for a given response variable, the treatment group decreased relative to controls, as would be expected if biological control agents reduced plant densities, biomass, etc. When appropriate, sign reversal markers were used to make the differences between control and treatment effect sizes consistent and meaningful across studies [e.g, so values such leaf area grazed (positive sign when agents are having an impact) are reversed in sign to be comparable to metrics such and biomass change (negative sign when agents are having an impact)]. To make the effect size results more meaningful and easier to interpret, percent reductions were back calculated from the effect sizes [ln(r)] and are presented in the table and figures below. In estimating the cumulative effect size (E ++ ) across studies (or across groups of studies), a weighted average was computed taking into account the individual variances associated with each study that were themselves computed using the sample sizes and standard deviations of the control and treatment groups for each study. The formula used for weighting was: V lnr = [(S T ) 2 / N T (X T ) 2 ] + [(S C ) 2 / N C (X C ) 2 ]; where V lnr is the studyspecific variance, S i is the standard deviation, N i is the sample size, X i is the average (where i = T for treatment or C for control; Rosenberg et al., 2000). As per traditional meta-analytical technique (Gurevitch and Hedges, 2001), all effect sizes were considered significantly different than zero when the 95% confidence intervals [based on 2500 bias-

30 20 corrected bootstrap iterations here because of the small sample sizes for some groupings in our models (as suggested by Adams et al., 1997)] did not overlap with zero. The log response ratio may be more appropriate than another popular effect size measure, d, in some cases because pooled standard deviation (between treatments and controls) is calculated as a part of d which may improperly affect the effect size, especially in cases of spatial heterogeneity between studies (a common condition in the data; see Osenberg et al. 1997). Also, the log response ratio has been used in other recent ecological meta-analyses (e.g., Stiling and Cornelissen 2005; Shurin et al. 2002). Finally, this effect size metric seemed appropriate since some of the data for control groups were initial values instead of actual control groups, making comparing initial and final values as a ratio seem more appropriate than comparing the values as separate experimental groups. See Hedges et al. (1999) for a comprehensive overview of the log response ratio. As suggested by Gurevitch and Hedges (1999; 2001), mixed models were used to analyze the data, as in the recent biological control meta-analysis performed by Stiling and Cornelissen (2005). The mixed models used here also tested for heterogeneity amongst the data used for each model. Heterogeneity both within (Q W ) and between (Q B ) categories (i.e., agents, experimental designs, etc.) in the mixed models were calculated and tested against the Chi-Square distribution. Total heterogeneity (Q T ) was also calculated for each grouping of studies (Rosenberg et al. 2000). A significant Q W value indicates there is significant heterogeneity in effect size between papers within a group (i.e., within an agent species, experimental design, etc.). Likewise, a significant Q B value indicates significant heterogeneity in effect size between groups of papers (i.e., between

31 21 agent species, experimental designs, etc.), as in an F-test. For all calculations, two or more data points for each model grouping (i.e., plant, agent, etc.) were required for analyses to be performed, so sample sizes were automatically adjusted for each of the analyses accordingly (hence the sample size variation among analyses in Table 2.1). It should be noted when interpreting the results below that the heterogeneity (Q) statistics were calculated separately from the bias-corrected bootstrap iterations, and therefore significance in the Q B statistics was not based directly upon on the confidence intervals. Next, a weighted continuous model meta-analysis (Rosenberg et al. 2000) was used to test for an effect of study duration on effect size to determine if longer studies resulted in larger effect sizes (since the biological control agents would have had more time to damage their respective plants), which clearly could confound the results. Because it is often beneficial to understand how the file drawer problem, or publication bias (e.g., experiments performed but not published due to negative results) may affect the interpretation of results of meta-analyses, fail safe analyses were performed. A fail safe number can be interpreted as the number of studies with negative results (effect size of zero or opposite to expectations) that would need to be included in the meta-analysis to negate an effect size that is significantly different than zero. Rosenthal s fail-safe method was used (as in Stiling and Cornelissen 2005), whereby a significant effect size is considered robust if the fail-safe number is larger than 5k + 10, with k being the number of studies in the meta-analysis (Rosenthal 1979).

32 22 Results All analyses of agents and experimental designs showed significant cumulative effect sizes (E ++ ; Table 2.1), and all analyses were considered robust to the file drawer effect based on fail-safe numbers, indicating that publication bias (Rosenthal 1979) is likely not a problem with the dataset presented here. Heterogeneity within groups (i.e., within agents and experimental designs; Q W ) was significant in all cases. Heterogeneity between groups (Q B ) was significant in all of the experimental design analyses, but nonsignificant in the agent species or one vs. two agent analyses (Table 2.1). Because of the many significant Q W and Q B results, total heterogeneity (Q T ) was significant in all analyses (Table 2.1). Within the biological control agent species analysis, there were two individual agent species that showed non-significant effect sizes: E. lecontei and B. melaleucae (Fig. 2.2). All other groups in all other analyses (i.e., all other agents and experimental design factors) showed significant effect sizes (Figs ). Pooling all agents for each target plant for analysis by plant species showed nonsignificant between-plant heterogeneity (Q B = 6.21; P = ) and did not change the interpretation of the study results. Based on the data here, no evidence exists of differences in efficacy by agent or plant species, indicating that no agent was significantly more effective at controlling its target plant than other agents and that no plant was significantly easier to control than others. The weighted continuous model to test the potential impacts of study duration on effect size was not significant (slope = ; P = 0.301; df = 58). Three data points

33 23 Table 2.1. Results of meta-analyses performed in this study (E ++ = cumulative effect size across groups). The 95% confidence intervals are the results from 2500 bias corrected bootstrap iterations. For the heterogeneity (Q) statistics (Q w = within group heterogeneity; Q B = between group heterogeneity; Q T = total heterogeneity), the numbers in parentheses represent the P-values associated with those analyses. Fail-safe numbers (see Rosenthal 1979) indicate the number of papers of no effect that would need to be included in each analysis to nullify the significant result, and all showed that the respective analysis can be considered robust. Analysis Groups (n) Papers (n) Q W Q B Q T E ++ 95% C.I. Fail-safe Agent (0.0015) 9.65 (0.4717) (0.0037) to vs. 2 Agents (0.0000) 0.41 (0.5225) (0.0000) to Exp. vs. Obs. Studies (0.0000) 8.48 (0.0036) (0.0000) to Lab vs. Field (0.0000) 7.70 (0.0055) (0.0000) to Variable Measured (0.0000) (0.0000) (0.0000) to Replicate Type (0.0000) (0.0005) (0.0000) to

34 24 Fig Log response ratio effect size estimates for biological control agent analysis. The data points represent the effect size for each agent and the bars surrounding the data points represent the bootstrapped 95% confidence intervals. The number in parentheses next to the agent name indicates the sample size of papers for that agent. The percentage next to sample size indicates the percent reduction across plant variables for that agent [back-calculated from ln(r)] to make effect size results easier to interpret. The agents that were excluded from this analysis for having a sample size of n = 1 were: G. pusilla, C. salviniae & S. multiplicalis, H. transverovittatus, P. acuminata, S. rufinasus, E. lecontei & A. ephemerella, I. variegatus, E. catarinensis, N. eichhorniae & E. catarinensis, G. nymphaeae, G. birmanica, O. terebrantis, M. lythri, G. calmariensis & M. lythri, M. scutellaris, C. myriophylli, L. ludoviciana, N. eichhorniae & O. terebrantis, B. affinis, N. eichhorniae & N. bruchi (with pathogen). The full names and target plants for these agents can be seen in Appendix 2.A.

35 25 Fig B. melaleucae (n = 3; -28.9%) H. pakistanae (n = 2; -37.0%) N. bruchi (n = 2; -38.1%) G. pusilla & G. calmariensis (n = 2; -38.9%) G. calamariensis (n = 5; -41.0%) E. lecontei (n = 8; -45.1%) N. eichhorniae (n = 7; -52.0%) Cumulative Effect Size (n = 43; -52.6%) B.melaleucae & O. vitiosa (n = 4; -58.8%) N. eichhorniae & N. bruchi (n = 5; -70.6%) A. ephemerella (n = 3; -73.0%) C. salviniae (n = 2; -78.8%) Effect Size [ln(r)]

36 26 Fig Log response ratio effect size estimates for one vs. two agent use analysis. The data points represent the effect size for each agent and the bars surrounding the data points represent the bootstrapped 95% confidence intervals. The number in parentheses next to the agent name indicates the sample size of papers for that group of studies. The percentage next to sample size indicates the percent reduction across variables for that group of studies [back-calculated from ln(r)] to make effect size results easier to interpret. 1 Agent (n = 46; -52.8%) Cumulative Effect Size (n = 62; -54.0%) 2 Agents (n = 16; -57.4%) Effect Size [ln(r)]

37 27 Fig Log response ratio effect size estimates for observational (no control group; initial value used as control) vs. experimental (true control group) study analysis. The data points represent the effect size for each agent and the bars surrounding the data points represent the bootstrapped 95% confidence intervals. The number in parentheses next to the agent name indicates the sample size of papers for that group of studies. The percentage next to sample size indicates the percent reduction across variables for that group of studies [back-calculated from ln(r)] to make effect size results easier to interpret. Experimental (n = 39; -46.6%) Cumulative Effect Size (n = 62; -54.0%) Observational (n = 23; -65.0%) Effect Size [ln(r)]

38 28 Fig Log response ratio effect size estimates for lab vs. field study analysis. The data points represent the effect size for each agent and the bars surrounding the data points represent the bootstrapped 95% confidence intervals. The number in parentheses next to the agent name indicates the sample size of papers for that group of studies. The percentage next to sample size indicates the percent reduction across variables for that group of studies [back-calculated from ln(r)] to make effect size results easier to interpret. Lab Studies (n = 28; -43.0%) Cumulative Effect Size (n = 62; -53.7%) Field Studies (n = 34; -69.8%) Effect Size [ln(r)]

39 29 Fig Log response ratio effect size estimates for plant variable measured analysis. The data points represent the effect size for each agent and the bars surrounding the data points represent the bootstrapped 95% confidence intervals. The number in parentheses next to the agent name indicates the sample size of papers for that group of studies. The percentage next to sample size indicates the percent reduction across variables for that group of studies [back-calculated from ln(r)] to make effect size results easier to interpret. The variables measured that were excluded from this analysis for having a sample size of n = 1 were: stand area, percent milfoil, percent infestation left, percent density change, percent total biomass, percent floating, total leaf area, number leaves removed, number of green leaves, percent healthy tissue, plant volume, tuber number. Density (n = 12; -33.0%) Biomass (n = 30; -46.7%) Cumulative Effect Size (n = 51; -50.4%) Weight Change (n = 2; -52.6%) Percent Cover (n = 5; -75.8%) Leaf Area Grazed (n = 2; -82.2%) Effect Size [ln(r)]

40 30 Fig Log response ratio effect size estimates for replicate type analysis. The data points represent the effect size for each agent and the bars surrounding the data points represent the bootstrapped 95% confidence intervals. The number in parentheses indicates the sample size of papers for that group of studies. The number in parentheses next to the agent name indicates the sample size of papers for that group of studies. The percentage next to sample size indicates the percent reduction across variables for that group of studies [back-calculated from ln(r)] to make effect size results easier to interpret. Individual Plant (n = 14; -40.0%) Aquaria (n = 13; -47.3%) Cumulative Effect Size (n = 62; -53.3%) Subsample (n = 28; -54.5%) Lake (n = 7; -77.6%) Effect Size [ln(r)]

41 31 were excluded from this analysis because the respective studies reported multiple or variable study durations between sites or sampling periods. Time was not averaged across sites for these studies the same way that the experimental data were averaged for the above analyses, as the imprecise time value may have confounded the continuous model results. Based on the non-significant continuous model results, study duration effects did not appear to confound the analyses. Discussion Based on the data presented here, in general, across agents and experimental designs, biological control of aquatic and wetland plants appears to be successful (or at least agents are significantly damaging their target plants). In every analysis performed, the cumulative effect size (E ++ ) was significantly less than zero and robust to file drawer effects (Table 2.1), implying that the biological control agents included in this study are negatively impacting their target aquatic and wetland plants. However, within these generally successful results, no differences were seen in efficacy by biological control agent species (or plant species; results not shown), with a resulting 52.6% average plant decrease across variables and agents (Cumulative Effect Size, Fig. 2.2). It should be noted that for the agents that individually showed non-significant effect sizes (E. lecontei and B. melaleucae; Fig. 2.2), the non-significant results were at least partially due to large confidence intervals from variable data, not necessarily a small effect size. These data points should not be interpreted as failures, and these results should not preclude future use of the agents to control their target plants. Euhrychiopsis lecontei is an example of a

42 32 biological control agent that shows variability in field effects (Reeves et al. 2008; data from this paper were used in the above analyses and contributed to the wide confidence intervals for E. lecontei overlapping zero), and thus is an agent where variance likely drove the non-significant results. Euhrychiopsis lecontei has been shown to be effective in many cases (reviewed by Sheldon and Creed 1995; Newman 2004), however, so E. lecontei (and B. melaleucae) should not be considered ineffective biological control agents based solely on the results presented here. Just as with the agent species results, non-significant heterogeneity was seen between studies that used one (52.8% plant reduction across variables) vs. two agents (54.0% plant reduction across variables; Fig. 2.3) concurrently. These data add to the debate in Denoth et al. (2002) about single vs. multiple agent releases for biological control of plants by showing that multiple agents may not work better than individual agents (at least for controlling the aquatic and wetland plants used here). Though no heterogeneity was seen in the agent analyses, the heterogeneity results from the experimental design factor models show that choice of experimental design may influence biological control studies. These heterogeneity results may explain some of the significant within agent heterogeneity (Q W ; Table 2.1), though future analyses will be needed to more fully understand these significant within-group (within agent, experimental design, etc.) heterogeneities. The heterogeneity results between experimental design factors are interesting, however, and do allow several suggestions to be made for future research.

43 33 First, examining the significant heterogeneity between experimental (with control group) vs. observational studies (without control group; initial values used as control data for analyses; Table 2.1), it is clear that observational studies reported larger effects on plants than experimental studies (observational studies mean effect size = 65.0% plant reduction across variables; experimental studies = 46.6% plant reduction across variables; Fig. 2.4). To attempt to better understand if this heterogeneity was a result of the agents performing better in observational studies, or if the study design inherently created a larger effect size, an analysis was performed taking initial values from experimental papers when available (n = 10) and using these values in place of control values for a new analysis. Non-significant heterogeneity was seen when comparing initial to final values in the same way as was done for observational studies [ln(r) = ; 39.2% plant reduction)] vs. comparing control to experimental values in those same 10 studies [ln(r) = ; 62.5% plant reduction)] (Q B = ; P = ). The lack of power due to small sample size make these results difficult to interpret, particularly given the large increase in effect size seen when initial values are substituted for control values. Because any number of factors other than biological control agents may impact plant populations (such as plant senescence; Reeves et al. 2008), to the extent that these factors affect control and treatment plots equally, focusing on the difference between treatment and controls will remove at least some of these non-agent effects. For future studies, including initial values in experimental studies (ideally for both treatment and control plots) will make it possible to better understand the difference in effect between observational and experimental studies.

44 34 Second, field studies (69.8% reduction across plant variables) had a larger overall effect than laboratory studies (43.0% reduction in plant variables; Fig. 2.5). The field studies used here may have been affected by the number of studies within this category that used observational data (n = 22 out of 34 total field studies) because observational data showed larger effect sizes than experimental data, as discussed above. However, when considering field data alone, there was non-significant heterogeneity between observational (67.7% reduction in plant variables, n = 22) and experimental (55.1% reduction in plant variables, n = 12) studies (Q B = 2.18; P = ). No matter how these results are interpreted, because biological control applications are clearly all in the field, field studies may be expected to provide a more realistic test of control efficacy. It may also be reassuring to lake managers that field studies provided larger effect sizes than laboratory studies. The potential confound of the field data by the observational data is an interesting problem and should be considered for future research and for the other analyses discussed below. Third, significant heterogeneity was demonstrated between the experimental variable measured in the studies (Table 2.1). Based on this analysis, we suggest that plant biomass (46.7% plant reduction across variables) or density (33.0% plant reduction across variables; Fig. 2.6) be used as the metric to determine control efficacy. Though these two metrics provided the smallest effect sizes of all the variables included in Fig. 2.6, the other variables are potentially confounded with other factors that inflated their effect size estimates. First, all the percent cover papers included here were observational, which again may have confounded the data to increase the overall effect size of that

45 35 variable. Next, it should be noted that the leaf area grazed and weight change data points both came from single papers because those papers reported two agents that were used independently and were thus included as separate data points (Fig. 2.6; Dech and Nosko 2002 and Creed and Sheldon 1994 respectively; Appendix 2.B). More data will be needed for these two metrics to better interpret the mean effect size or 95% confidence intervals. Fourth, significant heterogeneity was seen again in the replicate type used analysis. Subsamples of an area (i.e., transects, quadrats, plots, etc.) showed a 54.5% plant reduction across variables as compared to using individual plants (40.0% plant reduction across variables) or aquaria (47.3% plant reduction across variables; Fig. 2.7). Again, because biological control applications are all eventually in the field, subsamples of an area in the field may give a more realistic evaluation of biological control than inlab aquaria. Similarly, it is never the goal to control individual plants, and for plants that senesce like Eurasian watermilfoil (Myriophyllum spicatum), individual plant replicates may not be a good way to gauge control efficacy. Because of this, subsamples of field sites may provide the most realistic and meaningful evaluations of control efficacy. Though subsamples of an area provided smaller effect sizes than whole lakes (Fig. 2.7), six of the seven whole-lake studies were observational which again may have inflated the whole lake effect size in this analysis. Whole lakes may also be a viable option when possible if experimental data are used, though finding comparable control lakes would be difficult, and surveying large lakes would seem to prove logistically difficult.

46 36 Finally, the qualitative data in Fig. 2.1 can be used to make some more suggestions for future research. First, because biological control ideally provides longterm or even permanent control of the target plant (McFadyen 1998), more long term studies should be performed in the future, a need also noted by McFadyen (1998). Longterm studies (n = 38), which we liberally defined as two or more years, represent many fewer studies than short-term studies (n = 97 total; Fig. 2.1). Though the continuous model that tested study duration as a potential influence on effect size was nonsignificant (see above), more long-term studies will be needed to more fully understand what, in the end, is supposed to be a long-term solution for controlling invasive plants. Similarly, non-target effect studies have received the least amount of research relative to the other categories in Fig. 2.1, so this is a place where more research could be performed, especially since non-target effects can be a large problem in plant biological control applications (McFadyen 1998). Many studies have been performed on host-range of the biological control agents (included in distribution and life history bar in Fig. 2.1), so perhaps many potential non-target effects are alleviated by so closely studying host range. Overall, the results described here allow it to be said that biological control is an effective management tool in the fight to control invasive aquatic and wetland plants. The results also give credence to the basic qualitative suggestions that aquatic herbivory may be more important than historically thought (Lodge 1991; Newman 1991), as almost all the herbivores used here significantly damage their host-plants across many variables (Fig. 2.2). Meta-analyses like the one reported here are valuable tools for quantifying

47 37 relationships across studies. By weighting studies based on sample size and variability, meta-analysis reduces the subjectivity with which results are judged in purely narrative reviews (Gurevitch and Hedges 2001). Several of the conclusions made here about the differences in effect based on experimental design could not have been made without combining and comparing data across studies quantitatively, highlighting the value of meta-analyses such as the one presented here. For example, this meta-analysis highlights a potential effect of experimental design on biological control studies; that observational studies are more likely to report large agent effects than experimental studies. While not statistically significant, these analyses made it possible to test whether this effect was due to using initial values in place of controls or whether observational studies show larger effect sizes for some other reason. This study, along with Stiling and Cornelissen (2005), has provided what is hoped to be a valuable quantitative review of the biological control literature by providing a better understanding of the biological control of aquatic and wetland plants. Acknowledgments The authors thank NESCent, Jessica Gurevitch, Kerrie Mengersen, and Mark Lajuenesse for holding the meta-analysis workshop that made this study possible. We also thank two anonymous reviewers for their helpful comments on a previous draft.

48 38 References Adams, D. C., J. Gurevitch and M. S. Rosenberg Resampling tests for metaanalysis of ecological data. Ecology 78: Andres, L. A. and F. D. Bennett Biologcial control of aquatic weeds. Annual Review of Entomology 20: Cuda, J. P., R. Charudattan, M. J. Grodowitz, R. M. Newman, J. F. Shearer, M. L. Tamayo and B. Villegas Recent advances in biological control of submersed aquatic weeds. Journal of Aquatic Plant Management 46: Cooper, H. and L. V. Hedges (Eds.) The Handbook of Research Synthesis. Russell Sage Foundation, New York. DeLoach, C. J Biological control of weeds in the United States and Canada. In: Luken, J. O., Thieret, J. W. (Eds.). Assessment and management of plant invasions. Springer, New York, pp Denoth, M., L. Frid and J. H. Myers Multiple agents in biological control: improving the odds? Biological Control 24: Englund, G., O. Sarnelle and S. D. Cooper The importance of data-selection criteria: meta-analyses of stream predation experiments. Ecology 80: Gurevitch, J. and L. V. Hedges Statistical issues in ecological meta-analyses. Ecology 80: Gurevitch, J. and L. V. Hedges Meta-analysis: combining the results of independent experiments. In: Scheiner, S.M., Gurevitch, J. (Eds.). Design and Analysis of Ecological Experiments. Oxford University Press, NY, pp

49 39 Hedges, L. V., J. Gurevitch and P. S. Curtis The meta-analysis of response ratios in experimental ecology. Ecology 80: Julien, M. H. and M. W. Griffiths. (Eds.) Biological Control of Weeds: a World Catalogue of Agents and their Target Weeds. Fourth edition. CABI Publishing, Wallingford, UK. Lodge, D. M Herbivory on freshwater macrophytes. Aquatic Botany 41: McCarthy, J. M., C. L. Hein, J. D. Olden and M. J. Vander Zanden Coupling longterm studies with meta-analysis to investigate impacts of non-native crayfish on zoobenthic communities. Freshwater Biology 51: McFadyen, R. E. C., Biological control of weeds. Annual Review of Entomology 43: Mc Kay, F., M. Oleiro, G. C. Walsh, D. Gandolfo, J. P. Cuda, J.P. and G. S. Wheeler Natural enemies of the Brazilian peppertree (Sapindales: Anacardiaceae) from Argentina: their possible use for biological control in the USA. Florida Entomologist 92: Newman, R. M Herbivory and detritivory of freshwater macrophytes by invertebrates: a review. Journal of the North American Benthological Society 10: Newman, R. M Biological control of Eurasian watermilfoil by aquatic insects: basic insights from an applied problem. Archive fur Hyrobiologie 159: Osenberg, C. W., O. Sarnelle and S. D. Cooper The application of biological models in meta-analysis. The American Naturalist 150:

50 40 Reeves, J. L., P. D. Lorch, M. W. Kershner and M. A. Hilovsky Biological control of Eurasian watermilfoil by Euhrychiopsis lecontei: assessing efficacy and timing of sampling. Journal of Aquatic Plant Management 46: Rosenberg, M. S., D. C. Adams and J. Gurevitch MetaWin: statistical software for meta-analysis. Version 2.1. Sinaur Assiciates, Sunderland, Massachusetts. Rosenthal, R The file drawer problem and tolerance for null results. Psychological Bulletin 86: Sheldon, S. P. and R. P. Creed, Jr Use of a native insect as a biological control for an introduced weed. Ecological Applications 5: Shurin, J. B., E. T. Borer, E. W. Seabloom, K. Anderson, C. A. Blanchette, S. D. Cooper and B. S. Halpern A cross-ecosystem comparison of the strength of trophic cascades. Ecology Letters 5: Stiling, P. and T. Cornelissen What makes a successful biological control agent? A meta-analysis of biological control agent performance. Biological Control 34: Wilson, J. R. U., M. Rees and O. Ajuonu Population regulation of a classical biological control agent in its introduced range: larval density dependence in Neochetina eichhorniae, a biological control agent of water hyacinth Eichhornia crassipes. Bulletin of Entomological Research 96:

51 41 Appendix 2.A. Biological control agents and their target plants used in the literature search for this study. The agents are grouped by their target plant. Bolded agent names indicate agents for which data were able to be used in at least one of the above analyses. Biological Control Agent Agent Type Order Family Target Plant Agasicles hygrophila Insect Coleoptera Chrysomelidae Alternanthera philoxeroides Amynothrips andersoni Insect Thysanoptera Phlaeothripidae Alternanthera philoxeroides Arcola malloi Insect Lepidoptera Pyralidae Alternanthera philoxeroides Disonycha argentinensis Insect Coleoptera Chrysomelidae Alternanthera philoxeroides Stenopelmus rufinasus Insect Coleoptera Curculionidae Azolla filiculoides Bellura densa Insect Lepidoptera Noctuidae Eichhornia crassipes Cornops aquaticum Insect Orthoptera Acrididae Eichhornia crassipes Eccritotarsus catarinensis Insect Hemiptera Miridae Eichhornia crassipes Megamelus scutellaris Insect Hempitera Delphacidae Eichhornia crassipes Neochetina bruchi Insect Coleoptera Curculionidae Eichhornia crassipes Neochetina eichhorniae Insect Coleoptera Curculionidae Eichhornia crassipes Niphograpta albiguttalis Insect Lepidoptera Pyralidae Eichhornia crassipes Orthogalumna terebrantis Mite Acarina Galumnidae Eichhornia crassipes Sameodes albiguttalis Insect Lepidoptera Pyralidae Eichhornia crassipes Xubida infusella Insect Lepidoptera Pyralidae Eichhornia crassipes Bagous affinis Insect Coleoptera Curculionidae Hydrilla verticillata Bagous hydrillae Insect Coleoptera Curculionidae Hydrilla verticillata Cricotopus lebetis Insect Diptera Chironomidae Hydrilla verticillata

52 Biological Control Agent Agent Type Order Family Target Plant Hydrellia balciunasi Insect Diptera Ephydridae Hydrilla verticillata Hydrellia pakistanae Insect Diptera Ephydridae Hydrilla verticillata Ischnodemus variegatus Insect Hemiptera Blissidae Hymenachne amplexicaulis Lysathia ludoviciana Insect Coleoptera Chrysomelidae Ludwigia grandiflora Cataclysta camptozonale Insect Lepidoptera Pyralidae Lygodium microphyllum Neomusotima conspurcatalis Insect Lepidoptera Crambidae Lygodium microphyllum Galerucella calmariensis Insect Coleoptera Chrysomelidae Lythrum salicaria Galerucella pusilla Insect Coleoptera Chrysomelidae Lythrum salicaria Hylobius transversovittatus Insect Coleoptera Curculionidae Lythrum salicaria Myzus lythri Insect Homoptera Aphididae Lythrum salicaria Nanophyes marmoratus Insect Coleoptera Curculionidae Lythrum salicaria Boreioglycaspis melaleucae Insect Hemiptera Psyllidae Melaleuca quinquenervia Eucerocoris suspectus Insect Hemiptera Miridae Melaleuca quinquenervia Lophodiplosis trifida Insect Diptera Cecidomyiidae Melaleuca quinquenervia Lophyrotoma zonalis Insect Hymenoptera Pergidae Melaleuca quinquenervia Oxyops vitiosa Insect Coleoptera Curculionidae Melaleuca quinquenervia Poliopaschia lithochlora Insect Lepidoptera Pyralidae Melaleuca quinquenervia Acentria ephemerella Insect Lepidoptera Pyralidae Myriophyllum spicatum Cricotopus myriophylli Insect Diptera Chironomidae Myriophyllum spicatum Euhrychiopsis lecontei Insect Coleoptera Curculionidae Myriophyllum spicatum Neohydronomous affinis Insect Coleoptera Curculionidae Pistia stratiotes Spodoptera pectinicornis Insect Lepidoptera Noctuidae Pistia stratiotes Synclita obliteralis Insect Lepidoptera Crambidae Pistia stratiotes Cyrtobagous salviniae Insect Coleoptera Curculionidae Salvinia molesta Cyrtobagous singularis Insect Coleoptera Curculionidae Salvinia molesta Paulinia acuminata Insect Orthoptera Pauliniidae Salvinia molesta Samea multiplicalis Insect Lepidoptera Pyralidae Salvinia molesta Megastigmus transvaalensis Insect Hymenoptera Torymidae Schinus terebinthifolius 42

53 Biological Control Agent Agent Type Order Family Target Plant Galerucella birmanica Insect Coleoptera Chrysomelidae Trapa natans Galerucella nymphaeae Insect Coleoptera Chrysomelidae Trapa natans Nanophyes japonica Insect Coleoptera Curculionidae Trapa natans 43

54 44 Appendix 2.B. Papers used in the above analyses. Due to space constraints, article titles were not included. Full names of agents and their target plants can be seen in Appendix 2.A. Reference Journal (volume:pages) Agent Target Plant Aguilar et al Biocontrol (48: ) N. eichhorniae & N. bruchi E. crassipes Ajuonu et al Afr. Entomol. (11: ) N. eichhorniae & N. bruchi E. crassipes Ajuonu et al Biocontrol (54: ) N. eichhorniae E. crassipes Ajuonu et al Biocontrol (54: ) E. catarinensis E. crassipes Ajuonu et al Biocontrol (54: ) N. eichhorniae & E. catarinensis E. crassipes Albright et al Am. Midl. Nat. (152: ) G. pusilla & G. calmariensis L. salicaria Bashir et al Hydrobiologia (110:95-98) N. eichhorniae E. crassipes Bashir et al Hydrobiologia (110:95-98) N. bruchi E. crassipes Blossey and Schat 2007 Environ. Entomol. (26: ) G. calmariensis L. salicaria Caunter and Mohamed 1990 J. Plant. Prot. Trop. (7: ) N. eichhorniae E. crassipes Center et al Weed Sci. (30: ) N. eichhorniae E. crassipes Center et al Environ. Entomol. (36: ) B. melaleucae M. quinquenervia Creed and Sheldon 1993 Aquat. Bot. (45: ) E. lecontei M. spicatum Creed and Sheldon 1994 J. Aquat. Plant Manage. (32:21-26) A. ephemerella M. spicatum Creed and Sheldon 1994 J. Aquat. Plant Manage. (32:21-26) E. lecontei M. spicatum Creed and Sheldon 1995 Ecol. Appl. (5: ) E. lecontei M. spicatum Creed et al J. Aquat. Plant Manage. (30:75-76) E. lecontei & A. ephemerella M. spicatum Dech and Nosko 2002 Biological Control (23: ) G. pusilla L. salicaria Dech and Nosko 2002 Biological Control (23: ) G. calmariensis L. salicaria Del Fosse 1978 Entomophaga (23: ) N. eichhorniae & O. terebrantis E. crassipes Denoth and Myers 2005 Biological Control (32: ) G. calmariensis L. salicaria Ding and Blossey 2005 Environ. Entomol. (34: ) G. nymphaeae T. natans

55 45 Reference Journal (volume:pages) Agent Target Plant Ding et al Biological Control (37: ) G. birmanica T. natans Diop and Hill 2009 Afr. Entomol. (17:64-70) C. salviniae S. molesta Doyle et al Biological Control (24: ) H. pakistanae H. verticillata Doyle et al Biological Control (41: ) H. pakistanae H. verticillata Forno 1987 Bull. Entomol. Res. (77:9-18) C. salviniae & S. multiplicalis S. molesta Franks et al Environ. Entomol. (35: ) B. melaleucae M. quinquenervia Franks et al Biol. Invasions (10: ) B. melaleucae & O. vitiosa M. quinquenervia Godfrey et al Fla. Entomol. (77: ) B. affinis H. verticillata Grevstad 2006 Biological Control (39:1-18) G. pusilla & G. calmariensis L. salicaria Gross et al Oecologia (127: ) A. ephemerella M. spicatum Haag and Habeck 1991 J. Aquat. Plant Manage. (29:24-28) N. eichhorniae E. crassipes Haq and Sumangala 2003 Exp. Appl. Acarol. (29:27-33) O. terebrantis E. crassipes Heard and Winterton 2000 J. Appl. Ecol. (37: ) N. eichhorniae E. crassipes Heard and Winterton 2000 J. Appl. Ecol. (37: ) N. bruchi E. crassipes Hunt-Joshi et al Ecol. Appl. (14: ) H. transversovittatus L. salicaria Hunt-Joshi et al Ecol. Appl. (14: ) G. calmariensis L. salicaria Jiminez and Balandra 2007 Crop Protection (26: ) N. eichhorniae & N. bruchi (& pathogen) E. crassipes Johnson et al Aquat. Ecol. (31: ) A. ephemerella M. spicatum Lilie 2000 J. Aquat. Plant Manage. (38:98-104) E. lecontei M. spicatum Macrae et al J. Aquat. Plant Manage. (28:89-92) C. myriophylli M. spicatum Matos and Obrycki 2007 J. Insect. Sci. (7:Article 30) G. calmariensis L. salicaria Matos and Obrycki 2007 J. Insect. Sci. (7:Article 30) M. lythri L. salicaria Matos and Obrycki 2007 J. Insect. Sci. (7:Article 30) G. calmariensis & M. lythri L. salicaria McConnachie et al Biological Control (29: ) S. rufinasus A. filiculoides McGregor et al 1996 J. Aquat. Plant Manage. (34:74-76) L. ludoviciana L. grandiflora Moran 2004 Environ. Entomol. (33: ) N. eichhorniae and N. bruchi E. crassipes Morath et al Environ. Entomol. (35: ) B. melaleucae M. quinquenervia Newman and Beisboer 2000 J. Aquat. Plant Manage. (38: ) E. lecontei M. spicatum Newman et al Aquat. Bot. (53: ) E. lecontei M. spicatum Ogwang and Molo 2004 Biocontrol Sci. Technol. (14: ) N. eichhorniae & N. bruchi E. crassipes

56 46 Reference Journal (volume:pages) Agent Target Plant Overholt et al Fla. Entomol. (87: ) I. variegatus H. amplexicaulis Rayamajhi et al Plant Ecol. (192: ) B. melaleucae & O. vitiosa M. quinquenervia Rayamajhi et al Weed Sci. (56: ) B. melaleucae & O. vitiosa M. quinquenervia Reeves et al J. Aquat. Plant Manage. (46: ) E. lecontei M. spicatum Rizvi et al 1995 Indian J. of Plant Prot. (23: ) N. eichhorniae & N. bruchi E. crassipes Sheldon and Creed 1995 Ecol. Appl. (5: ) E. lecontei M. spicatum Sosa et al Biological Control (42: ) M. scutellaris E. crassipes Tipping et al. 2008a Biological Control (44: ) B. melaleucae & O. vitiosa M. quinquenervia Tipping et al. 2008b Aquat. Bot. (88: ) C. salviniae S. molesta Van and Center 1994 Weed Sci. (4: ) N. eichhorniae E. crassipes

57 CHAPTER III BIOLOGICAL CONTROL OF EURASIAN WATERMILFOIL BY EUHRYCHIOPSIS LECONTEI: ASSESSING EFFICACY AND TIMING OF SAMPLING Reprinted with permission from Journal of Aquatic Plant Management (see Appendix II for permissions): Reeves, J. L., P. D. Lorch, M. W. Kershner, and M. A. Hilovsky Biological control of Eurasian watermilfoil by Euhrychiopsis lecontei: assessing efficacy and timing of sampling. Journal of Aquatic Plant Management 46: Abstract The milfoil weevil, Euhrychiopsis lecontei Dietz, is a biological control agent for Eurasian watermilfoil (EWM; Myriophyllum spicatum L.), a nuisance aquatic macrophyte. EnviroScience, Inc. (Stow, OH) rears and stocks E. lecontei for management of EWM infestations. Here, we analyze data collected by EnviroScience, Inc. from treatment (weevil-stocked) and control (unstocked) EWM beds in 30 Michigan and 47

58 48 Wisconsin lakes over six years. Initial and final EWM and weevil densities were compared with lake-specific average and maximum lake depths, and lake surface area. The analyses showed substantive variability of weevil efficacy to control EWM. No significant associations were seen between average or maximum lake depth, or lake surface area on final weevil densities or plant density changes. Only the number of days between initial and final surveys and timing of final data collection proved significant in determining final EWM densities. As more time passed between surveys, final EWM densities significantly decreased at treatment sites, possibly due to weevils, with a much smaller decrease at control sites. Also, EWM densities declined at both treatment and control sites when final survey data were collected after the start of September, a phenomenon not seen when final data were collected earlier in the year. These declines at control sites were likely due to plant senescence. Lake managers utilizing E. lecontei should consider the length of time between EWM surveys, as well as the timing of data collection to avoid the confounding effect of plant senescence on data interpretation. Introduction The exotic macrophyte, Eurasian watermilfoil (Myriophyllum spicatum L.; hereafter EWM), has been established in the United States since at least the 1940s (Sheldon and Creed 1995). Since its introduction, EWM has become one of the most troublesome invasive aquatic macrophytes in the country (Smith and Barko 1990), establishing itself in at least 45 states and three Canadian provinces (Newman 2004).

59 49 The rapid growth rate and vegetative reproduction of EWM allows it to outcompete native macrophytes (Madsen et al. 1991), leading to ecosystem damage with striking changes in physical and chemical properties of lakes, often negatively impacting other biota (Grace and Wetzel 1978; Boylen et al. 1999). Water recreation is also hampered by dense EWM infestations, with reductions in swimming and boating, an overall decline in the fishery, and reduced aesthetic quality of the water body (Smith and Barko 1990). Coupling these negative effects with possible EWM densities of >300 stems/m 2, millions of dollars are spent annually managing and removing this plant (Sheldon and Creed 1995; Eiswerth et al. 2000). Multiple methods have been developed to control EWM, including herbicide use (Parsons et al. 2001), mechanical treatment (Boylen et al. 1996), and biological control (Newman 2004). Here we consider the use of a native biological control agent, the milfoil weevil (Euhrychiopsis lecontei Dietz), an aquatic weevil native to northern North America that specializes on native Myriophyllum spp. (Newman 2004). Since the invasion by EWM, E. lecontei has expanded its range of host species to include EWM, preferring EWM to its native host, M. sibiricum (Solarz and Newman 1996, Marko et al. 2005). This preference for EWM has caused E. lecontei to be favorably considered as a biological control agent for EWM (Sheldon and Creed 1995). Further, these weevils show higher developmental rates on EWM than on native Myriophyllum spp. (Newman et al. 1997; Roley and Newman 2006). Newman (2004) provides a comprehensive review of weevil biology.

60 50 Euhrychiopsis lecontei feeding has been associated with EWM declines in multiple studies (Creed and Sheldon 1995; Sheldon and Creed 1995; Newman and Beisboer 2000). Larval stem mining and adult consumption of leaf tissue can cause the plants to collapse and fall out of the water column (Creed et al. 1992), which reduces photosynthetic activity and plant vigor, and may promote a recovery of previously displaced and shaded-out native macrophytes. Because of these impacts, E. lecontei weevils have considerable potential as an EWM biological control agent, especially because they seem to have limited effects on other, native Myriophyllum spp. (Sheldon and Creed 2003). Given that EWM reduction by milfoil weevils may have dramatic, positive effects on a given water body, an understanding of the factors correlated with E. lecontei effectiveness is important to aid in control efforts. Data collected by EnviroScience, Inc. (Stow, OH) regarding weevil stocking in 30 Michigan and Wisconsin lakes during six years were analyzed to provide insight into and characterize variation of E. lecontei efficacy as an EWM biological control agent. Methods EnviroScience, Inc. ( is an environmental consulting firm that rears and stocks E. lecontei into lakes with EWM infestations through a proprietary program termed the Middfoil process. Weevil stocking data supplied by EnviroScience were analyzed to characterize variation in E. lecontei efficacy

61 51 at controlling EWM beds. The dataset included initial (pre-weevil stocking) and final plant (stems/m 2 ) and weevil (weevils/stem) densities from 29 lakes in Michigan and one in Wisconsin (Fig. 3.1; Table 3.1). These data were collected as a regular part of weevilbased EWM control efforts. Reported weevil data were a combination of all life stages (as in Jester et al. 2000) and were collected along randomly located transects running through EWM beds. Plant density data were collected from 0.3m by 0.3m quadrats randomly placed in EWM beds. Plants were cut near the substrate, and stems and weevils were counted on shore. Data were collected from EWM beds where weevils were stocked (treatment sites) and control EWM beds where no weevils were stocked. The number of treatment sites per lake and year varied from 1 to 10, while there were only one or two control sites per lake and year. Treatment sites were identified as large, dense beds adjacent to as much natural shoreline as possible, usually on the north side of the lake and away from boat traffic. Control sites were selected to be as similar to treatment sites as possible, and were located at least several hundred meters away from treatment sites to reduce the influence of weevil dispersal among sites. The data presented here were collected during summers between 2000 and Because some lakes provided data for multiple years (n = 8 lakes), we analyzed data both within and across years for treatment and control sites because treatment and control sites remained the same from year to year. For plant density analyses within years, a proportional plant density change was calculated using the equation: [(final plant density initial plant density) / initial plant density] to quantify intensity of plant density changes relative to initial plant density. Negative values indicate sites where EWM density decreased, possibly due to

62 52 Fig Geographic distribution of the Michigan and Wisconsin lakes included in this study.

63 53 Table 3.1. Lakes used in this analysis, their locations (county, state), surface area, average and maximum depths, and mean/range of final weevil densities (# / stem) at treatment sites. NR = data not reported.

64 54 weevil control. This metric allowed comparison of treatments and control sites directly, but because sample sizes for controls are smaller (generally one per lake compared to up to 10 per lake for treatment sites), non-significant results for controls may have been due to smaller sample size. For this reason, in the lake characteristic analyses described below, a relative plant density difference was also calculated using the equation: [(C f T f ) (C i T i )], where C f is the final plant density at the control site, T f is the final plant density at the treatment site, C i is the initial plant density at the control site, and T i is the initial plant density at the treatment site. For relative plant density difference, positive values may indicate weevil effect. This metric takes changes in control plots into account when quantifying differences in plant density. Average and maximum lake depths and lake surface area data, provided by lake managers or collected by EnviroScience, Inc. employees, were analyzed for correlations with plant density metrics and final weevil densities for lakes where these data were available (Table 3.1). No individual bed-level characteristics were collected for either treatment or control sites. The statistical test used to analyze these lake characteristic data was the Two-Dimensional Kolmogorov-Smirnov (2DKS) test, a nonparametric analysis for nonrandom distributions in bivariate data with nonlinear relationships (test statistic is D ; Garvey et al. 1998). The 2DKS tests for significant threshold points on the x-axis where the mean and variance of the y-axis data significantly change. This is accomplished by comparing the proportion of points in each of four quadrants around each point to a null expectation of equal distribution among these four quadrants. The 2DKS test provides x,y coordinates (D BKS ) where the maximum difference between

65 55 observed and expected point distribution exists. One can use this point to identify the value on the x-axis where the mean and variance of points along the y-axis changes (see Garvey et al. 1998). This test does not indicate direction of association, which must be inferred from the shape of the point distribution within the plot. The available lake characteristic data were non-normally distributed and heteroscedastic, violating the assumptions required by linear regression models, so this nonparametric test was chosen to analyze these data. Because no data available for EWM bed-specific characteristics, and the plant and weevil density data were collected on a finer scale than the lake depths, lake-wide averages by year of plant density metrics and final weevil density were used for 2DKS analyses. In addition to the 2DKS tests, linear regression was used to analyze the effect of final weevil density on proportional plant density change, as well as the effect of days passed between surveys on final plant density. To analyze the effect of weevil treatment on final plant density for the lakes with multiple years reported (n = 8), we also performed linear regressions on the plots of final plant density by year for each lake. This approach was taken because seven of the eight lakes with multiple years reported had only two years of data. Repeated measures ANOVAs were inappropriate given that missing data were correlated with duration of the study. Treatment and control sites were analyzed separately using initial plant density as a covariate in the analysis to remove the effects of differences in initial plant density between lakes and years. A line was fitted to these final plant densities across years, and the slopes of these lines (partial regression coefficient) were used to indicate how final plant density changed over years. To analyze

66 56 the effect of weevil treatment across years relative to controls, a one-tailed, one sample t- test was performed on the differences of the slopes (treatment sites slope - control sites slope) from the above regressions. Prior to any analyses, a Kruskal-Wallis test was performed on initial plant densities at control (n = 41) vs. treatment (n = 167) sites to assure the initial densities were not significantly different between control and treatment beds (H = ; P = ). To test if time of year of data collection affected plant density outcomes, we calculated the day of year when final surveys were conducted (range: day ). We then plotted lake specific averages for proportional plant density change at control sites vs. proportional plant density change at respective treatment sites. The points in this plot were split into two groups by the range of day of year of final collection (days 219 to 240 and days 241 to 261). A Chi-Square test with Monte Carlo Simulation was performed on each date range, treating each graph quadrant as a cell in the analysis. Averages by lake and year for treatment and control sites were used in all the analyses described above. Using these averages produced non-significant results, so in some cases, less conservative tests were performed using each treatment and control site separately for each lake to increase the number of data points and tease out effects that were not present in the original analyses. The lake characteristic 2DKS analyses and the Chi-Square analysis of the final survey date ranges are reported here using lake and year averages. However, the analyses regarding days passed between surveys within years, and across year effects on final plant density, along with the linear regression of final

67 57 weevil density vs. proportional plant density change are reported using individual treatment and control sites. Results and Discussion Proportional plant density change was not related to final weevil density at treatment sites (Linear Regression F = ; P = ; n = 115) or controls (F = P = ; n = 38). Also, substantial variation was seen in the effectiveness of E. lecontei at controlling EWM (Fig. 3.2). In some cases, plant densities declined at treatment sites and increased at control sites (quadrant IV), while in others, plant densities decreased at control sites and increased at treatment sites (quadrant II). Congruently, plant densities dropped at both control and treatment sites in some instances (quadrant III), while they increased at both in others (quadrant I). This broad range of variation suggested the role of other factors affecting EWM abundance and warranted examination of the lake characteristics included in the data set. No significant associations were found between any of the available lake characteristics and the lake and year averages of proportional plant density change, relative plant density difference, or final weevil density (2DKS all P-values >0.05). These results are similar to Jester et al. (2000), who found that no lake-wide characteristics (including those examined here) influenced weevil abundance, which can in turn be directly related to efficacy of plant control (Newman 2004). Jester et al. (2000) did find, however, that weevil abundance negatively correlated with several bed-level

68 58 Fig Changes in proportional plant density [(final plant density initial plant density) / initial plant density] at lake-specific control sites vs. treatment sites. Points in Quadrant I represent an increase in EWM density at both control and treatment sites, implying no weevil effect. Points in Quadrant II represent EWM density increases at treatment sites and EWM density declines at control sites, implying no weevil effect. Points in Quadrant III represent EWM density declines at both control and treatment sites weevil effect unknown. Points in Quadrant IV represent EWM density declines at treatment sites and EWM density increases at control sites, implying a weevil effect on EWM density. Proportional Plant Density Change at Treatment Site II I III IV Proportional Plant Density Change at Control Site

69 59 characteristics, including plant bed depth and percent sandy shoreline adjacent to the plant bed. They also found that weevil abundance was positively correlated with number of M. spicatum apical tips per plant, distance from shore to milfoil bed, and percent natural shoreline in the Wisconsin lakes they studied. Bed-level characteristics such as these should be measured during future weevil stocking efforts or subsequent monitoring of those sites. See Newman (2004) for a review of factors known to influence weevil density. Weevil densities should clearly be related to reduction of EWM densities. Newman (2004) noted that weevil densities >1 weevil/stem can consistently control EWM, while densities < 0.1 weevils/stem may not be sufficient for plant control. Twenty-three of the 30 lakes reported here had average final weevil densities >0.1 weevils/stem (Table 3.1) for weevil stocking sites. However, we saw no relationship between final weevil density and proportional plant density change. Because of this, further research is warranted to better link when and how high weevil densities influence efficacy of EWM control. Another important consideration for using weevils in EWM control is the timing of data collection. Clearly, maximizing the time between stocking and end of season sampling is desirable to best measure weevil effects. The timing of data collection proved significant in determining final plant densities at both treatment, and to a lesser extent, control sites. Final EWM densities at individual sites were negatively correlated with the number of days that elapsed between initial and final plant surveys (range 35 to 106 days). This relationship held true for treatment sites (Linear Regression F = ; P <

70 ; n = 109), although the relationship was marginally non-significant for the control sites (F = ; P = ; n = 34; Fig. 3.3). This stronger relationship in treatment than control sites is consistent with the hypothesis that the weevils are having a negative impact on EWM because the plants are showing stronger declines with time at weevil stocking sites. However, the lack of a significant relationship at control sites may also be due to reduced statistical power resulting from a much smaller sample size. Analyses to examine the effect of weevils on final plant density while controlling for multiple treatment sites and multiple years with lake and year-specific averages of final plant densities for both control and treatment sites were not significant (all P-values >0.05). While the effect of weevils within years is one indication of success in controlling EWM, it is also important to look at effects over several years. No significant across year effects were seen; the differences of the slopes of the lines of the final plant densities in successive years at treatment vs. respective control sites were not significantly different than zero (one-tailed, one sample t-test t = ; P = ; n = 8). In examining these slope differences, however, five out of the eight lakes showed negative values, implying that plant densities were dropping more or increasing less at the treatment vs. respective control sites in these lakes. The overall average of these slope differences between treatment and control sites was , indicating that a weevil effect across years may be seen with more than eight lakes. If more data can be collected, this analysis should be performed again.

71 61 Fig The relationship between final EWM densities (stems/m 2 ) and the number of days passed between initial and final surveys for control and treatment sites Control Sites Regression P = ; R 2 = y = x Final Plant Density (stems / m 2 ) Treatment Sites Regression P<0.0001; R 2 = y = x Days Passed Between Surveys

72 62 Extended periods of time between initial and final surveys may give E. lecontei more time to reproduce and damage plant tissue; however, plant senescence may also contribute to reduced EWM densities observed during late summer. To examine a possible effect of plant senescence on final EWM densities, we characterized average lake and year specific proportional plant density changes at control and treatment sites by the day of year of the final survey (Fig. 3.4). Across lakes and years, the day of the year of the final EWM survey ranged from day 219 (around 10 August) through day 261 (mid- September). We split this range in half and graphed points in two groups: those whose final survey was conducted between days 219 to 240 and those surveyed between days 241 to 261. Day 241 provided a biologically relevant reference point of around the start of September. In examining EWM senescence in an Indiana reservoir, Landers (1982) noted that senescence started in August and became advanced by the start of September (See Fig. 5 in Landers 1982). Analysis of the distribution of the points in Fig. 3.4 showed that the data points from the first date range were randomly distributed among the four graph quadrants (Chi-Square χ 2 = 2.79; P = ; n = 19). However, during the second half of the date range, data points were significantly clustered in Quadrant III, where plant densities declined at both control and treatment sites (χ 2 = 11.80; P = ; n = 15) indicating that plant senescence may be skewing the data and our understanding of the role of E. lecontei in reducing EWM density. Weevil densities also start to decline in September as weevils leave lakes to overwinter on shore (Newman et al. 2001), making careful choice of final sampling date even more important.

73 63 Fig Average proportional EWM density change by lake and year at control vs. treatment sites. Filled circles represent lakes where the final survey was conducted between days 219 and 240 (~ 7 August to end of August); open circles represent lakes where the final survey was conducted after day 240 (very late August / early September). P-values are results from Chi-Square test (with Monte Carlo Simulation) of distribution of points among the four graph quadrants. Average Proportional Plant Density Change at Treatment Sites II III Days 219 to 240 P = Days 241 to 261 P = Average Proportional Plant Density Change at Control Sites I IV

74 64 Investigators examining the effectiveness of E. lecontei as a biological control agent for EWM using population augmentation should allow similar periods of time to go by between initial and final surveys both within and across years from lake to lake, while making sure to have all data collected before plant senescence becomes a factor. It remains desirable, however, to give the weevils as much time as possible to damage EWM before assessing success of the control effort. Because senescence seems to start in late August (though this may vary based on latitude and weather), we propose that final surveys of E. lecontei and EWM density should be collected by mid-august to reduce the risk of plant senescence or departing weevils skewing data interpretation. More research is warranted to better understand when and where the weevils will be effective at controlling EWM. We propose that future research be conducted in lakes with an equal number of similar treatment and control EWM beds, paired a priori, while keeping in mind the data collection timing issues noted above. Also, lakes should be surveyed for multiple years after a stocking event, both with and without yearly stocking, to gain a better understanding of how effective the weevils are over a period of years. The dataset considered here involved evaluation of patterns over a few months to three years, and large-scale changes in such complicated systems could reasonably be expected to require more time to develop. Further, more specific experiments utilizing collection of EWM bed-specific measures would help gain a deeper understanding of the variation of final weevil densities (Table 3.1). Data such as those collected in Jester et al. (2000) on M. spicatum bed-specific characteristics (and any number of novel variables) should be collected when possible. Taking these factors into account would provide a much greater

75 65 understanding of what affects weevil density (and therefore efficacy) and may eventually lead to a suite of characteristics that will predict weevil success. Acknowledgments We thank EnviroScience, Inc. for providing the data set analyzed here. We also thank two anonymous reviewers whose comments made this a much stronger manuscript. References Boylen, C. W., L. W. Eichler and J.D. Madsen Loss of native aquatic plant species in a community dominated by Eurasian watermilfoil. Hydrobiologia 415: Boylen, C. W., L. W. Eichler and J. W. Sutherland Physical control of Eurasian watermilfoil in an oligotrophic lake. Hydrobiologia 340: Creed, R. P., Jr. and S. P. Sheldon Weevils and watermilfoil: did a North American herbivore cause the decline of an exotic plant? Ecological Applications 5: Creed, R. P., Jr., S. P. Sheldon and D. M. Cheek The effect of herbivore feeding on the buoyancy of Eurasian watermilfoil. Journal of Aquatic Plant Management 30:75-76.

76 66 Eiswerth, M. E., S. G. Donaldson and W. S. Johnson Potential environmental impacts and economic damages of Eurasian watermilfoil (Myriophyllum spicatum) in western Nevada and northern California. Weed Technology 14: Garvey, J. E., E. A. Marschall and R. A. Wright From star charts to stoneflies: detecting relationships in continuous bivariate data. Ecology 79: Grace, J. B. and R. G. Wetzel The production biology of Eurasian watermilfoil (Myriophyllum spicatum L.): a review. Journal of Aquatic Plant Management 16:1-11. Jester, L. L., M. A. Bozek., D. R. Helsel and S. P. Sheldon Euhrychiopsis lecontei distribution, abundance, and experimental augmentations for Eurasian watermilfoil control in Wisconsin lakes. Journal of Aquatic Plant Management 38: Landers, D. H Effects of naturally senescing aquatic macrophytes on nutrient chemistry and chlorophyll a of surrounding waters. Limnology and Oceanography 27: Madsen, J. D., J. W. Sutherland, J. A. Bloomfield, L. W. Eichler and C. W. Boylen The decline of native vegetation under dense Eurasian watermilfoil canopies. Journal of Aquatic Plant Management 29: Marko, M. D., R. M. Newman and F. K. Gleason Chemically mediated host-plant selection by the milfoil weevil: a freshwater insect-plant interaction. Journal of Chemical Ecology 31:

77 67 Newman, R. M Biological control of Eurasian watermilfoil by aquatic insects: basic insights from an applied problem. Archic fur Hydrobiologie 159: Newman, R. M. and D. D. Biesboer A decline of Eurasian watermilfoil in Minnesota associated with the milfoil weevil, Euhrychiopsis lecontei. Journal of Aquatic Plant Management 38: Newman, R. M., M. E. Borman and S. W. Castro Developmental performance of the weevil Euhrychiopsis lecontei on native and exotic watermilfoil host plants. Journal of the North American Benthological Society 16: Newman, R. M., D. W. Ragsdale, A. Milles and C. Oien Overwinter habitat and the relationship of overwinter to in-lake densities of the milfoil weevil, Euhrychiopsis lecontei, a Eurasian watermilfoil biological control agent. Journal of Aquatic Plant Management 39: Parsons, J. K., K. S. Hamel, J. D. Madsen and K. D. Getsinger The use of 2,4-D for selective control of an early infestation of Eurasian watermilfoil in Loon Lake, Washington. Journal of Aquatic Plant Management 39: Roley, S. and R. M. Newman Developmental performance of the milfoil weevil, Euhrychiopsis lecontei (Coleoptera: Curculionidae), on northern watermilfoil, Eurasian watermilfoil, and hybrid (northern X Eurasian) watermilfoil. Environmental Entomology 35: Sheldon, S. P. and R. P. Creed, Jr Use of a native insect as a biological control for an introduced weed. Ecological Applications 5:

78 68 Sheldon, S. P. and R. P. Creed, Jr The effect of a native biological control agent for Eurasian watermilfoil on six North American watermilfoils. Aquatic Botany 76: Smith, G. S. and J. W. Barko Ecology of Eurasian watermilfoil. Journal of Aquatic Plant Management 28: Solarz, S. L. and R. M. Newman Oviposition specificity and behavior of the watermilfoil specialist Euhrychiopsis lecontei. Oecologia 106:

79 CHAPTER IV VISION IS IMPORTANT FOR PLANT LOCATION BY THE PHYTOPHAGOUS AQUATIC SPECIALIST EUHRYCHIOPSIS LECONTEI DIETZ (COLEOPTERA: CURCULIONIDAE) Reprinted with permission from Springer (See Appendix II for permissions): Reeves, J. L., P. D. Lorch and M. W. Kershner Vision is important for plant location by the phytophagous aquatic specialist Euhrychiopsis lecontei Dietz (Coleoptera: Curculionidae). Journal of Insect Behavior 22: Abstract The aquatic milfoil weevil Euhrychiopsis lecontei Dietz (Coleoptera: Curculionidae) is a specialist on Myriophyllum spp. and is used as a biological control agent for Eurasian watermilfoil (Myriophyllum spicatum L.), an invasive aquatic macrophyte. We show evidence that visual cues are important for plant detection by these weevils. Weevils had difficulty locating plants in dark conditions and were highly attracted to plant stems in the light, even when the plant sample was sealed in a vial. However, weevils were equally attracted to both M. spicatum and another aquatic 69

80 70 macrophyte, coontail (Ceratophyllum demersum L.) in vials. Turbidity (0-100 NTU) did not significantly influence visual plant detection by the weevils. This work fills a void in the literature regarding visual plant location by aquatic specialists and may help lead to a better understanding of when and where these weevils will find, accept, and damage their target host-plants. Introduction The aquatic milfoil weevil (Euhrychiopsis lecontei Dietz; Coleoptera: Curculionidae) is used as a biological control agent for Eurasian watermilfoil (Myriophyllum spicatum L.), a widespread invasive aquatic macrophyte (Sheldon and Creed 1995; see Newman 2004 for a review of weevil life history and use as a biological control agent). Euhrychiopsis lecontei, native to northern North America, is a specialist on native Myriophyllum spp., but has expanded its host range to include M. spicatum (Newman 2004) since the plant was introduced circa 1940 from Eurasia (Sheldon and Creed 1995). In fact, E. lecontei has been used to control M. spicatum not only because of its negative effects on the plant, but also because weevils were observed to prefer waterborne chemical attractants produced by M. spicatum over other Myriophyllum spp. (Solarz and Newman 1996; Marko et al. 2005), and because weevils develop faster on M. spicatum than on native Myriophyllum spp. (Newman et al. 1997; Roley and Newman 2006). There are several details of the weevil life cycle important to the research described below. First, during the spring and summer, weevils remain fully submerged,

81 71 foraging, mating and laying eggs on Myriophyllum stems. Eggs are laid on the apical meristems of the plants, after which larvae mine the plant stem and pupate inside it (Newman 2004). The damage caused by adults foraging on leaf tissue and especially by larvae can cause host watermilfoil to fall from the water column and die back. The weevils overwinter as adults in terrestrial leaf litter along lake shorelines, from which they must enter the water or fly to find new host-plants in the spring (Newman et al. 2001). Marko et al. (2005) studied the chemosensory abilities of E. lecontei. They found that glycerol and uracil were attractive to E. lecontei, and that M. spicatum exuded these chemicals at higher concentrations than native Myriophyllum spp. This work is clearly important in understanding the role of chemosensory capabilities in weevil selection of individual plants or plant species. However, since weevils overwinter on land and reenter water bodies to find host-plants during the spring, chemical cues are unlikely to play a role in initial host-plant location. Thus, our aim in this study was to evaluate the importance of vision in E. lecontei for plant location in the water. Newman (2004) notes that virtually nothing is known about plant location by submersed macrophyte specialists (but see Marko et al. 2005), so this work is also intended to contribute to the understanding of host-plant location by aquatic specialists. When compared with the role of chemical cues in host-plant detection/selection, visual plant detection by phytophagous insects has received relatively little attention (Prokopy and Owens 1983). However, vision is important in plant location by many terrestrial insect groups, including Heteroptera (Cook and Neal 1999), Hemiptera (Gish

82 72 and Inbar 2006; Patt and Setamou 2007; Vargas et al. 2005), Diptera (Serandour et al. 2006; Drew et al. 2003; Aluja and Prokopy 1993), and Coleoptera (Egusa et al. 2006; Hausmann et al. 2004; Stenberg and Ericson 2007). In the above studies (which do not represent an exhaustive list), only Serandour et al. (2006) examined host-plant detection by an aquatic organism. They found that the mosquito larva Coquillettidia richiardii Ficalbi uses environmental light cues to locate the roots of its emergent aquatic macrophyte host-plants. However, C. richiardii is merely lucifugous (negatively phototactic), thus this work does not represent a strong case for the use of visual cues in host-plant detection. Supporting the idea that E. lecontei may use vision in host-plant detection, evidence exists that some specialists may be more visually oriented than generalist counterparts. For example, in at least one study, a specialist aphid was shown to visually locate host-plants better than a generalist counterpart (Vargas et al. 2005). Further, it appears that vision is more important than olfactory cues for plant detection in the monophagous chrysomelid beetle species Altica engstroemi J. Sahlberg (Stenberg and Ericson 2007). On top of this, Anthonomus pomorum L., a terrestrial curculionid, may have a trichromatic vision system for host plant detection (Hausmann et al. 2004). Thus, evidence exists of specialists visually locating plants, and beetles that show vision to be important in host-plant location, so demonstrating the importance of vision in plant location by E. lecontei will add an excellent aquatic example to this relatively undocumented topic.

83 73 Since E. lecontei overwinters on land, vision is likely important for host-plant location, given that they are unlikely to detect plant exudates from host-plants while flying over lakes to locate host-plants. Similarly, weevils are unlikely to detect exudates from plants that may be tens of meters away if they randomly entered the water from shore while searching for plants. As further evidence for the role of visual host-plant detection, E. lecontei also has relatively large eyes (Fig. 4.1). For E. lecontei, finding host plants quickly seems extremely important, as they are poor swimmers, with their legs being adapted to holding onto plant stems rather than swimming. Thus, vision may ultimately help E. lecontei find their host-plants and get to the relative safety of a milfoil stem as quickly as possible. The goal of this study was to gain a better understanding of the role of visual cues in plant detection by the specialist herbivore E. lecontei. Understanding the process of plant location and selection may eventually enhance our ability to predict where and when E. lecontei will be able to find, select, accept, and damage M. spicatum. This study also contributes to the relatively small body of literature regarding visual plant detection by phytophagous insects, particularly those in aquatic systems. Methods The E. lecontei weevils used for these experiments were donated by EnviroScience, Inc. (Stow, OH, USA). They were housed as same-sex pairs (one individual was marked on back with nail-polish for identification purposes) in small, clear plastic, sealed 18.9 x 10.1 x 10.1 cm aerated tanks filled with dechlorinated tap water.

84 74 Fig Photograph of an adult Euhrychiopsis lecontei head. Note large eye size relative to head size.

85 75 The weevils were kept on a 14hL/10hD cycle using broad-spectrum fluorescent light. Prior to use in these experiments, weevils were given at least two days in their holding tanks to become accustomed to the conditions and light cycle. Each tank was stocked with two M. spicatum stems (~12 cm long) that were replaced as necessary throughout the experimental period. Myriophyllum spicatum used in this experiment was collected from various lakes/reservoirs in Portage County, OH, USA. All experimental trials took place between 10:00 A.M. and 2:00 P.M. to help control for any behavioral differences at different times of day. The behavioral arenas used in these experiments were round 6.5 cm tall, clear glass dishes with an inside diameter of 17 cm at the bottom of the dish. Each arena was filled to a depth of 2.5 cm with dechlorinated water, which was replaced after every 1-2 behavioral trials to reduce the presence of chemical cues in the water. The arenas were placed on a level surface atop circular grids (17 cm diameter; Fig. 4.2) with 12 segments (30 o each) radiating from the center to the edge of the arena. Each segment was broken up into 5 bands (1.7 cm each) from the center to the edge of the arena to quantify how far from the arena center each weevil traveled in a segment by the end of each trial. Initial observation confirmed, as Solarz and Newman (2001) noted, that E. lecontei is positively phototactic. To empirically test this, a dissecting scope light (~9300 lux) was aimed at a random location along the edge of the experimental arena and a weevil was released in the arena s center. Nineteen of 22 weevils swam directly to the light within 5 min, with the remaining three orienting themselves toward the light. The weevils would even follow the light around the dish when the light was moved haphazardly.

86 76 Fig Grid used to track weevil movement in experimental arenas. Segments (1-12) were used to track weevil direction and plant placement, while bands (labeled 1-5) were used to track distance traveled. Plants and/or vials were always placed in band 5 against the arena wall

87 77 To control for this phototaxis during experimental trials, a small fluorescent light bank was placed directly above the experimental arenas. The light bank consisted of two 59 cm bulbs (~90 lux) that were 21 cm apart and elevated ~22.5 cm. The side walls of the arenas were wrapped with aluminum foil to prevent light from entering from the side and in some trials, the arena tops were covered with white 21.6 x 27.8 cm paper to diffuse light entering the arena. Importance of light in plant location To test if light is important for E. lecontei in locating M. spicatum, our first experiment was to place a single M. spicatum meristem (~3 cm long) in a randomly chosen location along the periphery of an arena (multiple plant stems were used across trials and were soaked in dechlorinated water between uses). In each trial, a single weevil was released into the center of the arena and given 10 min. to find the plant. For each weevil, paired trials were conducted both in the light and in the dark (n = 28 trials per lighting treatment). We hypothesized that if the weevil found the plant in the light but not the dark, visual (light) cues may be driving plant location by the weevils. For each set of paired trials, the first lighting treatment was randomly selected and the remaining light or dark trial was immediately performed upon the conclusion of the first trial. Using the same weevil, the same plant stem was moved into the same relative location (i.e., band five in same segment) in a fresh arena for the remaining trial according to the grid in Fig For dark trials, arenas were placed into cardboard boxes with corners and seams covered with aluminum foil. The boxes in which the arenas were

88 78 placed were at least four times taller than the arena to prevent light from entering through the sides. The arenas were then also covered with smaller cardboard boxes close to the same size of the arena, again with corners and seams covered with aluminum foil. After the 10 min. movement period, weevil location was recorded. For statistical analysis of this light-dark experiment, a McNemar Test was used to compare the number of weevils that found the plant only in the light vs. only in the dark. Weevil attraction to plants in vials To further isolate the importance of visual cues in E. lecontei plant location by eliminating the role of chemical cues, the remaining experiments focused on weevil response to M. spicatum meristems that were sealed in water-filled, clear glass, one-dram vials. Prior to vial use in these experiments, vials were filled with food coloring, capped, and submerged in water to confirm that no food coloring (used as a surrogate for plant chemical cues) escaped from the vials. Finally, all vials were rinsed between trials. Vials containing M. spicatum meristems were placed at random locations around the arena s edge, as in the light-dark experiment. However, preliminary experiments showed weevils would often swim to the vial containing the plant but then swim away after a short time of not being able to get onto the plant. This did not happen when exposed plants were used in the light-dark experiment, where all weevils that found the plant had to be physically removed from the plant by the experimenter at trial s end. Because the weevils would not stay on the vials, the paper light diffuser was not used and weevils were observed directly. Since the light bank used in these experiments consisted

89 79 of two bulbs running horizontally above the experimental arenas, few (if any) directional light cues existed, as weevils showed no tendency to swim in any particular direction during initial observation. Since no plant chemical cues could be present in these experiments, and results of the light-dark experiment were so convincing relative to the role of vision in plant detection, all vial experiments were only conducted in the light. The first vial experiment was conducted in order to eliminate the possibility that weevils were attracted to the vials alone. For this experiment, we placed a vial at a random location along the periphery of the arena. The weevils (n = 15) were run though paired trails with M. spicatum-filled vials and empty (filled only with fresh tap water) vials, in random order, one immediately after the other. The time it took the weevil to contact each vial was recorded. Vials were placed in the same randomly selected location for each treatment and arenas and water were changed between each treatment. Weevils were given 10 min to contact the vial in each trial, and the trial was stopped if the weevil did not find the vials in the allotted time (this protocol was followed for all experiments that follow). For this experiment, a two-tailed paired t-test was used to compare the time to vial for each treatment type. Visual plant differentiation This experiment involved determining if weevils can visually differentiate between plant species. Using the methods described for the previous vial experiment, each weevil (n = 17) went through paired trials with either M. spicatum or coontail (Ceratophyllum demersum L.) meristems in sealed vials, in randomly chosen order.

90 80 Ceratophyllum demersum was chosen for use in this experiment because it is similar in overall color and form to M. spicatum but has different leaf and branch formation. The time it took the weevil to contact the vial containing a given species was recorded for each trial. The location of the vials along the arena wall was again randomly assigned and stayed the same for each trial pair. For this experiment, a two-tailed paired t-test was used to compare the time to vial for each treatment type. Effect of water turbidity on plant location The final experiment was conducted to determine how water turbidity affects visual plant location by the weevils. For this experiment, each weevil (n = 11) was run through three levels of turbidity: clear (unaltered, dechlorinated tap water), 40 NTU (nephelometric turbidity units, an arbitrary unit of turbidity measured by a nephelometer; ± 1.70 NTU [mean ± 1SD across treatment replicates]), and 100 NTU ( ± 2.48 NTU [mean ± 1SD across treatment replicates]). Turbidity was manipulated by varying the concentration of bentonite clay suspended in the arena water. These relatively high turbidity values were selected after initial trials demonstrated that much lower turbidity levels had no negative effect on plant location by weevils. Each weevil was run through the three turbidity treatments in random order, with the vial containing an M. spicatum meristem placed randomly in the same position along the arena periphery in each trial. Once more, time elapsed until contact with the vial was recorded. For each weevil, the slope of the relationship between turbidity level and time to the vial was estimated. Then, a one-tailed, one sample t-test was performed on these slopes with the

91 81 expectation that there would be a positive slope to these lines, indicating that weevils generally found vials more quickly at lower turbidity levels. This regression slope approach was used so we could keep track of individual weevils performance, as there is much variation in individual weevils ability to swim. Overall, variation in sample size across this series of experiments was a result of high levels of within-lab weevil mortality. Also, weevils that could not move from arena center (e.g., weevils that got stuck on their backs on the arena bottom) in experiments were not included in analyses. For all vial experiments, weevils that did not contact a vial within the allotted 10 min were excluded from the analyses as well. Results Importance of light in plant location In the light-dark experiment, weevils found the M. spicatum meristem in the light significantly more frequently than in the dark (McNemar P < ; n = 28), with 19 out of 28 weevils finding the plant in the light but not the dark. Of the remaining nine weevils, one weevil found the plant in the dark but not the light, four found the plant in the light and dark, and four never found the plant (Fig. 4.3). It should be noted that weevils still moved in the dark, as 20 of 23 weevils that did not find the plant in the dark were in the fourth or fifth band of the arena at the trial s end, with many swimming vigorously into or along the arena s edge.

92 82 Fig Results of light importance experiment. Number of observations shown for each combination of weevils finding plants across light and dark treatments McNemar Test: P < Number of Observations light only dark only both neither Category

93 83 Weevil attraction to plants in vials The weevils were significantly more attracted to sealed vials containing an M. spicatum stem than to empty vials (paired t-test P = ; t = ; n = 10; Fig. 4.4). Those weevils that contacted the empty vials appeared to do so through random movement around the experimental arena and showed no tendency to stay at the vial after initial contact. In contrast, for vials containing an M. spicatum meristem, weevils swam directly to the vial when it entered their field of view and swam against the vial for several seconds seemingly to get to the plant (data were not recorded relative to time spent attempting to get to plant meristem, as trials were ended shortly after vial contact). Four weevils were excluded from this analysis for not contacting the empty vial during the allotted time (i.e., 10 min) for the trials, while one weevil was excluded for not contacting the M. spicatum vial during the allotted time. Visual plant differentiation The weevils were not more attracted to M. spicatum than C. demersum in vials (paired t-test P = ; t = ; n = 16). The average time to vial for the M. spicatum treatment was 50.7 seconds, while the average time to vial for C. demersum treatment was 46.6 seconds. One weevil was excluded from this analysis for not contacting the vial within the allotted time in a C. demersum trial.

94 84 Fig Results of weevil attraction to plants in vials experiment. Average time to vial (± 1 SE) shown for each treatment. 250 Paired t-test: P = Average Time to Vial (sec) M. spicatum Empty Treatment

95 85 Effect of water turbidity on plant location Turbidity did not significantly affect weevil ability to locate M. spicatum-filled vials (one-tailed, one sample t-test P = ; t = ; n = 11). In this analysis, six of the eleven weevils had positive slopes for the relationship between their time to vial and turbidity, with the remaining five weevils having negative slopes. Slopes ranged from through 10.52, with an average slope of None of the individual slopes were significantly different than zero. Discussion These experiments demonstrated that vision plays an important role in M. spicatum location by E. lecontei. Not only did weevils find plants much more easily in the light than in the dark, the weevils also showed attraction to plants in vials sealed to prevent any chemical cues associated with M. spicatum from influencing their behavior. Given that weevils did not differentiate between M. spicatum and C. demersum, or at least were not more attracted to one than the other, vision is likely not the ultimate determinate of how these specialist weevils select their host-plants. Prokopy and Owens (1983) note that the historically accepted phases of host plant selection by phytophagous insects include: host habitat location, host location, host recognition/acceptance, and host suitability. Since these weevils over-winter on land and must crawl into or fly over water to find their host plants (Newman et al. 2001), vision is likely used for at least the host habitat location and host location phases. Chemical cues (Marko et al. 2005) may explain

96 86 host-plant selection by E. lecontei and how the weevils differentiate among macrophyte species and select appropriate plants. It is possible that vials used in these experiments distorted the image of the plants they contained, so it should not be concluded that weevils cannot visually differentiate between plant species based solely on these data. The weevils are also poor swimmers and may have been trying to get to the first plant they could to cling onto, especially in the novel and potentially stressful environment of the experimental arena. It seems from these data that color and/or shape contrast may be important for the weevils to visually identify potential plant hosts, supporting the appropriate/inappropriate landings hypothesis whereby insects may land haphazardly on green objects and use other (chemical) cues to discern appropriate hosts (Finch and Collier 2000). The results of the turbidity experiment may further support this notion. Even in highly turbid water, the weevils still easily found the plant-containing vials. The sample size for the turbidity experiment was small (n = 11), but it seems clear from the data that the weevils can discern and become attracted to the form and/or color of a plant at a distance of at least 8.5 cm (the distance from the center of the arena to its edge). Even under highly turbid conditions, all weevils contacted the vials via direct swimming paths from the release site, not random movement. These data may be promising to lake managers utilizing E. lecontei as a control for M. spicatum, as the weevils seem to not be highly affected by turbid water. The water was well illuminated for the turbidity experiment, however, so the weevils may react differently under field conditions.

97 87 Beyond visual attraction to the plants, the positively phototactic tendencies of the weevils may also prove interesting relative to navigational abilities. Initial pilot experimentation using the light-dark methods involved undiffused, ambient laboratory lighting. The weevils still tended to find the plants more in the light than in the dark. However, the weevils seemed to be swimming to the top of the water column in the direction of the closest ceiling light in the laboratory. Given this anecdotal result, we hypothesize that the weevils may use light cues to navigate to appropriate locations on their host plants (e.g., meristems for oviposition). The aphid Sitobion rosaeiformis Das is positively phototactic and may use sunlight as a directional cue to move upward onto plants for safety and food, especially after dropping off their host-plant for predator escape (Hajong and Raghu Varman 2002). Similarly, if E. lecontei becomes dislodged from its host-plant, the weevils may very well use sunlight at the water surface to reorient, particularly given that adult weevils feed on the top portions of the plants. Supporting this notion, Solarz and Newman (2001) note that E. lecontei tends to swim upward. Females may use sunlight to guide them to the floating apical meristems where oviposition occurs. Further, we speculate that males may potentially even use sunlight to help them locate these females. When considering host-plant selection in phytophagous insects, more attention is paid to chemical cues than visual cues (Prokopy and Owens 1983; see Bernays and Chapman 1994 for comprehensive information on chemical cues). However, flying insects do use visual cues to find host-plants (Prokopy and Owens 1983). This work adds to the small body of literature on visual host plant location by aquatic insects, showing

98 88 that they too use vision to find host-plants. The data presented in our study clearly indicate the importance of vision in plant location for these specialist aquatic weevils and provide another example of an insect that uses visual cues in locating its host-plant. Gaining an understanding of what affects host-plant location and selection by biological control agents such as E. lecontei can be of critical importance. Having knowledge of factors influencing weevil behavior, particularly relative to host-plant detection, should lead to a better understanding of when and where they will be most effective as a biological control agent. For instance, if E. lecontei adults that overwintered cannot find M. spicatum because of factors such as wave action, low light conditions, plants with unattractive morphology, etc., they cannot attack or control the plant. Research is warranted, then, regarding how well the weevils can detect submerged plants from the air if they are flying to find their host-plants, or how far they may be able to see underwater if they entered a lake. These questions should also be answered relative to the potential plant location influences noted directly above. While the work presented here is not a comprehensive study of cues used by E. lecontei to find M. spicatum, it does provide strong evidence that vision is important and justifies further research. For example, experiments could be devised to test how well weevils can find M. spicatum in mixed stands, or if weevils are visually more attracted to certain morphologies (i.e., size, branch number, color, etc.) of M. spicatum over others. An understanding of factors such as these will ideally allow for better prediction of biological control efficacy not only of E. lecontei, but any biological control agent for which vision is important in host-plant location.

99 89 Acknowledgments We thank EnviroScience, Inc. (Stow, OH; for their donation of the weevils used in these experiments. We also thank three anonymous reviewers for their helpful comments. References Aluja, M. and R. J. Prokopy Host odor and visual stimulus interaction during intratree host finding of Rhagoletis pomonella flies. Journal of Chemical Ecology 19: Bernays, E. A. and R. F. Chapman Host Plant Selection by Phytophagous Insects, Chapman and Hall, New York. Cook, C. A. and J. J. Neal Plant finding and acceptance behaviors of Anasa tristis (DeGeer). Journal of Insect Behavior 12: Drew, R. A. I., R. J. Prokopy and M. C. Romig Attraction of fruit flies of the genus Bactrocera to colored mimics of host fruit. Entomologia Experimentalis et Applicata 107: Egusa, S., T. Nishida, K. Fugisaki and H. Sawada Spatio-temporal abundance of flushing leaves shapes host selection in the willow leaf beetle, Plagiodera versicolora. Entomologia Experimentalis et Applicata 120:

100 90 Finch, S. and R. H. Collier Host-plant selection by insects a theory based on appropriate/inappropriate landings by pest insects of cruciferous plants. Entomologia Experimentalis et Applicata 96: Gish, M. and M. Inbar Host location by apterous aphids after escape dropping from the plant. Journal of Insect Behavior 19: Hajong S. R. and A. Raghu Varman A report of positive phototaxis exhibited by polymorphic forms of an aphid. Journal of Insect Behavior 15: Hausmann, C., J. Samietz and S. Dorn Visual orientation of overwintered Anthonomous pomorum (Coleoptera: Curculionidae). Environmental Entomology 33: Marko, M. D., R. M. Newman and F. K. Gleason Chemically mediated host-plant selection by the milfoil weevil: a freshwater insect-plant interaction. Journal of Chemical Ecology 31: Newman, R. M Biological control of Eurasian watermilfoil by aquatic insects: basic insights from an applied problem. Archiv fur Hydrobiologie 159: Newman, R. M., M. E. Borman and S. W. Castro Developmental performance of the weevil Euhrychiopsis lecontei on native and exotic watermilfoil host plants. Journal of the North American Benthological Society 16: Newman, R. M., D. W. Ragsdale, A. Milles and C. Oien Overwinter habitat and the relationship of overwinter to in-lake densities of the milfoil weevil, Euhrychiopsis lecontei, a Eurasian watermilfoil biological control agent. Journal of Aquatic Plant Management 39:63-67.

101 91 Patt, J. M. and M. Setamou Olfactory and visual stimuli affecting host plant detection Homalodisca coagulata (Hemiptera: Cicadellidae). Environmental Entomology 36: Prokopy, R. J. and E. D. Owens Visual detection of plants by herbivorous insects. Annual Review of Entomology 28: Roley, S. and R. M. Newman Developmental performance of the milfoil weevil, Euhrychiopsis lecontei (Coleoptera: Curculionidae), on northern watermilfoil, Eurasian watermilfoil, and hybrid (northern X Eurasian) watermilfoil. Environmental Entomology 35: Serandour, J., D. Rey and M. Raveton Behavioural adaptation of Coquillettidia (Coquillettidia) richiardii larvae to underwater life: environmental cues governing plant insect interaction. Entomologia Experimentalis et Applicata 120: Sheldon, S. P. and R.P. Creed, Jr Use of a native insect as a biological control for an introduced weed. Ecological Applications 5: Solarz, S. L. and R. M. Newman Oviposition specificity and behavior of the watermilfoil specialist Euhrychiopsis lecontei. Oecologia 106: Solarz, S. L. and R. M. Newman Variation in hostplant preference and performance by the milfoil weevil, Euhrychiopsis lecontei Dietz, exposed to native and exotic watermilfoils. Oecologia 126: Stenberg, J. A. and L. Ericson Visual cues override olfactory cues in the hostfinding process of the monophagous leaf beetle Altica engstroemi. Entomologia Experimentalis et Applicata 125:81-88.

102 92 Vargas, R. R., A. J. Troncoso, D. H. Tapia, R. Olivares-Donoso and H. M. Niemeyer Behavioural differences during host selection between alate virginoparae of generalist and tobacco-specialist Myzus persicae. Entomologia Experimentalis et Applicata 116:43-53.

103 CHAPTER V VISUAL PLANT DIFFERENTIATION BY THE MILFOIL WEEVIL, EUHRYCHIOPSIS LECONTEI DIETZ (COLEOPTERA: CURCULIONIDAE) Reprinted with permission from Springer (See Appendix II for permissions): Reeves, J. L. and P. D. Lorch Visual plant differentiation by the milfoil weevil, Euhrychiopsis lecontei Dietz (Coleoptera: Curculionidae). Journal of Insect Behavior 22: Introduction Euhrychiopsis lecontei Dietz (Coleoptera: Curculionidae) is an aquatic weevil native to the northern United States that specializes on plants in the genus Myriophyllum (milfoils). Euhrychiopsis lecontei is currently being used in the U.S. as a biological control agent for the invasive Eurasian watermilfoil (Myriophyllum spicatum L.: Haloragaceae; Newman 2004). The weevil has expanded its host range to include M. spicatum since the plant s introduction from Eurasia in the 1940s. Field and laboratory studies have shown that E. lecontei reduces M. spicatum densities, and also prefers M. 93

104 94 spicatum over other (native) milfoils (see Newman 2004 for summary of E. lecontei life history and use as a biological control agent). Euhrychiopsis lecontei can use visual cues for plant location (Reeves et al. 2009). However, during no-choice plant differentiation trials, weevils were attracted to both M. spicatum and coontail (Ceratophyllum demersum L.: Ceratophyllaceae; Reeves et al. 2009). Specifically, weevils swam equally quickly to sealed vials containing one or the other of these plants when one individual plant stem was the only visual stimulus in the behavioral arenas used. We present results here describing how the weevils responded to choice trials where both M. spicatum and C. demersum were placed side by side in the behavioral arenas. This short communication expands on the work presented by Reeves et al. (2009) which showed that weevils can use vision for at least initial host-plant detection, but may not be able to visually differentiate between plant species. Because Reeves et al. (2009) did not conclusively show in no-choice trials that weevils were unable to visually differentiate plants, the goal of this study was to more closely examine the question of visual plant differentiation by E. lecontei. Methods The same weevil husbandry methods, behavioral arenas, and basic experimental methodology were used as in Reeves et al. (2009). Plant stems (M. spicatum and C. demersum; meristems used for all experiments) were sealed in water-filled, one dram, clear glass vials (dimensions: ~4.5 cm tall; ~1.5 cm diameter) to prevent chemical cues from altering weevil behavior. Two vials (one vial containing M. spicatum; one vial

105 95 containing C. demersum; plant stems filled length of vial) were horizontally placed side by side (vial lids pointing away from one another) in random locations at the edges of the circular arenas (17 cm diameter) in ~2cm deep water. The left to right order of the plant species was randomized. Vials were placed side by side (as opposed to opposite one another) in the arenas to prevent the weevils from swimming to the first vial to enter their field of view (i.e., the vial they were facing when released into the arena) since the weevils are attracted to both plant species when viewed singularly (Reeves et al. 2009). Weevils were individually released into the center of the arena and the vial they contacted first was recorded. These trials were replicated with 32 weevils (weevil age unknown for all experiments; weevils for all experiments reared on M. spicatum), both male and female. To examine the possibility that weevils are using plant shape or growth form over plant color to differentiate host-plants (coontail is a similar color green), choice and nochoice trials were performed using sealed brown (i.e., yellow-brown, dead/decomposing stems) vs. green (healthy) M. spicatum stems with similar growth forms. The brown and green colors, along with similarity in growth form (i.e., stem size and leaf density), were visually determined subjectively by the experimenter. For the no-choice trials, the weevils (n = 20) were individually released into the center of an arena and the time to vial contact was recorded for brown vs. green stems separately, as in the no-choice M. spicatum vs. C. demersum trials in Reeves et al. (2009). Next, choice trials were performed by placing sealed brown and green stems side by side (in random left to right order) in the arenas (as in the plant species choice trials above), and recording which stem

106 96 the weevils (n = 32) contacted first. The smaller sample size for the no-choice color trials resulted from lower weevil availability for that experiment and no clear indication that more weevils would have affected the outcome of the experiment. For all experiments, lame weevils that could not move easily throughout the behavioral arenas or those that appeared to swim randomly with no interest in the plant samples were excluded from the experiments and analyses. Seven weevils were excluded from the milfoil vs. coontail choice experiment, nine weevils were excluded from the nochoice brown vs. green milfoil experiment, and three weevils were excluded from the brown vs. green milfoil choice experiment. All sample sizes presented in the Methods and Results and Discussion sections are responsive weevils and do not include the lame weevils noted here. The trials for all experiments were designed to be 10 minutes long, however most weevils made clear choices and swam to the plant stems within one min. Most weevils would swim directly toward the plant stems when they entered their field of view and then swim against the vial in which the plant stem was sealed for several seconds seemingly to get to the plant. Results and Discussion When the weevils (n = 32) were given a choice between M. spicatum and C. demersum sealed in vials, they showed a visual preference for M. spicatum over C. demersum (22 chose M. spicatum; 10 chose C. demersum; G-test of goodness-of-fit G = ; P = ). The results of this experiment suggest that E. lecontei may be able to visually differentiate between aquatic plant species while under water, and contradicts

107 97 the initial findings of the no-choice tests between M. spicatum and C. demersum in Reeves et al. (2009) where the weevils were attracted to both plants when time to plant contact was measured instead of direct plant choice/preference. The negative results of this initial experiment reported in Reeves et al. (2009) were likely due to the fact that the weevils are poor swimmers, and may have recognized C. demersum as a plant and therefore safety when it was the only visual stimulus present in the no-choice trials. However, when given a choice here, the weevils visually preferred M. spicatum to C. demersum. In the case of E. lecontei, and likely other systems, both choice and no-choice trials can be useful (but may produce different results) when quantifying host-plant location/selection behavior. That is, Reeves et al. (2009) showed in no-choice trials that E. lecontei is potentially visually attracted to plants in general (likely for safety), but here it is shown that weevils can visually select the appropriate host when given a choice between a host and non-host species. Both of these results stem from methodological differences between Reeves et al. (2009) and the work presented here, and both give useful indications about the plant locating mechanisms of E. lecontei. In the no-choice trials between brown and green stems, weevils were equally quickly attracted to brown and green stems when they were individually placed in the arenas. Nine of the twenty responsive weevils (45%) swam to the brown plant faster, and the remaining 11 responsive weevils (55%) swam to the green plant faster. There were no differences in the average time to vial contact for brown plants (11.9 ± 7.5 sec) and the average to green plants (10.8 ± 6.9 sec) (paired t-test t = ; P = ). When choice trials were performed with brown and green stems placed side-by-side (n = 32

108 98 weevils), similar results were obtained: 15 responsive weevils contacted the brown plant first, whereas 17 responsive weevils contacted the green plant first (G-test of goodnessof-fit G = ; P = ). Based on the data presented here, it seems as though E. lecontei can visually differentiate between plant species, at least when given a choice between M. spicatum and C. demersum. These data expand on those presented in Reeves et al. (2009) by demonstrating that vision is important for host-plant detection (and potentially selection) by E. lecontei. Prokopy and Owens (1978) speculated that monophagous/oligophagous insects may be more visually acute in host-plant detection than polyphagous insects. Our data support this notion, as E. lecontei, a weevil specializing on watermilfoils, is visually more attracted to its host-plant than a non-host-plant species. The leaves of two plants used here have different growth forms. Myriophyllum spicatum has whorled compound leaves with many leaflets, and C. demersum has stiff, whorled, single needle-like leaves, so it is possible that the weevils are discerning plant species based on their shape, especially given that weevils do not appear to distinguish between brown and green stems with similar growth form in either choice or no-choice trials. Future research should focus on determining the extent to which plant color or plant form is used by E. lecontei to visually locate potential host-plants, and which colors or shapes are most attractive to the weevils. This research should clearly include the manipulation of both plant shape and color. Also, it would be useful to examine weevil response to more plant species, including native milfoils (e.g., Myriophyllum sibiricum Komarov) that have similar leaf shapes and colors to M. spicatum.

109 99 Gaining an understanding of host-finding behavior in this and other biological control systems may eventually allow us to better predict when and where biological control agents can find and damage their target host-plants. In fact, mechanisms of host location are among the most important factors listed by Cuda et al. (2008) for understanding and predicting efficacy of biological control agents of submersed aquatic weeds. Beyond efficacy prediction, a deeper understanding of the role of vision in hostfinding behavior also may help lead to the development of traps to assess weevil presence and abundance. Because trap color and placement (Hoback et al. 1999) and trap type (Bloem et al. 2002) can influence the outcome of insect surveys, a greater understanding of the specific visual stimuli that are attractive to E. lecontei is essential for developing efficient traps for this weevil. The work presented here and by Reeves et al. (2009) advances our understanding of host-locating mechanisms used by E. lecontei and thus may eventually lead to better prediction of biological control efficacy and trap development. Acknowledgments We thank EnviroScience, Inc. (Stow, OH; for donating the weevils and plants used in these experiments. We also thank three anonymous reviewers for their helpful comments.

110 100 References Bloem, S., R. F. Mizell and C. W. O Brien Old traps for new weevils: new records for curculionids (Coleoptera: Curculionidae), brentids (Coleoptera: Brentidae), and anthribids (Coleoptera: Anthribidae) from Jefferson Co., Florida. Florida Entomologist 85: Cuda, J. P., R. Charudattan, M. J. Grodowitz, R. M. Newman, J. F. Shearer, M. L. Tamayo and B. Villegas Recent advances in biological control of submersed aquatic weeds. Journal of Aquatic Plant Management 46: Hoback, W. W., T. M. Svatos, S. M. Spomer and L. G. Higley Trap color and placement affects estimates of insect family-level abundance and diversity in a Nebraska salt marsh. Entomologia Experimentalis et Applicata 91: Newman, R. M Biological control of Eurasian watermilfoil by aquatic insects: basic insights from an applied problem. Archiv fur Hydrobiologie 159: Prokopy, R. J. and E. D. Owens Visual generalist with visual specialist phytophagous insects: host selection behavior and application to management. Entomologia Experimentalis et Applicata 24: Reeves, J. L., P. D. Lorch and M. W. Kershner Vision is important for plant location by the phytophagous aquatic specialist Euhrychiopsis lecontei Dietz (Coleoptera:Curculionidae). Journal of Insect Behavior 22:54-64.

111 CHAPTER VI VISUAL ACTIVE SPACE OF THE MILFOIL WEEVIL, EUHRYCHIOPSIS LECONTEI DIETZ (COLEOPTERA: CURCULIONIDAE) Under review at Journal of Insect Behavior at time of dissertation completion. Abstract Euhrychiopsis lecontei Dietz (Coleoptera: Curculionidae) is used as a biological control agent for the invasive aquatic macrophyte, Eurasian watermilfoil (Myriophyllum spicatum L.). Because E. lecontei overwinters on land and must find plants in lakes each spring, plant finding behaviors are essential to eventually understanding and predicting long term biological control. Here we further the understanding of plant finding by showing that E. lecontei is visually attracted to M. spicatum at up to 17.5 cm, and is even more attracted to plants than other attractive visual stimuli (i.e., walls of experimental troughs used in these experiments) within 15 cm. We also show that turbidity may affect visual plant finding at 15 cm. Using available data from this and other previous studies involving chemical cue use and other life history factors, we propose a testable conceptual model for how E. lecontei finds plants each year. This model may also be used to explain plant finding by aquatic phytophagous insects in general. 101

112 102 Introduction The specialist, native, aquatic milfoil weevil, Euhrychiopsis lecontei Dietz (Coleoptera: Curculionidae; body length ~2 mm), is a promising candidate for use as a biological control agent for the highly invasive macrophyte, Eurasian watermilfoil (Myriophyllum spicatum L.; Sheldon and Creed 1995; Newman 2004). Since the introduction of M. spicatum to the U.S. in the 1940 s, it has spread to over 45 states and three Canadian provinces (Newman 2004), causing much ecological and economic damage (Boylen et al. 1999; Smith and Barko 1990; Grace and Wetzel 1978). Euhrychiopsis lecontei has expanded its host range to include M. spicatum since the plant was introduced (Newman 2004), and even prefers M. spicatum over other native milfoils (Marko et al. 2005; Solarz and Newman 1996, 2001). Interestingly, E. lecontei also develops faster on M. spicatum than native Myriophyllum spp. (Newman et al. 1997; Roley and Newman 2006, Solarz and Newman 2001) and may not significantly damage native Myriophyllum spp. (Sheldon and Creed 2003). See Newman (2004) for a more comprehensive review of weevil life history and its use as a biological control agent. Euhrychiopsis lecontei overwinters as adults on land in on-shore leaf litter (Newman et al. 2001), and thus has the seemingly large challenge of relocating hostplants in the spring that are growing in a completely different habitat than that in which the weevils overwintered. Because biological control ideally provides long term control of problematic plants (McFadyen 1998), an understanding of how weevils find plants in the spring should be directly related to predicting under what conditions weevils can be expected to find plants, which in turn is directly related to predicting long term control

113 103 efficacy. Thus, plant finding behaviors may be important to understand not only in this system, but in any aquatic biological control system (Cuda et al. 2008). Both visual and chemical cues have been explored as plant finding cues for E. lecontei. The chemical cues glycerol and uracil (general aquatic plant exudates) appear to be used by E. lecontei to find plants (Marko et al. 2005). Visual cues are also very important (Reeves et al. 2009); E. lecontei can even visually differentiate plant species under water (Reeves and Lorch 2009). Because the use of visual cues in general appears to be important for E. lecontei, an important step in further exploring visual cue use is to determine the distance at which E. lecontei can become visually attracted to M. spicatum. With an understanding of how far weevils can see M. spicatum underwater, we will be closer to understanding how weevils make it back to their host-plants in the spring. Thus, one goal of the research presented here was to determine the distance at which E. lecontei become visually attracted to M. spicatum. We also explored the role of turbidity in potentially reducing plant detection distance. Active spaces are the distances around a plant within which a stimulus (either chemical or visual) is sufficiently strong to elicit a behavioral response from a phytophagous insect (Schoonhoven et al. 2005, p.144). In general, few studies have quantified the active space for either visual (n = 4) or chemical (n = 6) cues. None of these prior studies were aquatic systems (reviewed in Table 6.2 in Schoonhoven et al. 2005), so the research presented here is intended to add to the relatively small list of insects for which the size of active spaces has been estimated, especially for aquatic systems where no such data have yet been published.

114 104 The variability seen both within and across lakes in the effectiveness of E. lecontei at controlling M. spicatum (Reeves et al. 2008) may be related to factors affecting plant finding in the spring. The work presented here and by Marko et al. (2005), Reeves et al. (2009) and Reeves and Lorch (2009) all help to understand plant finding by E. lecontei. The results of these (and other) studies are used to provide a conceptual model for how E. lecontei, and likely other aquatic phytophagous insects, may find underwater plants in the spring. Methods Visual Active Space To determine how far weevils are visually attracted to M. spicatum stems underwater (i.e., their visual active space), we developed 3.08m long troughs (Fig. 6.1) to incrementally move sealed plant stems to different distances from the starting point and record how far the weevils can be from the plants and still be significantly visually attracted to them. A small portion of apical M. spicatum stem (sealed inside a one dram glass vial to prevent detection of any chemical cues; Reeves et al. 2009, Reeves and Lorch 2009) was placed in a random side relative to the mid-point of the trough. Individual weevils were released in the middle of the trough and given 5 minutes to choose which direction to swim (either toward or away from the plant stem). A choice was defined either by vial contact or swimming past the 8.5 cm mark from the center

115 105 Fig Cross section and dimensions of trough used for these experiments. End caps of essentially the same size and shape were attached to the ends of the trough using clear 100% silicone caulking, creating a water-tight seal. The asterisk shows where weevil was released relative to sides, and horizontal dotted line indicates approximate water level.

116 106 release point in the trough. This distance was chosen as the weevil choice point because 8.5 cm was the distance used to determine choice in Reeves et al. 2009, and also because 8.5 cm represents over 40 body lengths of the weevil, making it unlikely that such a large movement from the release point was random (especially since weevils are poor swimmers). Almost all weevils that swam past the 8.5 cm mark in the direction of the plant stem reached the vial and bumped against the vial for several seconds, apparently trying to get to the plant. For each distance examined in this study (described below), 20 weevils were tested individually. If a significant number of weevils (using Chi-Squared tests) swam toward or contacted the vial, we considered weevils to have been visually attracted to the plant at that distance. For each trial, the trough was filled with about 2 cm of dechlorinated tap water. Because chemical cues were not a large concern for these experiments with sealed vials, and because the volume of water required to fill the trough was substantial and difficult to completely remove, water was changed after every ~5 successful trials. For all experiments, multiple sealed plant stems were used both within and across experiments. For this visual active space experiment, much initial experimentation was done before weevils would start positively responding to the sealed plant stems. The successful methods involved lining the inside of the trough with aluminum foil (duller side up) and placing the fluorescent light banks that were used 60 cm above the troughs. The light banks were placed parallel to and behind the trough so the back side of the trough was in line with the front side of the light bank. Two light banks were used, mounted end to end. Each light bank consisted of two 122 cm fluorescent bulbs spaced ~3cm apart. The bulbs

117 107 (General Electric brand F40 Plant and Aquarium bulbs with Ecolux Technology) were 40 watts and emitted 1900 lumens. White copy paper (21.6 x 27.8 cm) was taped over the length of the bulbs to diffuse the light from the banks, as E. lecontei is positively phototactic (Reeves et al. 2009). The weevils responded to plants most strongly when ambient laboratory lighting was left on, so lab lighting was left on for all trials. Finally, in all trials, a wall of aluminum foil was put up at 70 cm from the trough s center in either direction. The weevils for this experiment (and those described below) were housed en masse in 37.9 L aerated aquaria under a 14:10 L:D cycle. Plant (M. spicatum) masses were haphazardly placed unrooted into the aquaria (filling half of aquaria) and replaced as necessary throughout the experimental period. No weevils were used more than once for a single distance or for different distances on the same day (except as noted below). Weevils may have been used in more than one experiment across days, however, as they were haphazardly picked from the aquaria. For all experiments, weevil age was unknown, and sex was not determined because sex did not matter in Reeves et al. (2009). Also, for all experiments, lame weevils that could not move easily throughout the trough or weevils that did not make a choice were excluded from the experiments and analyses. Between three and ten weevils were excluded from any set of trials (distances). Weevils for all experiments were donated by EnviroScience, Inc. (Stow, OH; and all plant stems were collected in Portage County, OH, USA. All trials were performed between around 10:00 a.m. and 2:00 p.m. to control for any potential behavioral differences at different times of day.

118 108 For this visual active space experiment (replicated twice), 15 cm was arbitrarily chosen as the starting point because previous work clearly showed that E. lecontei is visually attracted to plants from at least 8.5cm (Reeves et al. 2009). Based on initial success of 20 weevils at 15 cm, new sets of 20 weevils were tested at 25 cm, 20 cm, and finally 17.5 cm from the trough s center. In each trial, weevils were singularly released into the trough at the center and their choice (i.e., swimming toward or away from plant stem) was recorded. During this first replicate of distance experiments, weevils from the 20 cm trials were immediately used in 8.5 cm trials to confirm that weevils were attracted to the M. spicatum stems at this distance as they were in Reeves et al. (2009) even if they were not attracted at 20 cm. This was done to test whether or not a lack of response was due to general unresponsiveness, and also to show that the results from these methods are comparable to Reeves et al. (2009). This was the only instance in which a weevil was used more than once in a day for any of the experiments. The entire experiment described above was replicated a second time using a different set of weevils (no individuals from first replicate were used again in second replicate and weevils from 20 cm trials were not re-used for the 8.5 cm trials in replicate two). Effect of Turbidity on Active Space Turbidity was shown not to affect visual plant detection by E. lecontei in Reeves et al. (2009) at a distance of 8.5 cm. To test whether turbidity effects occur over larger distances, turbidity trials were performed to see how plant detection would be affected at

119 cm (the first distance beyond 8.5 cm as above). Two turbidity levels were used, 0 and 41.5 ± 2.0 NTU (Nephelometric Turbidity Units) that were created by suspending bentonite clay in water (as in Reeves et al. 2009). A smaller, ~36 cm long section of the same type of trough as above was used to more easily ensure getting all water and clay out of the trough when water was changed. For each turbidity value, 25 weevils were used with sealed plant stems randomly placed 15 cm from the center of the trough. To remain consistent with the methods above, the series of turbidity experiments were performed on different sets of weevils on different days, rather than each weevil being run in each turbidity level successively as in Reeves et al. (2009). For our statistical analyses, we used logistic regression to test for differences between replicates and distances in the visual active space experiment to determine if pooling the data between the two replicates was appropriate. For each distance in the active space experiment and each turbidity level in the turbidity experiment, Chi-Squared tests were used to determine whether significantly more than 50% of weevils swam toward the plant, or if the weevils contacted the vial before the trough wall. Chi-Squared tests were used beyond logistic regression to examine the results of each distance or turbidity level individually.

120 110 Results Visual Active Space We used logistic regression to test for differences between replicates and distances in whether weevils went toward the plant or not. The replicate by distance interaction term in the model was not significant (L-R Chi-Square = 0.54, P = 0.46), nor was the replicate term (L-R Chi-Square = , p = 0.98). The reduced model with just distance yielded a significant logistic regression (L-R Chi-Square = 6.38, P = 0.01). Because of this (and because both replicates were from two different sets of weevils with no overlap of individuals between replicates), we pooled the data between the two replicates for analysis of each distance using Chi-Squared tests. Table 6.1 and Fig. 6.2 show that beyond a distance of 15 cm, the plant stem became less attractive than the walls of the experimental trough (which are approximately 5 cm from the starting point). Thus, it can be said that within 15 cm, E. lecontei can clearly visually distinguish an M. spicatum stem and become attracted to it, even if another attractive visual stimulus (such as the trough wall) is closer. When the milfoil stem was placed beyond 15 cm, the closer attractive visual stimulus (i.e., the trough wall) became more attractive. At a distance of 17.5 cm, weevils still significantly swam in the direction of the plant stem, but they significantly swam to the wall first. At 20 centimeters and beyond, weevils stopped significantly swimming in the direction of the plant stems and significantly contacted the wall before the vial when they did swim toward the plant stem.

121 111 Table 6.1. Pooled visual active space results showing how many weevils swam toward the vial with a plant stem (out of 40), how many of these contacted the vial first, and whether they contacted the vial or trough wall first significantly more often. P- values are based on Chi-Squared test with 50% expectation. Distance (cm) # Swam in Direction of Plant out of 40 P-value (Chi-Square) Vial Contacted First P-value for First Contact < (plant contact first) (plant contact first) (wall contact first) (wall contact first) (wall contact first)

122 112 Fig Results of pooled active space replicates. Bar height indicates number of weevils that swam toward vials. Within bars, color indicates proportion of weevils that first contacted either trough wall (grey) or vial (black). Line across figure indicates null expectation of no choice by weevils. 40 Number in Direction of Plant (out of 40) *** ** ** A A B B Vial Contact First Wall Contact First B Distance (cm) **P < ***P < A = Significant proportion contacted vial first B = Significant proportion contacted wall first

123 113 Effect of Turbidity on Active Space Table 6.2 and Fig. 6.3 show that at 40 NTU, weevils did not become significantly attracted M. spicatum stems in troughs. Since it was clear that 40 NTU turbidity affected weevil behavior at 15 cm, neither 100 NTU trials (as in Reeves et al. 2009), nor 40 NTU trials at longer distances seemed necessary and thus were not performed. Discussion Based on our data, we can say that E. lecontei has a visual active space of at least 17.5 cm. Further, it can be said that E. lecontei has the visual capability to become more attracted to plant stems than other attractive visual stimuli (like the trough walls) within 15 cm. Though these distances may be relatively small compared to some terrestrial insects (2-10 m; Table 6.2 in Schoonhoven et al. 2005), 17.5 cm represents over 80 body lengths of E. lecontei. Because even at 25 cm the weevils were still trending towards the vial, our estimate of the visual active space of E. lecontei may be conservative, especially because the trough walls were at least somewhat attractive the weevils. We presume that the trough walls may have been perceived as structure (safety), or that the walls reflected or bent the light at the water s surface which may have attracted the weevils (E. lecontei is positively phototactic; Reeves et al. 2009). The successful methods here involved using the foil-covered walls to take away the ability for the weevils to interact with the ridges at the bottom of the trough (Fig. 6.1). They were attracted to these ridges and interacted

124 114 Table 6.2. Turbidity results showing how many weevils swam toward the vial with a plant stem (out of 25), how many of these contacted the vial first, and whether they contacted the vial or trough wall first significantly more often. P-values are based on Chi-Squared test with 50% expectation. Turbidity (NTU) # Swam in Direction of Plant out of 25 P-value (Chi-Square) Vial Contacted First P-value for First Contact (vial first) (wall first)

125 115 Fig Results of turbidity experiment. Bar height represents number of weevils at each turbidity level that swam toward vial. Within bars, color indicates proportion of weevils that first contacted either trough wall (grey) or vial (black). Line across figure indicates null expectation of no choice by weevils. 25 Number in Direction of Plant (out of 25) * A Vial Contacted First Wall Contacted First B Turbidity (NTU) *P < 0.05 A = Marginally insignificant proportion contacted vial first (P = ) B = Significant proportion contacted wall first

126 116 with them during initial experimentation (potentially because they were perceived as plant stems). Also, the foil may have taken away any strong directional light cues via reflection. Even though the trough walls were closer than the vials to the weevils in all cases, at up to 15 cm, the plant-filled vials were more attractive than the trough walls. The visual acuity of E. lecontei is strong at short distances such as these, especially considering that at a distance of 8.5 cm, similar looking plant species can be visually differentiated (Reeves and Lorch 2009). Besides distance from the plant, turbidity seemed to have an impact on plant finding by E. lecontei. At a distance of 15 cm, turbid water at 40 NTU significantly reduced visual plant detection. The proportion of weevils swimming toward the plant that touched the vial first was highly reduced as well. At a shorter distance of 8.5 cm, however, turbidities up to 100 NTU did not affect visual plant location by E. lecontei (Reeves et al. 2009), so turbidity may only affect E. lecontei at distances beyond 8.5 cm. If glycerol and uracil (the phytochemicals attractive to E. lecontei; Marko et al. 2005) can be detected beyond 8.5 cm from a plant in turbid conditions, perhaps E. lecontei would use these chemicals to get to the plants within their visual active space so they could then finally locate the plants using visual cues. Because both visual (Reeves et al. 2009; Reeves and Lorch 2009) and chemical (Marko et al. 2005) cue use has been documented for plant finding by E. lecontei, and because the overwintering habits of E. lecontei have been worked out to some extent (Newman et al. 2001), it becomes possible to conceptually model how all these factors may act together in plant finding. To broaden our overall understanding of plant finding

127 117 behavior, we have built a conceptual model comprised of data available across E. lecontei studies (Fig. 6.4). With the available data, it would seem that E. lecontei may use chemical cues such as glycerol and uracil (both ubiquitous among aquatic plants) to locate plants (Marko et al. 2005) if plants are not seen when weevils enter lakes (Step 2 in Fig. 6.4). Because the swimming distance in the y-mazes (28 cm total; 14 cm stems and arms) used in Marko et al. (2005) was longer the visual active spaces presented here (17.5 cm), weevils may perceive chemical cues from farther away than visual cues. Once a plant is within 17.5 cm, weevils likely use vision to more precisely locate (Reeves et al. 2009) and potentially differentiate (Reeves and Lorch 2009) plants (Step 3 in Fig. 6.4). In general, insects may switch from chemoreception to vision when good visual cues become available, as vision is a more precise search modality than chemoreception (Bell 1990), particularly when considering ubiquitous plant cues such as uracil and glycerol. Once E. lecontei has used both visual and chemical cues together to find plants, it is very likely that contact chemoreception (i.e., tasting plant) would be the final determinant of host-plant selection (Step 4 in Fig. 6.4), a common if not ubiquitous behavior amongst phytophagous insects (Bernays and Chapman 1994). It has even been proposed that aquatic herbivores in general may use contact chemoreception rather than olfaction for host location and selection (Spanhoff et al. 2005). While some of the steps in Fig. 6.4 are understood, some parts about how such a small weevil can find plants in large lakes each spring remain unanswered. For instance, the relative roles of flying over vs. crawling into lakes in spring are mostly unknown, though limited flight dispersal appears to be possible (Newman et al. 2001). Lake finding

128 118 Fig Conceptual model of plant finding by E. lecontei. The model covers the classically accepted phases of host-plant location and selection as noted in Prokopy and Owens (1983): host habitat finding (Step 1); host finding (Steps 2-3); host recognition and acceptance (Step 4), and host suitability (Step 5). Our model extends these steps to include overwintering habitat finding (Step 6), since weevils leave their respective lakes in the fall. Steps 2-5 may be true of aquatic phytophagous insects in general, while steps 1 and 6 will vary with the life history of the insect and plant in question.

129 Fig

130 120 in general remains unexplored (Step 1 in Fig. 6.4), along with the behaviors associated with finding appropriate on-shore over-wintering sites (Step 6 in Fig. 6.4). These habitat finding behaviors will also be important to understanding long term control efficacy, and both are places where future work should be done. Once these things begin to be elucidated, the variation in efficacy of E. lecontei in controlling M. spicatum can be more fully understood. We will have to know how E. lecontei finds plants in the spring before we can predict when E. lecontei will find and thus damage/control M. spicatum. The study presented here, along with the conceptual model and corresponding literature in Fig. 6.4, contribute to our understanding of plant finding by E. lecontei. Much of the conceptual model of plant finding in Fig. 6.4 may also apply to aquatic phytophagous insects in general (see Harms and Grodowitz 2009 for a comprehensive list of aquatic phytophagous insects). Acknowledgments We thank EnviroScience, Inc. ( for donating the weevils used in these experiments and their continual support of our research. References Bell, W. J Searching behavior patterns in insects. Annual Review of Entomology 35:

131 121 Bernays, E. A. and R. F. Chapman Host Plant Selection by Phytophagous Insects, Chapman and Hall, New York. Boylen, C. W., L. W. Eichler and J. D. Madsen Loss of native aquatic plant species in a community dominated by Eurasian watermilfoil. Hydrobiologia 415: Cuda, J. P., R. Charudattan, M. J. Grodowitz, R. M. Newman, J. F. Shearer, M. L. Tamayo and B. Villegas Recent advances in biological control of submersed aquatic weeds. Journal of Aquatic Plant Management 46: Grace, J. B. and R. G. Wetzel The production biology of Eurasian watermilfoil (Myriophyllum spicatum L.): a review. Journal of Aquatic Plant Management 16:1-11. Harms, N. E. and M. J. Grodowitz Insect herbivores of aquatic and wetland plants in the United States: a checklist from the literature. Journal of Aquatic Plant Management 47: Marko, M. D., R. M. Newman and F. K. Gleason Chemically mediated host-plant selection by the milfoil weevil: a freshwater insect-plant interaction. Journal of Chemical Ecology 31: McFadyen, R. E. C Biological control of weeds. Annual Review of Entomology 43: Newman, R. M Biological control of Eurasian watermilfoil by aquatic insects: basic insights from an applied problem. Archiv fur Hydrobiolgie 159:

132 122 Newman, R. M., M. E. Borman and S. W. Castro Developmental performance of the weevil Euhrychiopsis lecontei on native and exotic watermilfoil host plants. Journal of the North Amerocan Benthological Society 16: Newman, R. M., D. W. Ragsdale, A. Milles and C. Oien Overwinter habitat and the relationship of overwinter to in-lake densities of the milfoil weevil, Euhrychiopsis lecontei, a Eurasian watermilfoil biological control agent. Journal of Aquatic Plant Management 39: Prokopy, R. J. and E. D. Owens Visual detection of plants by herbivorous insects. Annuual Review of Entomology 28: Reeves, J. L. and P. D. Lorch Visual plant differentiation by the milfoil weevil, Euhrychiopsis lecontei Dietz (Coleoptera: Curculionidae). Journal of Insect Behavior 22: Reeves, J. L., P. D. Lorch, M. W. Kershner and M. A. Hilovsky Biological control of Eurasian watermilfoil by Euhrychiopsis lecontei: assessing efficacy and timing of sampling. Journal of Aquatic Plant Management 46: Reeves, J. L., P. D. Lorch and M. W. Kershner Vision is important for plant location by the phytophagous aquatic specialist Euhrychiopsis lecontei Dietz (Coleoptera: Curculionidae). Journal of Insect Behavior 22: Roley, S. and R. M. Newman Developmental performance of the milfoil weevil, Euhrychiopsis lecontei (Coleoptera: Curculionidae), on northern watermilfoil, Eurasian watermilfoil, and hybrid (northern X Eurasian) watermilfoil. Environmental Entomology 35:

133 123 Schoonhoven, L. M., J. J. A. van Loon and M. Dicke Insect-Plant Biology, Oxford University Press, New York. Sheldon, S. P. and R. P. Creed, Jr Use of a native insect as a biological control for an introduced weed. Ecological Applications 5: Sheldon, S. P. and R. P. Creed, Jr The effect of a native biological control agent for Eurasian watermilfoil on six North American watermilfoils. Aquatic Botany 76: Smith, G. S. and J. W. Barko Ecology of Eurasian watermilfoil. Journal of Aquatic Plant Management 28: Solarz, S. L. and R. M. Newman Oviposition specificity and behavior of the watermilfoil specialist Euhrychiopsis lecontei. Oecologia 106: Solarz, S. L. and R. M. Newman Variation in hostplant preference and performance by the milfoil weevil, Euhrychiopsis lecontei Dietz, exposed to native and exotic watermilfoils. Oecologia 126: Spanhoff, B., C. Kock, A. Meyer and E. I. Meyer Do grazing caddisfly larvae of Melampophylax mucoreus (Limnephilidae) use their antennae for olfactory food detection? Physiological Entomology 30:

134 CHAPTER VII VISION SHOULD NOT BE OVERLOOKED AS AN IMPORTANT HOST- PLANT DETECTION AND SELECTION MECHANISM FOR PHYTOPHAGOUS INSECTS Under review at Environmental Entomology at time of dissertation completion. Abstract In the last few decades, visual detection and selection of host-plants by phytophagous insects has received relatively little attention. This lack of research seems to have occurred because historically the assumption has been made that chemical cues are ultimately (if not exclusively) important for finding host-plants. In this article, recent advances in the field of phytophagous insect visual ecology are outlined, with specific arguments and evidence presented to refute many of the assumptions that are often made suggesting the unimportance of phytophagous insect vision. Insects from almost all phytophagous orders have been shown to use vision on at least some level for host-plant location, and some recent examples have even shown that vision is more important than olfaction for finding hosts. In addition, some insects even appear to have the ability to visually differentiate plant species to select appropriate hosts. Because of these and 124

135 125 other striking results from the recent literature, the visual capabilities of phytophagous insects should not be underestimated, and visual cues should be built into future hostplant location and selection studies. An extensive list of papers about visual cue use by phytophagous insects is provided for researchers interested in pursuing this topic. Introduction Before a phytophagous insect can interact with or utilize its host-plant(s), the insect clearly must first find the plant. Consequently, locating an appropriate host-plant is generally the first and most essential step in any insect-plant interaction (at least for adult insects searching out a host for feeding or oviposition). For this reason, host-plant location deserves attention, and a substantial amount work has been done in this area. Because of the general success of chemically driven host-location studies (see Bruce et al. 2005) and the thought that insects may not have very good visual acuity (Land 1997), subsequent assumptions have long been made that vision may not be important for plant location and selection by phytophagous insects. Because of this, the use of vision by phytophagous insects has been historically understudied, especially when compared to use of olfactory cues (Prokopy and Owens 1983). The purpose of this article is to outline research that has been recently performed on the use of vision by phytophagous insects to locate and select suitable host-plants, with an overarching goal of determining if vision is still being understudied (Prokopy and Owens 1983). For the purposes of this article, a host-plant is considered to be a plant that an insect is using for feeding or oviposition, largely excluding flower location by

136 126 nectar-feeding insects. Phenomena such as the use of vision by bees to locate flowers (which has an extensive literature by itself) will not be considered here. Similarly, because a fairly large literature exists examining color learning in insects (reviewed by Papaj and Prokopy 1989), examples of color learning (especially by Lepidoptera) were excluded from this article. It should be noted, however, that shape can be learned as well (Allard and Papaj 1996). In general, the literature discussed within will consider adult insects, but it should be noted that vision is also important for at least some larval phytophagous insects (e.g., Kitabatake et al. 1983; Yasui et al. 2006). Prokopy and Owens (1983) provide a comprehensive review of the use of vision by phytophagous insects to locate plants, outlining the mechanics and stimuli involved in visual plant detection (see Visser 1988 for a review on plant finding stimuli in general). This article will serve to update Prokopy and Owens (1983) with some striking recent examples of use of vision by phythohagous insects for host-plant location, while generally supporting the argument that vision may be much more important than previously thought for host-plant detection and selection. It is not the intention of this article to imply that visual cues may be more important than chemical cues for plant location by phytophagous insects, though a few recent examples of this exist and will be discussed below. Chemical cues have clearly been shown to be extremely important for locating and selecting appropriate host-plants (reviewed by Visser 1986; Bernays and Chapman 1994; Bruce et al. 2005), so there is no intent here to argue against this. This article does, however, at least begin to refute some of the many historical assumptions surrounding the unimportance of vision for phytophagous insects. Below, some of these

137 127 specific assumptions are presented and discussed, starting broadly with the overarching assumption that vision is generally not important to phytophagous insects. Assumption 1: Vision, in general, is not an important host location mechanism because insect visual acuity is poor. Vision has historically been ignored when compared to chemical cues for host location purposes (Prokopy and Owens 1983), and has been assumed to be important only for things such as locating appropriate landing sites (Bernays and Chapman 1994). More recently, support for the idea that we lack knowledge on insect vision (at least in the Chrysomelidae) has come from the literature review of Stenberg and Ericson (2007). This lack of knowledge also appears be true outside the Chrysomelidae. For example, Bruce et al. (2005) provide a review of host-plant location by insects, and only mention vision once in one sentence of their introduction, while exclusively discussing chemical cues for the rest of the article. Similarly, Bernays and Chapman (1994 pp ) cite in their book 13 papers on visual cue use by phytophagous insects (9 papers about color as a visual stimulus and 4 papers about shape and size of visual stimuli). Conversely, 29 studies are cited under odor-induced attraction. As an even more quantitative example of the lack of visual studies in recent years, a literature search performed in ISI Web of Science in early August 2010 returned 897 articles when host plant location was searched. By refining these studies by the term visual, 44 articles (4.9%) were returned. Similarly, refining the articles by vision returned even fewer articles at 16 (1.8%). Clearly, different search terms will change the corresponding results, but in any set of

138 128 search terms it can be seen that visual studies make up only a small proportion of the overall host plant location literature. It would seem from these results that vision is still generally ignored (Prokopy and Owens 1983). Because of the large number of successful studies regarding to the use of chemicals for host-plant detection (reviewed by Visser 1986; Bernays and Chapman 1994; Bruce et al. 2005), it has been historically easy to ignore visual cues in host-plant finding studies (Prokopy and Owens 1983). There has long been a general assumption that because all plants are green but have uniquely different chemistries, insects must clearly use chemoreception predominately (if not wholly) for plant location (Schoonhoven et al. 2005; p. 144). Even with the relatively small body of literature regarding visual detection of plants, examples from almost all orders of phytophagous insects have been found that use visual cues on at least some level for host-plant or trap detection (Appendix 2.A; the lack of plant or trap finding examples from the Phasmatodea is likely a lack of research not lack of insect ability). In spite of the ubiquity of visual cue use amongst phytophagous insects, we have tended to ignore vision, assuming that it is less important than chemical cues. However, this historical assumption has no strong support beyond the general success of chemically driven studies. There seems to have been no strong, empirical studies conclusively showing that vision (at least in general across insects) is not or cannot be important for host-plant location. This assumption that vision is not important seems to be largely based on rough anatomical/physiological calculations comparing insect eyes to vertebrate eyes that have

139 129 made wild claims such as insect eyes needing to be 19m in radius to match human visual acuity (see Land 1997 for a review of insect visual acuity with historical perspectives). While some phytophagous insects use vision simply for things like swimming away from light in order to help find the roots of submerged host-plants (Serandour et al. 2006), other insects can visually differentiate plant species and select the appropriate host in the absence of any chemical cues (Reeves and Lorch 2009). In addition, one butterfly species can visually recognize conspecific egg clusters and avoid ovipositing on a plant that is already bearing eggs (Vasconcellos-Neto and Monteiro 1993), and at least one moth can detect and avoid subtle variegation because it mimics damage by leaf-miners (Soltau et al. 2009). Similarly, Pieris rapae L. (Lepidoptera: Pieridae) may not be affected by plant diversity when searching for host-plants because of high visual acuity (Broad et al. 2008). These examples (which are described in more detail below) show that insect visual acuity may have been underestimated, or at least that insects can potentially do more with the acuity they can create/perceive than what has been assumed. As noted above, one clear reason for the prevalence of the assumption that vision is not important to phytophagous insects is that they have often been shown to become attracted to chemicals from their host-plants in the absence of any visual cues. In these types of studies (e.g., Marko et al. 2005; Addesso and McAuslane 2009; many others reviewed in Bruce et al. 2005), because positive results were seen in the absence of visual cues, it was subsequently assumed that chemical cues are of ultimate importance for hostplant location. Without explicitly examining visual cues, however, it cannot be assumed that vision is not important, nor can it be assumed that chemical cues are more important

140 130 than visual cues for host-plant location. Thus, successful chemosensory work should not preclude any given insect from subsequent testing with visual cues. As a case in point, the aquatic milfoil weevil, Euhrychiopsis lecontei Dietz (Coleoptera: Curculionidae), was shown to be attracted to glycerol and uracil, both of which are exudates of not only their host-plants, but aquatic plants in general (Marko et al. 2005). Vision was not mentioned at all in this paper, and the chemoreceptive abilities of the weevils were discussed as the plant finding mechanism of the weevil. However, later work showed not only that visual cues were important for the weevils for locating plants in the complete absence of chemical cues (Reeves et al. 2009), but also that the weevils can differentiate between their host plant and at least one non-host species using vision alone (Reeves and Lorch 2009). This example suggests that visual cues should generally be examined in studies of host-plant locating behavior. Bell (1990) declared in his review of searching behavior in insects that any study which only incorporates a single search modality underestimates the amount of information an insect has at its disposal, which supports the idea of including visual cues as well. Especially in cases such as Heisswolf et al. (2007b) where it was unclear exactly which stimuli (visual or chemical) attracted Cassida canaliculata Laich (Coleoptera: Chrysomelidae) to its host plant, vision should be explicitly studied. In subsequent work on the same system (Heisswolf et al. 2007a), however, the authors chose not to examine vision, assuming it must not be important for the beetle in question. This example reinforces the point that vision still is often ignored without justification.

141 131 Appendix 7.A shows experimental evidence that vision may be more important for phytophagous insects than has been historically assumed by demonstrating that examples from almost all phytophagous orders use vision on at least some level. Another compelling kind of evidence is that plants have found many ways to use visual stimuli for their own benefit to either exploit or deter insects. First, variegation in some plants has recently been hypothesized to mimic damage by herbivory, thus driving away future herbivores that happen upon the apparently damaged plants (Soltau et al. 2009). Next, besides deterring insects, plants have also been able to hide from them by manipulating visual stimuli. For instance, sand-dune plants that have a white coloration because of sand particles sticking to glandular trichomes, are thought to have evolved this ability in order to make visual detection by phytophagous insects more difficult. These plants may even use sand to mimic deterrent fungal infections (Lev-Yadun 2006). Similarly, Karageorgou and Manetas (2006) hypothesized that the red coloration in young oak leaves may (among other things) make the leaves less discernable from the background. In contrast to the evolution of mechanisms to reduce detectability, pitcher plants can even attract insects visually. Schaefer and Ruxton (2008) found that red pitcher plants caught more insects (Diptera, Homoptera, Hymenoptera) than green pitcher plants, though the exact reason why the red pitcher plants were more successful was not entirely clear. The result is interesting nonetheless. Finally, another way plants have been hypothesized to utilize the visual capabilities of phytophagous insects for their own benefit is the warning off of fall-colonizing aphids by trees through the vivid displays of fall coloration (reds, oranges). Hamilton and Brown (2001) hypothesized that trees which are heavily

142 132 chemically defended invest in carotenoids that are subsequently used to warn aphids of the defense, though how the colors are detected and used by aphids is still debated (Doring et al. 2009). Clearly, if visual detection of plants was not important for phytophagous insects, plants would not have developed these means of exploiting the visual capabilities of their herbivores. Assumption 2: Vision is only used when appropriate chemical cues are detected. Bernays and Chapman (1994) note in their review on host-plant selection that response to visual stimuli often only occurs after detection of appropriate chemical cues, or potentially when chemical cues are lost during final plant approach. While this is true in some cases (e.g., Blackmer and Canas 2005; Jonsson et al. 2007), it is not always true, so this is another assumption that should perhaps not always be made. Some insects do, in fact, show the opposite behavior, where chemical cues may function to enhance visual detection of host-plants. For example, Patt and Setamou (2007) found that Homalodisca coagulata Say (Homoptera: Cicadellidae) responds to chemical cues only to enhance their response to visual cues. Similarly, Aluja and Prokopy (1993) hypothesized that Rhagoletis pomonella Walsh (Diptera: Tephritidae) uses chemical cues for host-plant location only if visual stimuli are insufficient. Next, Hessian flies (Diptera: Cecidomyiidae) appear to use visual cues over chemical cues for host-plant location (Withers and Harris 1996). Finally, Diaphorina citri Kuwayama (Hemiptera: Psyllidae) responds to chemical cues only in the presence of visual cues under some circumstances (Wenninger et al. 2009). It seems unlikely that these four examples (which span multiple

143 133 orders and families) would be the only examples of this phenomenon, so it should not be assumed a priori that visual responses only occur in the presence of appropriate chemical cues for any given insect. A more specific hypothesis related to the one proposed by Bernays and Chapman (1994) is the appropriate/inappropriate landings hypothesis (Finch and Collier 2000). According to this hypothesis, insect pests of cruciferous crops are stimulated to land haphazardly on green objects only after detection of appropriate plant volatile chemicals. After landing, contact chemoreception is then used to accept or reject the plant. The authors do a good job of discussing the importance of visual cues in this hypothesis, but an assumption still exists that vision only becomes important after chemical detection, and that the insects land haphazardly on any green object. It may be possible that the insects can perceive more than just the green color (i.e., shape, size, etc.) of the potential hosts, and that visual cues may also cause the insects to land (even without chemical cues), so there are still other testable ideas for cruciferous vegetable pests (and phytophagous insects in general). Beyond the evidence that visual cues are used in conjunction with or after detection of chemical cues, there is some evidence that insects may use vision exclusively for host-plant location. One such striking example is that of Altica engstroemi J. Sahlberg (Coleoptera: Chrysomelidae: Alticinae). This Chrysomelid was never attracted to its hostplants chemical signatures in laboratory olfactometer testing, but was highly visually attracted to its host-plant, even when presented with dummy plants at the same time (Stenberg and Ericson 2007). Stenberg and Ericson (2007) proposed that insects which

144 134 live in persistent habitats and utilize host-plants that dominate species-poor plant communities may commonly evolve vision as an important (or even exclusive) host-plant location mechanism. This is a testable hypothesis worth exploring in plant-insect systems with similar characteristics. Another example of a phytophagous insect locating its hosts using vision alone is that of Macrosiphoniella artemisiae Boyer de Fonscolombe (Homoptera: Aphididae). This aphid was not attracted to host-plant volatiles and could discriminate between host and non-host targets solely using vision (Gish and Inbar 2006). In fact, it has historically been thought that aphids in general use vision predominantly to find host-plants (Doring and Chittka 2007), though evidence does exist that at least some aphids are capable of detecting host-plant odors (e.g., Hori 1999). Similar to the above examples, Empoasca fabae (Homoptera: Cicadellidae) was attracted by visual and not chemical cues (Bullas- Appleton et al. 2004). Clearly, if some insects exclusively use vision, the importance of vision should not be ignored for any system. The only way to gain a better understanding of the use of vision for host-plant location is to explicitly include visual cues as an integral part of any host-plant finding study. Assumption 3: Insects cannot visually differentiate plant species Though this assumption may fit within Assumption 1 above, it is worth mentioning separately because it is perhaps the strongest use of host-plant visual cues by phytophagous insects to visually differentiate between plant species and select the appropriate host. While visual plant differentiation may seem unlikely to those assuming

145 135 that vision is unimportant to insects, this phenomenon has been documented at least three times. First, Prokopy et al. (1983a) were able to show that the cabbage root fly, Delia radicum L. (Diptera: Anthomyiidae), can visually distinguish between host-plant cultivars using leaf color alone. Similarly, Hyalopterus pruni (Homoptera: Aphididae) may also be able to differentiate plants based on color alone (Moericke 1969). More recently, Reeves and Lorch (2009) showed that E. lecontei can visually discriminate between a host-plant and at least one non-host species in the complete absence of chemical cues. In the case of E. lecontei, however, plant form may have been more important than color, showing that color may not always be the only or most important plant visual cue. It is possible that the small number of examples of insects visually differentiating plant species/cultivars may be based on a research bias whereby it was assumed that insects cannot visually differentiate plants, so research was not attempted. These three examples seem unlikely to be the only three insects with this ability, so more visual plant differentiation studies are certainly warranted in the future. Assumption 4: Color is the only important visual stimulus for phytophagous insects. Schoonhoven et al. (2005 p.144) note that insect vision is often ignored because all plants are green but have uniquely different chemistries. This thought may not only have led to relatively little study of phytophagous insect vision in general, but may also have lead to an assumption that green (color) is the only important visual stimulus for insects, as in the appropriate/inappropriate landings hypothesis (Finch and Collier

146 ) discussed above. Clearly, the thought that color is important to phytophagous insects is well supported even by recent research (e.g., Prokopy et al. 1983a; Kostal 1991; Hausmann et al. 2004; Egusa et al. 2006). At least two phyphophagous insects have even been shown to have tri-chromatic visual systems (Anthonomus pomorum L. Coleoptera: Curculionidae; Hausmann et al and Myzus persicae Sulz. (Hemiptera: Aphididae; Kircnner et al. 2005), so color is undoubtedly very important for phytophagous insects. Nocturnal lepidopterans may even be able to see color at night (Kelber et al. 2002). The use of color as a visual stimulus for phytophagous insects is more comprehensively reviewed in Prokopy and Owens (1983), and the evolution of color vision in insects in general is reviewed by Briscoe and Chittka (2001). Even though color is clearly important to phytophagous insects, it may not be the ultimate determinant of plant finding in all cases (Brennan and Weinbaum 2001). For example, shape/form may be more important than color in E. lecontei (Reeves and Lorch 2009). Leaf shape has been shown to be important for other insects as well (Harris and Miller 1984; Hodgson and Elbakhiet 1985; Mackay and Jones 1989). It has even recently been thought that leaf shape in general may also affect insect feeding behavior in some cases (Rivero-Lynch et al. 1996). Next, size/area of visual stimuli has been shown to be important (Prokopy et al. 1983b; Ahman et al.1985; Carrizo 2008). In addition, spatial (e.g., vertical vs. horizontal) orientation of visual targets has been shown to be important as well (Harris et al. 1993). Finally, location (height) of visual traps has also been shown to be important (Gillespiei and Vernonz 1990; Atakan and Canhilal 2004; Blackmer et al. 2008).

147 137 With such a variety of visual stimuli having been shown to be important to various insects, color should not be exclusively used in future host-location studies. In fact, multiple visual stimuli should be tested, if possible, and assumptions a priori as to which stimuli will be important for a given phytophagous insect should be avoided. Especially because attractive visual stimuli are not always mutually exclusive and may even be synergistic (Prokopy 1968), the assumption that other visual stimuli are unimportant if one has been found to be important should be avoided as well. Similarly, stimulus height, shape, and area were all shown to be important in Tuttle et al. (1988), so more than one visual stimulus could be explored during host-plant location studies, even if previous work on a single stimulus has been effective. While the list of studies presented in this section is clearly not an exhaustive list of all studies for each type of visual stimulus, it does show that different visual stimuli can be attractive to different insects, so this idea may be useful to consider when designing future studies. The successful use of color may have created a bias towards only using color as a visual stimulus, so more studies using stimuli beyond color are certainly warranted. As part of using many different types of stimuli for visual experiments, realistic visual stimuli (i.e., plants themselves) could be used more than they have been in the past. Very often, highly artificial visual stimuli are used in host-location studies. For instance, colored (painted) polystyrene spheres were used in (Drew et al. 2006). Similarly, colored paper has been used as a plant surrogate (Roessingh and Stadler 1990; Hirota and Kato 2001; Jonsson et al.2007; Reddy et al. 2009). Next, colored transparent

148 138 plastic sheets have even been used (Scherer and Kolb 1987a,b; Sharma and Franzmann 2001). Also, and perhaps even less realistically, lights have even been used as plant visual cues (Todd et al. 1990; Blackmer and Canas 2005; Wenninger et al. 2009). If the goal of a study is determine how an insect finds host-plants, it may sometimes be more productive to use actual host-plants as the visual stimulus, or at least shape the dummy stimulus to look reasonably similar to the host-plant in question. This is especially true given that color may not always be the only or even most important visual stimulus for phytophagous insects (see above). Using real plants for visual cues in host location studies has been done successfully in the past (Szentesi et al. 1986; Aluja and Prokopy 1993; Prokopy et al. 1994; Couty et al. 2006; Stenberg and Ericson 2007; Reeves et al. 2009; Reeves and Lorch 2009) so this approach is worth considering. Negative results in studies with colored plastic, paper, or lights may have missed some critical aspect of the plant that either alone or together would have attracted the phytophagous insect, leading to premature conclusions that the insect in question is not attracted by visual cues. If real plants are used it may be more likely that the insect will perceive the visual cues that elucidate responses, and a more reliable determination that the insect uses visual cues can be attained (though the plant would need to be manipulated to determine which specific visual cues were important). Clearly, plant chemical cues will need to be removed or blocked in studies using real plants for visual cues, but methods as simple as sealing plants in containers worked in Stenberg and Ericson (2007) and Reeves et al. (2009).

149 139 Though past studies using artificial stimuli have been successful, it is worth attempting to move beyond lights and colored paper or plastic as surrogates for plant visual stimuli, especially given the striking examples of insect visual acuity noted above. This idea may even extend to trapping. It seems few traps have been developed that resemble host-plants themselves, though cost of production is likely part of the reason for this. Interestingly, however, Rossingh and Stadler (1990) showed that visual stimuli with stems attracted more flies for oviposition than artificial leaves without stems, so something as simple as adding a stem to artificial stimuli may make traps (or host-plant surrogates) more realistic. Another potentially interesting research area would to be to test realistic visual stimuli (i.e., real plant material or dummy stimulus shaped like a plant) and artificial stimuli in choice tests to see if using more realistic stimuli elicits a stronger behavioral response than artificial stimuli. Finally, unrealistic conditions in laboratory arenas could also impact insect behavior. Field studies, though very difficult to perform, have been successfully performed in the past (e.g., Aluja and Prokopy 1993, Butkewich and Prokopy 1997) so it may be interesting to compare insect response to both realistic and artificial visual stimuli in both the lab and the field. Underexplored Areas of Study for Phytophagous Insect Visual Ecology New research into the importance of vision for phytophagous insects could proceed down novel and potentially fruitful research avenues. One potentially interesting area of study is phytophagous insect eye anatomy/physiology. Two of the many factors influencing visual acuity in insects are number of ommatidia and inter-ommatidial angles,

150 140 with higher numbers of ommatidia and smaller angles between them giving the best acuity (Land 1997). These (and other) ommatidia metrics could be measured and compared for insects with and without visual preferences for plant location. Prokopy and Owens (1978) make the related suggestion that specialist phytophagous insects may be more visually acute than generalists for host-plant location. This later idea has been directly supported in aphids (Vargas et al. 2005), and less directly supported in specialist beetles that were very visually oriented towards their host-plant (Stenberg and Ericson 2007; Reeves and Lorch 2009). An interesting hypothesis to test, then, would be that specialist phytophagous insects have higher numbers of ommatidia and smaller interommatidial angles relative to generalist counterparts, or at least that insects that react to visual stimuli may have more ommatidia than insects that may rely exclusively on olfaction. Quantifying neuronal responses to visual stimuli may prove interesting as well. Neuronal response detection is a common practice for determining chemosensory abilities in phytophaogus insects (see Visser 1986), but there seems to be much less work done on visual stimuli. In locusts, however, work has been done involving neuronal responses to visual (light) stimuli (Homburg and Wurden 1997; Vitzthum et al. 2002; Field et al. 2008). Similar work has also been done on an aphid (Kirchner et al. 2005), and also on the electrophysiological response to color in fruit flies (Agee 1985). Since vision is clearly an important mechanism for host-plant location for many insects, studies on neuronal responses to visual stimuli may prove interesting and important.

151 141 Another common practice for chemosensory studies of phytophagous insects is identifying deterrent or repellent chemicals using olfactometers or electroantennograms (Visser 1986; Bruce et al. 2005). The overall reaction of a phytophagous insect to a potential host-plant is the sum of effects from attractive chemicals and deterrent or repelling chemicals (Bernays and Chapman 1994). Interestingly, this same idea has been shown to be true with visual (light) cues in Delia antiqua Meigen (Diptera: Anthomyiidae). This fly responded to visual stimuli based on the ratio of UV light (deterrent) and various visible wavelengths of light (attractive; Judd et al. 1988), in a manner similar to how an insect would react to various ratios of attractive and deterrent chemical stimuli (Bernays and Chapman 1994). With an understanding of what is visually deterrent to phytophagous insects may come the ability to cultivate plants that are better protected from pest insects, or deterrent visual stimuli can even be designed and placed in agricultural fields (Prokopy and Owens 1983; Foster and Harris 1997). Studies of deterrent visual stimuli are certainly warranted in other plant-insect systems to better understand this relatively understudied topic. A further underexplored area of study is the idea of visual active spaces. That is, the area around a plant in which the visual stimuli associated with the plant are sufficiently strong to elicit a response (Schoonhoven et al p.144). Essentially, more work needs to be done to determine how far phytophagous insects can see. Schoonhoven et al. (2005; pp ) discuss the idea of active spaces for sufficiently strong visual and chemical cues, and give some examples of each in their Table 6.2. They note that a general assumption exists that the active space around a plant is larger for chemical cues

152 142 than visual cues (i.e., insects can detect plants from farther away using chemicals). However, this assumption is called premature by Schoonhoven et al. (2005) because of a lack of evidence, so research in this area may prove important. Directional information from visual cues is more precise than from chemical cues (Bell 1990) because the former does not depend on diffusion or air/water currents, so it is critically important to determine the distance from which an insect is attracted to its host-plant on a visual basis, especially if this distance can be compared to the active space associated with the hostplant s chemical cues. Beyond elucidating how far phytophagous insects can see and comparing those distances to chemosensory abilities, more work should be done to determine the relative strengths of chemical vs. visual stimuli. Choice-tests could be performed more often with visual and chemical stimuli present at the same time to quantify which has the biggest draw for the insect. A study of this type has been performed at least once with results showing that the diurnal hawkmoth, Macroglossum stellatarum (Lepidoptera: Sphingidae) is more attracted to visual than chemical cues when presented with both at the same time (Balkenius et al. 2006). These types of studies will likely turn up other instances of visual cues being more attractive than chemical cues, especially since insects may switch from chemical to visual cues when good visual cues become available (Bell 1990). One general area where researchers have tended not to ignore visual cues as being important is trapping. Trapping of phytophagous insects is discussed in some detail in Prokopy et al. (1983). Trapping appears to be the driving force in many visual studies

153 143 concerning phytophagous insects (Appendix 7.A). In general, trap color and placement (e.g., Hoback et al. 1999) and trap type (e.g., Bloem et al. 2002) can influence trapping efficacy, so understanding visual stimuli is of critical importance in trapping applications. Here then is an interesting phenomenon: very often in trapping studies, we assume vision is important and try various colors, shapes, sizes, heights, etc. of traps to find the most effective designs. For some reason, however, when we decide to do a host-plant location study, vision is often assumed not to be important and chemosensory studies are employed. While vision in the context of trapping has proved fruitful, it should not be the only case in which visual stimuli are taken into account. If visual stimuli have so clearly been shown to attract insects to traps, it seems likely that the insects might use vision to find host-plants as well. Obvious candidates for visual studies of plant location would be pest insects that have been shown to be visually attracted to traps (e.g., Coli et al. 1985; Singh and Saxena 2004; Anthanassiou et al. 2004; Hogmire and Leskey 2006; Blackmer et al. 2008; Van den Berg et al. 2008). It would also be useful to determine the relative strength of trap visual cues vs. plant visual cues. Often, trapping studies make suggestions of which types of traps to use for any given insect. However, just because a trap was most attractive out of the group of traps used, that does not mean the trap will be more attractive than host plants. Blumthal et al. (2005) came close to doing something like this, as they compared attractive flower color to the color of previously demonstrated sticky traps, finding they were similar. However, the flowers and traps were not placed side-by-side to let the insect make a choice between stimuli, which may have been potentially interesting. This type of choice

154 144 study has been done at least once, however, with interesting results. Natwick et al. (2007) showed that two thrips species are more attracted to blue traps than yellow traps. They set up traps within plots of plants, and found more thrips on the traps than on the plants at the end of their study. It seemed, then, that the traps were an attractive enough stimulus to lure the thrips off of their host-plants. Clearly, more of these types of studies are warranted, especially if they show how well trapping may (or may not) work and because they may aid in surveying or reducing numbers of pest insects. Conclusions Because the use of vision by phytophagous insects appears ubiquitous among the phytophagous orders, and because many striking examples have come out in the recent literature regarding the visual abilities of phytophagous insects, vision could be examined much more often than it is currently. Whenever a host-plant location study is performed, visual cues should be explicitly examined. Ungrounded assumptions as to which visual cues to use should be avoided, and multiple cues (i.e., many colors, sizes, forms, placements) should be used if possible. Also, realistic shapes, sizes, orientations, and colors of cues could be used more often, as these types of cues could potentially be more indicative of insect response than colored lights, paper, etc. The anatomy and physiology of visual detection and selection of host-plants should be further explored as well. Overall, working these elements into studies of phytophagous insects will likely show that vision is much more important than previously thought. Because chemicals and vision can often work synergistically (e.g., Bjorklund et al. 2005; Pinero et al. 2006) for

155 145 host location, vision should perhaps always be explored, even in cases where chemical cues may have previously been shown to be important. Examining a single search modality in any given study may be a mistake because it can underestimate the amount of information an insect has at its disposal (Bell 1990), so both conjecture and empirical evidence support this call for an increase in research on the use of vision by phytophagous insects. Hopefully, this article will help spur an increase in this research. The current and historic research bias towards exclusively using chemical cues may just be a human research bias and is almost certainly not a universal sensory bias by phytophagous insects. Therefore, explicitly examining visual detection and selection of host-plants by phytophagous insects may be the only way to ever truly understand the whole host-finding process. The examples noted throughout this article, though much less numerous than chemical cue papers, show that the call for more visual studies in Prokopy and Owens (1983) has not been answered, and gives strong justification for an increase in research on phytophagous insect visual ecology. Acknowledgments Pat Lorch is thanked for his comments on this manuscript. References Addesso, K. M. and H. J. McAuslane Pepper weevil attraction to volatiles from host and nonhost plants. Environmental Entomology 38:

156 146 Agee, H. R Spectral response of the compound eye of the wild and laboratory reared apple maggot fly, Rhagoletis pomonella. Journal of Agricultural Entomology 2: Ahman, I., J. Weibull and J. Pettersson The role of plant size and plant density for host finding in Rhopalosiphum padi (L.) (Hem.: Aphididae). Sweedish Journal of Agricultural Research 15: Allard, R. A. and D. R. Papaj Learning of leaf shape by pipevine swallow tail butterflies: a test using artificial models. Journal of Insect Behavior 9: Aluja, M. and R. J. Prokopy Host odor and visual stimulus interaction during intratree host finding behavior of Rhagoletis pomonella flies. Journal of Chemical Ecology 19: Athanassiou, C. G., N. G. Kavallieratos and B. E. Mazomenos Effect of trap type, trap color, trapping location, and pheromone dispenser on captures of male Palpita unionalis (Lepidoptera: Pyralidae). Journal of Economic Entomology 97: Atakan, E. and R. Canhilal Evaluation of yellow sticky traps at various heights for monitoring cotton insect pests. Journal of Agricultural and Urban Entomology 21: Balkenius, A., W. Rosen and A. Kelber The relative importance of olfaction and vision in a diurnal and a nocturnal hawkmoth. Journal of Comparative Physiology A 192:

157 147 Bell, W. J Searching behavior patterns in insects. Annual Review of Entomology 35: Bernays, E. A. and R. F. Chapman Host Plant Selection by Phytophagous Insects, Chapman and Hall, New York. Bjorklund, N., G. Norlander. and H. Bylund Olfactory and visual stimuli used in orientation for conifer seedlings by the pine weevil, Hylobius abietis. Physiological Entomology 30: Blackmer, J. L. and L. A. Canas Visual cues enhance the response of Lygus Hesperus (Heteroptera: Miridae) to volatiles from host plants. Environmental Emtomology 34: Blackmer, J. L., J. A. Byers and C. Rodriguez-Saona Evaluation of color traps for monitoring Lygus spp.: Design, placement, height, time of day, and non-target effects. Crop Protection 27: Bloem, S., R. F. Mizell and C. W. O Brien Old traps for new weevils: new records for curculionids (Coleoptera: Curculionidae), brentids (Coleoptera: Brentidae), and anthribids (Coleoptera: Anthribidae) from Jefferson Co., Florida. Florida Entomologist 85: Blumthal, M. R., R. A. Cloyd, L. A. Spomer and D. F. Warnock Flower color preferences of western flower thrips. Horticultural Technology 15: Brennan, E. B. and S. A. Weinbaum Psyllid responses to colored sticky traps and the colors of juvenile and adult leaves of the heteroblastic host plant Eucalyptus globulus. Environmental Entomology 30:

158 148 Briscoe, A. D. and L. Chittka The evolution of color vision in insects. Annual Review of Entomology 46: Broad, S. T., N. A. Schellhorn, S. N. Lisson and N. J. Mendham Host location and oviposition of lepidopteran herbivores in diversified broccoli cropping systems. Agricultural and Forest Entomology 10: Bruce, T. A. J., L. J. Wadhams and C. M. Woodcock Insect host location: a volatile situation. Trends in Plant Science 10: Bullas-Appleton, E. S., G. Otis, C. Gillard and A. W. Schaafsma Potato leafhopper (Homoptera: Cicadellidae) varietal preferences in edible beans in relation to visual and olfactory cues. Environmental Entomology 33: Butkewich, S. L. and R. J. Prokopy Attraction of adult plum curculios (Coleoptera: Curculionidae) to host-tree odor and visual stimuli in the field. Journal of Entomological Science 32:1-6. Campbell, S. A. and J. H. Borden Bark reflectance spectra of conifers and angiosperms: implications for host discrimination by coniferophagous bark and timber beetles. Canadian Entomologist 137: Carrizo, P Effects of yellow trap size on sampling efficiency for flower thrips (Frankliniella occidentalis) from pepper plants (Capsicum annum) grown in greenhouses. Ciencia e Investigacion Agraria 35: Coli, W. M., T. A. Green, T. A. Hosmer and R. J. Prokopy Use of visual traps for monitoring insect pests in the Massachusetts apple IPM program. Agriculture, Ecosystems & Environment 14:

159 149 Couty, A., H. Van Emden, J. N. Perry, J. Hardie, J. A. Pickett and L. J. Wadhams The roles of olfaction and vision in host-plant finding by the diamondback moth, Plutella xylostella. Physiological Entomology 31: Doring, T. F., M. Archetti and J. Hardie Autumn leaves seen through herbivore eyes. Proceedings of the Royal Society B 276: Doring, T. F. and L. Chittka Visual ecology of aphids: a critical review on the role of colours in host finding. Arthropod-Plant Interactions 1:3-16. Drew, R. A. I., C. Dorji, M. C. Romig and P. Loday Attactiveness of various combinations of colors and shapes to females and males of Bactrocera minax (Diptera: Tephritidae) in a commercial mandarin grove in Bhutan. Journal of Economic Entomology 99: Egusa, S., T. Nishida, K. Fujisaki and H. Sawada Spatio-temporal abundance of flushing leaves shapes host selection in the willow leaf beetle, Plagiodera versicolora. Entomologia Expermentalis et Applicata 120: Field, L. H., C. Duch and H-J Pfluger Responses of efferent octopaminergic thoracic unpaired median neurons in the locust to visual and mechanosensory signals. Journal of Insect Physiology 54: Finch, S. and R. H. Collier Host plant selection by insects: a theory based on appropriate/inappropriate landings by pest insects of cruciferous plants. Entomologia Experimentalis et Applicata 96: Foster, S. P. and M. O. Harris Behavioral manipulation methods for insect pest management. Annual Review of Entomology 42:

160 150 Gillespiei, D. R. and R. S. Vernonz Trap catch of western flower thrips (Thysanoptera: Thripidae) as affected by color and height of sticky traps in mature greenhouse cucumber crops. Journal of Economic Entomology 83: Gish, M. and M. Inbar Host location by apterous aphids after escape dropping from the plant. Journal of Insect Behavior 19: Hamilton, W. D. and S. P. Brown Autumn tree colors as a handicap signal. Proceedings of the Royal Society of London B 268: Harris, M. O. and J. R. Miller Foliar form influences ovipositional behavior of the onion fly. Physiological Entomology 9: Harris, M. O., S. Rose and S. Malsch The role of vision in the host plant-finding behaviour of the Hessian fly. Physiological Entomology 18: Hausmann, C., J. Samietz and S. Dorn Visual orientation of overwintered Anthonomus pomorum (Coleoptera: Curculionidae). Environmental Entomology 33: Heisswolf, A., D. Gabler, E. Obermaier and C. Muller. 2007a. Olfactory versus contact cues in host plant recognition of a monophagous chrysomelid beetle. Journal of Insect Behavior 20: Heisswolf, A., S. Ulmann, E. Obermaier, O. Mitesser and H. J. Poethke. 2007b. Host finding in the specialised leaf beetle, Cassida canaliculata: an analysis of smallscale movement behaviour. Ecological Entomology 32:

161 151 Hirota, T. and Y. Kato Influence of visual stimuli on host location in the butterfly, Eurema hecabe. Entomologia Experimental et Applicata 101: Hoback, W. W., T. M. Svatos, S. M. Spomer and L. G. Higley Trap color and placement affects estimates of insect family-level abundance and diversity in a Nebraska salt marsh. Entomologia Experimentalis et Applicata 91: Hodgson, C. J. and I. B. Elbakhiet Effect of colour and shape of target hosts on the orientation of emigrating adult apterous Myzus persicae in the laboratory. Entomologia Experimentalis et Applicata 38: Hogmire, H. W. and T. C. Leskey An improved trap for monitoring stink bugs (Heteroptera: Pentatomidae) in apple and peach orchards. Journal of Entomolgical Science 41:9-21. Homberg, U. and S. Wurden Movement-sensitive, polarization-sensative, and light-sensative neurons of the medulla and accessory medulla of the locust, Schistocerca gregaria. Journal of Comparative Neurology 386: Hori, M Role of host plant odors in the host finding behaviors of aphids. Applied Entomology and Zoology 34: Jonsson, M., K. Rosdahl and P. Anderson Responses to olfactory and visual cues by the over-wintered and summer generations of the pollen beetle, Meligethes aeneus. Physiological Entomology 32: Judd, G. J. R., J. H., Borden and A. D. Wynne Visual behavior of the onion fly, Delia antiqua: antagonistic interaction of ultraviolet and visible wavelength reflectance. Entomologia Experimentalis et Applicata 49:

162 152 Karageorgou, P. and Y. Manetas The importance of being red when young: anthocyanins and the protection of young leaves of Quercus coccifera from insect herbivory and excess light. Tree Physiology 26: Kelber, A., A. Balkenius and E.J. Warrant Scotopic colour vision in nocturnal hawkmoths. Nature. 419: Kirchner, S. M., T. F. Döring and H. Saucke Evidence for trichromacy in the green peach aphid, Myzus persicae (Sulz.) (Hemiptera: Aphididae). Journal of Insect Physiology 51: Kitabatake, S., I. Shimizu and M. Kato Wavelength-dependent properties of phototaxis in larvae of Bombyx mori. Photochemistry and Photobiology 37: Koštál, V The effect of colour of the substrate on the landing and oviposition behaviour of the cabbage root fly. Entomologia Experimentalis et Applicata 59: Land, M. F Visual acuity in insects. Annual Review of Entomology 42: Lapis, E. B. and J. H. Borden Role of wavelength-specific reflectance intensity in host selection by Heteropsylla cubana Crawford (Homoptera: Psyllidae). Pan- Pacific Entomologist 71: Lev-Yadun, S Defensive functions of white coloration in coastal and dune plants. Israel Journal of Plant Sciences 54: Mackay, D. A. and R. E. Jones Leaf shape and the host-finding behaviour of two ovipositing monophagous butterfly species. Ecological Entomology 14:

163 153 Marko, M. D., R. M. Newman and F. K. Gleason Chemically mediated host-plant selection by the milfoil weevil: a freshwater insect-plant interaction. Journal of Chemical Ecology 31: Moericke, V Hostplant specific colour behavior by Hyalopterus pruni (Aphididae). Entomologia Experimentalis et Applicata 12: Natwick, E. T., J. A. Byers, C-C. Chu, M. Lopez and T. J. Henneberry Early detection and mass trapping of Frankliniella occidentalis and Thrips tabaci in vegetable crops. Southwestern Entomologist 32: Patt, J. M. and M. Sétamou Olfactory and visual stimuli affecting host plant detection in Homalodisca coagulata (Hemiptera: Cicadellidae). Environmental Entomology 36: Papaj, D. and R. J. Prokopy Ecological and evolutionary aspects of learning in phytophagous insects. Annual Review of Entomology 34: Pinero, J. C., I. Jacome, R. Vargas and R. J. Prokopy Response of female melon fly, Bactrocera cucurbitae, to host-associated visual and olfactory stimuli. Entomologia Experimentalis et Applicata 121: Prokopy, R. J Visual response of apple maggot flies, Rhagolettis pomonella (Diptera:Tephritidae):orchard studies. Entomologia Experimentalis et Applicata 11: Prokopy, R. J., C. Bergweiler, L. Galarza and J. Schwerin Prior experience affects the visual ability of Rhagoletis pomonella flies (Diptera: Tephritidae) to find host fruit. Journal of Insect Behavior 7:

164 154 Prokopy, R. J., R. H. Collier and S. Finch. 1993a. Leaf color used by cabbage root flies to distinguish among host plants. Science 221: Prokopy, R. J., R. H. Collier and S. Finch. 1993b. Visual detection of host plants by cabbage root flies. Entomologia Experimentalie et Applicata 34: Prokopy, R. J. and E. D. Owens Visual generalist with visual specialist phytophagous insects: host selection behavior and application to management. Entomologia Experimentalis et Applicata 24: Prokopy, R. J. and E. D. Owens Visual detection of plants by herbivorous insects. Annual Review Entomology 28: Reddy, G. V. P., Z. T. Cruz, N. Bragana and R. Muniappan Response of Melitta oedipus (Lepidoptera: Sesiidae) to visual cues is increased by the presence of food source. Journal of Economic Entomology 102: Reeves, J. L. and P. D. Lorch Visual plant differentiation by the milfoil weevil, Eurychiopsis lecontei Dietz (Coleoptera: Curculionidae). Journal of Insect Behavior 22: Reeves, J. L., P. D. Lorch and M. W. Kershner Vision is important for plant location by the phytophagous aquatic specialist Euhrychiopsis lecontei Dietz (Coleoptera: Curculionidae). Journal of Insect Behavior 22: Rivero-Lynch, A. P., V. K. Brown and J. H. Lawton The impact of leaf shape on the feeding preference of insect herbivores: experimental and field studies with Capsella and Phyllotreta. Phil. Transactions of the Royal Society of London B 351:

165 155 Roessingh, P. and E. Städler Foliar form, colour and surface characteristics influence oviposition behaviour in the cabbage root fly Delia radicum. Entomologia Experimentalis et Applicata 57: Sérandour, J., D. Rey and M. Raveton Behavioural adaptation of Coquillettidia (Coquillettidia) richiardii larvae to underwater life: environmental cues governing plant-insect interaction. Entomologia Experimentalis et Applicata 120: Schaefer, H. M. and G. D. Ruxton Fatal attraction: carnivorous plants roll out the red carpet to lure insects. Biology Letters 4: Scherer, C. and G. Kolb. 1987a. Behavioral experiments on the visual processing of color stimuli in Pieris brassicae L. (Lepidoptera). Journal of Comparative Physiology A 160: Scherer, C. and G. Kolb. 1987b. The influence of color stimuli on visually controlled behavior in Aglais urticae L. and Pararge aegeria L. (Lepidoptera). Journal of Comparative Physiology A 161: Schoonhoven, L.M., J. J. A. van Loon and M. Dicke Insect-Plant Biology, Oxford University Press, New York. Sharma, H. C. and B.A. Franzmann Orientation of sorghum midge, Stenodiplosis sorghicola, females (Diptera: Cecidomyiidae) to color and host-odor stimuli. J. Agricultural and Urban Entomology 18: Singh, A. K. and K. N. Saxena Attraction of larvae of the armyworm Spodoptera litura (Lepidoptera: Noctuidae) to coloured surfaces. European Journal of Entomology 101:

166 156 Soltau, U., S. Dotterl and S. Liede-Schumann Leaf variegation in Caladium steudneriifolium (Araceae): a case of mimicry? Evolutionary Ecology 23: Stenberg, J. A. and L. Ericson Visual cues override olfactory cues in the hostfinding process of the monophagous leaf beetle Altica engstroemi. Entomologia Experimentalis et Applicata 125: Szentesi, A., T. L. Hopkins and R. D. Collins Orientation responses of the grasshopper, Melanoplus sanguinipes, to visual, olfactory and wind stimuli and their combinations. Entomologia Experimentalia et Applicata 80: Todd, J. L., P. L. Phelan and L. R. Nault Interaction between visual and olfactory stimuli during host-finding by leafhopper, Dalbulus maidis (Homoptera: Cicadellidae). Journal of Chemical Ecoogy 16: Tuttle, A. F., D. N. Ferro and K. Idoine Role of visual and olfactory stimuli in host finding of adult cabbage root flies, Delia radicum. Entomologia Experimentalis et Applicata 47: Van den Berg, J., B. Torto, J. A. Pickett, L. E. Smart, L. J. Wadhams and C. M. Woodcock Influence of visual and olfactory cues on field trapping of the pollen beetle, Astylus atromaculatus (Col.: Melyridae). Journal of Applied Entomology 132: Vargas, R. R., A. J. Troncoso, D. H. Tapia, R. Olivares-Donoso and H. M. Niemeyer Behavioural differences during host selection between alate virginoparae of

167 157 generalist and tobacco-specialist Myzus persicae. Entomologia Experimentalie et Applicata 116: Vasconcellos-Neto, J. and R. F. Monteiro Inspection and evaluation of host plant by the butterfly Mechanitis lysimnia (Nymph., Ithomiinae) before laying eggs: a mechanism to reduce intraspecific competition. Oecologia 95: Visser, J. H Host odor perception in phytophagous insects. Annual Review of Entomology 31: Visser, J. H Host-plant finding by insects: orientation, sensory input, and search patterns. Journal of Insect Physiology 34: Vitzthum, H., M. Muller. and U. Homberg Neurons of the central complex of the locust Schistocerca gregaria are sensitive to polarized light. Journal of Neuroscience 22: Wenninger, E. J., L. L. Stelinksi and D. G. Hall Roles of olfactory cues, visual cues, and mating status in orientation of Diaphorina citri Kuwayama (Hemiptera: Psyllidae) to four different host plants. Environmental Entomology 38: Withers, T. M. and M. O. Harris Foraging for oviposition sites in the Hessian fly: random and non-random aspects of movement. Ecological Entomology 21: Yasui, H., M. Fukaya and S. Wakamura Behavioral responses in feeding to green color as visual stimulus with two lepidopteron larvae, Spodoptera litura (Fabricuius) (Noctuidae) and Milionia basalis pryeri Druce (Geometridae). Applied Entomology and Zoology 41:41-47.

168 158 Appendix 7.A. Bibliographic information for references on phytophagous insect visual ecology. Article titles were omitted because of space constraints (but full references can be seen in the References section for many of these articles). Note that the trapping studies do not represent an exhaustive list, as it was not the intent of this article to extensively review insect trapping. No plant finding examples were found for Phasmatodea or Symphyta. No trapping examples were found for Phasmatodea or Orthoptera. Paper Number Order or Suborder Plant or Trapping Study Reference Journal (volume: pages) 1 Coleoptera Plant Bjorklund et al Physiol. Entomol. 30: Coleoptera Plant Butkewich and Prokopy 1997 J. Entomol. Sci. 32: Coleoptera Coleoptera Plant Plant Egusa et al Goyer et al Entomol. Exp. Appl. 120: Forest Ecol. and Manag. 191: Coleoptera Plant Hausmann et al Environ. Entomol. 33: Coleoptera Plant Jonsson et al Physiol. Entomol. 32: Coleoptera Plant Reeves et al J. Insect Behav. 22: Coleoptera Plant Reeves and Lorch 2009 J. Insect Behav. 22: Coleoptera Plant Stenberg and Ericson 2007 Entomol. Exp. Appl. 125: Coleoptera Plant Szentesi et al Entomol. Exp. Appl. 105: Coleoptera Trap Ali 1993 Insect Sci. Appl. 2: Coleoptera Trap Blight and Smart 1999 J. Chem. Ecol. 25: Coleoptera Trap Lelito et al J. Appl. Entomol. 132: Coleoptera Coleoptera Trap Trap Mizell et al Pawson and Watt 2009 J. Entomol. Sci. 42: Agr. Forest Entomol. 11: Coleoptera Trap Strom et al Entomol. Exp. Appl. 100: 63-67

169 159 Paper Number Order or Suborder Plant or Trapping Study Reference Journal (volume: pages) 17 Coleoptera Trap Strom and Goyer 2001 Ann. Entomol. Soc. Am. 94: Coleoptera Trap Van den Berg et al J. Appl. Entomol. 132: Diptera Plant Aluja and Prokopy 1993 J. Chem. Ecol. 19: Diptera Plant Brévault and Quilici 2007 Entomol. Exp. Appl. 125: Diptera Plant Degen and Städler 1997 Entomol. Exp. Appl. 83: Diptera Plant Drew et al Entomol. Exp. Appl. 107: Diptera Plant Drew et al J. Econ. Entomol. 99: Diptera Plant Harris et al Physiol. Entomol. 18: Diptera Plant Harris and Miller 1983 Ann. Entomol. Soc. Am. 76: Diptera Plant Harris and Miller 1984 Physiol. Entomol. 9: Diptera Plant Harris and Rose 1990 Environ. Entomol. 19: Diptera Plant Henneman and Papaj 1999 Entomol. Exp. Appl. 93: Diptera Plant Judd and Whitfield 1997 Eur. J. Entomol. 94: Diptera Plant Katsoyannos 1987 J. Appl. Ent. 104: Diptera Plant Katsoyannos et al Entomol. Exp. Appl. 42: Diptera Plant Katsoyannos et al Entomol. Exp. Appl. 38: Diptera Plant Koštál 1991 Entomol. Exp. Appl. 59: Diptera Plant Kostal and Finch 1994 Entomol. Exp. Appl. 70: Diptera Plant Landolt and Reed 1990 Environ. Entomol. 19: Diptera Plant Owens and Prokopy 1986 Physiol. Entomol. 11: Diptera Plant Prokopy et al. 1993A Science 221: Diptera Plant Prokopy et al. 1993B Ent. Exp. Appl. 34: Diptera Plant Prokopy et al J. Insect Behav. 7: Diptera Plant Roessingh and Städler 1990 Entomol. Exp. Appl. 57: Diptera Plant Sérandour et al Entomol. Exp. Appl. 120: Diptera Plant Sharma and Franzmann 2001 J. Agric. Urban. Entomol. 18: Diptera Plant Tuttle et al Entomol. Exp. Appl. 47: Diptera Diptera Plant Trap Withers and Harris 1996 Alyokhin et al Ecol. Entomol. 21: J. Econ. Entomol. 93: Diptera Diptera Trap Trap Cornelius et al. 1999A Cornelius et al. 1999B Environ. Entomol. 28: J. Econ. Entomol. 92: Diptera Trap Epsky et al Environ. Entomol. 24:

170 160 Paper Number Order or Suborder Plant or Trapping Study Reference Journal (volume: pages) 49 Diptera Trap Finch 1995 Entomol. Exp. Appl. 74: Diptera Trap Hill and Hooper 1984 Entomol. Exp. Appl. 35: Diptera Trap Judd and Borden 1991 Entomol. Exp. Appl. 58: Diptera Trap Katsoyannos et al J. Econ. Entomol. 93: Diptera Trap Koštál and Finch 1996 Physiol. Entomol. 21: Diptera Trap Mayer et al J. Econ. Entomol. 93: Diptera Trap Pinero et al Entomol. Exp. Appl. 121: Diptera Trap Riedl and Hislop 1985 Environ. Entomol. 14: Diptera Trap Rull and Prokopy 2005 Entomol. Exp. Appl. 114: Heteroptera Plant Blackmer and Canas 2005 Environ. Emtomol. 34: Heteroptera Plant Blatt and Borden 1999 Environ. Entomol. 28: Heteroptera Plant Cook and Neal 1999 J. Ins. Behav. 12: Heteroptera Trap Blackmer et al Crop Prot. 27: Heteroptera Trap Hogmire and Leskey 2006 J. Entomol. Sci. 41: Heteroptera Trap Holopainen et al Agr. Food Sci. Finland. 10: Heteroptera Trap Legrand and Los 2003 Fla. Entomol. 86: Heteroptera Trap Leskey and Hogmire 2005 J. Econ. Entomol. 98: Heteroptera/Lepidoptera/Symphyta Trap Coli et al Agricult. Ecosys. Environ. 14: Homoptera Plant Ahman et al Sweedish J. agric. Res. 15: Homoptera Plant Bullas-Appleton et al Environ. Entomol. 33: Homoptera Plant Gish and Inbar 2006 J. Insect Behav. 19: Homoptera Plant Hajong and Varman 2002 J. Insect Behav. 15: Homoptera Plant Hodgson and Elbakhiet 1985 Entomol. Exp. Appl. 38: Homoptera Plant Ishii-Eiteman and Power 1997 Ecol. Appl. 7: Homoptera Plant Kirchner et al J. Insect Physiol. 51: Homoptera Plant Nissinen et al Eur. J. Entomol. 105: Homoptera Plant Patt and Sétamou 2007 Environ. Entomol. 36: Homoptera Homoptera Plant Plant Todd et al. 1990A Todd et al. 1990B J. Chem. Ecol. 16: Entomol. Exp. Appl. 54: Homoptera Homoptera Plant Plant Vargas.et al Wenninger et al Entomol. Exp. Appl. 116: Environ. Entomol. 38: Homoptera Plant/Trap Brennan and Weinbaum 2001 Environ. Entomol. 30:

171 161 Paper Number Order or Suborder Plant or Trapping Study Reference Journal (volume: pages) 81 Homoptera Trap Adams et al Environ. Entomol. 12: Homoptera Trap Döring et al Entomol. Exp. Appl. 113: Homoptera Trap Hardie et al Physiol. Entomol. 21: Homoptera Trap Horton and Lewis 1997 J. Econ. Entomol. 90: Lepidoptera Plant Badenes-Perez et al J. Econ. Entomol. 97: Lepidoptera Plant Balkenius et al J. Comp. Physiol. A. 192: Lepidoptera Plant Couty et al Physiol. Entomol. 31: Lepidoptera Plant Hirota and Kato 2001 Entomol. Exp. Appl. 101: Lepidoptera Plant Hirota and Y. Kato 2004 Appl. Entomol. Zool. 39: Lepidoptera Plant Kelber 1999 J. Exp. Biol. 202: Lepidoptera Plant Kelber et al Nature. 419: Lepidoptera Plant Mackay and Jones 1989 Ecol. Entomol. 14: Lepidoptera Plant Reddy et al Econ. Entomol. 102: Lepidoptera Plant Rojas and Wyatt 1999 Entomol. Exp. Appl. 91: Lepidoptera Plant Sambaraju and Phillips 2008 J. Insect Behav. 21: Lepidoptera Plant Scherer and Kolb 1987b J. Comp. Physiol. A. 161: Lepidoptera Plant Scherer and Kolb 1987a J. Comp. Physiol. A. 160: Lepidoptera Plant Vasconcellos-Neto and Monteiro 1993 Oecologia 95: Lepidoptera Plant Yasui et al Appl. Entomol. Zool. 41: Lepidoptera Trap Athanassiou et al J. Econ. Entomol. 97: Lepidoptera Trap Meagher Jr Fla. Entomol. 84: Lepidoptera Trap Myers et al J. Entomol. Sci. 44: Lepidoptera Trap Singh and Saxena 2004 Eur. J. Entomol. 101: Lepidoptera Trap Suckling et al J. Chem. Ecol. 31: Orthoptera Plant Bailey and Harris 1991 J. Ins. Behav. 4: Orthoptera Plant Szentesi et al Entomol. Exp. Appl. 80: Symphyta Trap Anderbrant et al J.Appl. Ent. 107: Symphyta Thysanoptera Trap Plant Barker et al Blumthal et al Entomol. Exp. Appl. 85: Hort. Tech. 15: Thysanoptera Thysanoptera Plant Plant/Trap Yaku et al Teulon et al Ecol. Entomol. 32: Entomol. Exp. Appl. 93: Thysanoptera Trap Carrizo 2008 Cien. Inv. Agr. 35:

172 162 Paper Number Order or Suborder Plant or Trapping Study Reference Journal (volume: pages) 113 Thysanoptera Trap Chen et al Environ. Entomol. 33: Thysanoptera Trap Chen et al Southwest. Entomol. 31: Thysanoptera Trap Childers and Brecht 1996 J. Econ. Entomol. 89: Thysanoptera Trap Chu et al Fla. Entomol. 89: Thysanoptera Trap Gillespiei and Vernonz 1990 J. Econ. Entomol. 83: Thysanoptera Trap Harman et al Pest Manag. Sci. 63: Thysanoptera Trap Kirk 1984 Ecol. Entomol. 9: Thysanoptera Trap Natwick et al Southwest. Entomol. 32: Thysanoptera Trap Rieske and Raffa 2003 J. Econ. Entomol. 96: Thysanoptera/Homoptera Trap Atakan and Canhilal 2004 J. Agric. Urban Entomol. 21: 15-24

173 CHAPTER VIII CONCLUSIONS The work presented in this dissertation has added to our understanding of aquatic plant biological control at many scales. At the largest scale, Chapter II showed that in general across agents and plants, biological control of aquatic and wetland plants appears to be successful. Chapter II also showed that experimental design may impact biological control studies, so we suggested that field studies with control groups that measure plant biomass or density in subsamples (quadrats, transects, etc.) of the treatment area be performed in the future. Because all of the experimental design factors in Chapter II showed significant heterogeneity, we were able to highlight several hypotheses that can be tested in the future to determine such things as why observational studies produce larger effects than experimental studies and whether this difference is the underlying cause of several of the other significant difference in experimental design seen in our meta-analysis. Though prior qualitative claims of successful aquatic plant biological control have been made (e.g., McFadyen 1998), the meta-analysis presented in Chapter II provides a valuable, quantitative review of the literature from the last three decades. Since biological control can provide the most ecologically and economically sound methodology for 163

174 164 controlling problematic plants (McFadyen 1998), the results in Chapter II should be reassuring to lake or wetland managers interested in utilizing biological control. On a relatively smaller scale, Chapter III (Reeves et al. 2008) showed that biological control of Eurasian watermilfoil (Myriophyllum spicatum L.) by the milfoil weevil (Euhrychiopsis lecontei Dietz) may be successful, as final plant densities were significantly correlated with time between initial and final surveys at treatment (weevilstocked) but not control (un-stocked) sites within lakes. However, these results also exhibited much variability in efficacy, both within and between lakes. Two factors that seemingly impact data interpretation in this system are timing of data collection and plant senescence, so better sampling methods were suggested. That is, initial and final data should be taken at similar time intervals between sites to make data more easily comparable between sites. Also, all data should be collected by mid-august in this system when possible to avoid the potential confound of plant senescence, a likely confound of the data in Chapter III. Staying with the same system, Chapters IV-VI focused on plant finding by E. lecontei. These chapters contribute to our understanding of control efficacy at an even finer scale still: the level of an individual control agent. Though these chapters did not explicitly focus on weevil damage to plants, plant finding mechanisms in general are understudied and will be important to eventually fully understand biological control of aquatic plants (Cuda et al. 2008). Because E. lecontei overwinters on land as adults in onshore leaf litter (Newman et al. 2001), the weevils have the seemingly large challenge each spring of relocating plants in a totally different habitat than that in which they

175 165 overwintered. Since one goal of biological control is long term control, plant finding becomes an important aspect of control efficacy for E. lecontei. Clearly, if E. lecontei cannot find M. spicatum in the spring, it cannot damage or control the plant, much less survive until the next season. Thus, an understanding of plant finding is directly related to understanding control efficacy because we would not be able to predict when E. lecontei can find and control plants until we understand how E. lecontei finds plants. As a compliment to the work on chemosensory abilities of E. lecontei that has been done (Marko et al. 2005), Chapters IV-VI focused on vision as a potential plant finding mechanism for E. lecontei, especially since olfaction may not be critically important for aquatic herbivores (Spanhoff et al. 2005). Chapter IV (Reeves et al. 2009b) showed that weevils use vision for plant location. This was demonstrated in multiple ways, such as weevils significantly finding plants in the light more than the dark. Weevils were also attracted to plant stems that were sealed in vials to prevent any chemical detection. Next, weevils were attracted to sealed plants even in very turbid water. Finally, though vision clearly appeared to be important to E. lecontei, weevils were not able to differentiate M. spicatum from coontail, Ceratophyllum demersum L. in no-choice trials where only one of the plants was present in the experimental arena at a time. To further explore the role of vision for plant finding and potentially differentiation by E. lecontei, choice trials were performed in Chapter V (Reeves and Lorch 2009) using M. spicatum and C. demersum simultaneously placed side-by-side (and sealed in vials to prevent any chemical detection) in the experimental arenas. It was shown that weevils significantly preferred M. spicatum based purely on visual cues,

176 166 further highlighting that vision may be important to E. lecontei, even to the point of allowing them to differentiate plant species. To try and determine if weevils used shape or color while differentiating M. spicatum from C. demersum, both choice and no-choice trials were performed in Chapter V using brown versus green M. spicatum stems. Weevils showed no preference for green stems over brown stems, meaning that plant form may be more important than color during visual plant finding. To compliment Chapters IV and V, Chapter VI extended the understanding of visual cue use by E. lecontei by showing that weevils were attracted to plants at distances of up to 17.5 cm, and may have sufficient visual acuity to discern plants amongst other attractive visual stimuli at distances of up to 15 cm. With the information from Chapters IV-VI, along with previous studies on E. lecontei behavior and life history factors such as chemical cue use (Marko et al. 2005) and overwintering habits (Newman et al. 2001), a conceptual model of weevil plant finding was built at the conclusion of Chapter VI (Fig. 6.4). Though some steps of weevil plant finding remain somewhat unanswered in the literature (such as lake finding in the spring and finding overwintering habitat in the fall), the work presented in Chapters IV-VI clearly add much to the understanding of plant finding in this system. The plant finding model also adds to the very sparse understanding of host-plant location by aquatic insects (Newman 2004), especially those which are biological control agents (Cuda et al. 2008). Because the use of chemical cues for long range host location by aquatic herbivores may be unlikely (Spanhoff et al. 2005; Dusenbery 1992), the use of vision for host-plant location should be explored for other aquatic phytophagous insects. The

177 167 conceptual model (Fig. 6.4) presented in Chapter VI may be applicable to aquatic insects in general, and may contribute to better understanding plant finding. It became clear to me while reviewing the literature for Chapters IV-VI that vision in general has largely been ignored when compared to chemical cues as the major attractants for host location by phytophagous insects. This is a phenomenon that has gone unchanged since noted by Prokopy and Owens (1983). Because of this, Chapter VII was written to extend the reach of this dissertation beyond aquatic biological control agents to insect-plant systems in general. Many physical factors may make chemical detection difficult for insects (Dusenbery 1992), and since vision may be a more precise search mechanism for insects than olfaction (Bell 1990), more work is clearly warranted for determining the role that vision plays for plant location by phytophagous insects. Chapter VII used examples from the recent literature (i.e., since the review by Prokopy and Owens 1983) to highlight the potential importance of vision for phytophagous insects and to debunk some of the historical assumptions that vision is not important. Overall, it can be said that aquatic plant biological control is successful at the scale of the field of biological control as a whole (Chapter II). However, at a smaller scale, biological control efficacy may differ from lake to lake, as evidenced by the variable M. spicatum control by E lecontei in Chapter III. At a smaller scale still, plant finding by individual biological control agents of aquatic plants is understudied and likely very important (Cuda et al. 2008), so Chapters IV-VI aid greatly in improving our understanding of plant finding by E. lecontei, and potentially contribute to a better understanding of this phenomenon in aquatic insects generally. Knowledge of factors that

178 168 may affect control efficacy such as study design (Chapters II and III) or plant finding behavior of the agents (Chapters IV-VI) can be important since biological control may provide the most economically and ecologically sound method for controlling problematic plants (McFadyen 1998). The reach of this dissertation is extended from to the host-plant finding and selection literature in general by Chapter VII and Appendix I (Reeves et al. 2009a). Chapter VII argues for more research on the visual capabilities of phytophagous insects in general (both aquatic and terrestrial) by reviewing recent literature and showing that many assumptions about the unimportance of vision for phytophagous insects may be ungrounded. Also, Appendix I provided a method for growing plants to study host-plant selection in insects that have below-ground larvae. These contributions, along with all of those noted above, make this dissertation what is hoped to be a valuable addition not only to the biological control literature, but to the general host-plant finding and selection literature as well. References Bell, W. J Searching behavior patterns in insects. Annual Review of Entomology 35: Cuda, J. P., R. Charudattan, M. J. Grodowitz, R. M. Newman, J. F. Shearer, M. L. Tamayo and B. Villegas Recent advances in biological control of submersed aquatic weeds. Journal of Aquatic Plant Management 46: Dusenbery, D. B Sensory Ecology, W.H. Freeman and Company, New York.

179 169 Marko, M. D., R. M. Newman and F. K. Gleason Chemically mediated host-plant selection by the milfoil weevil: a freshwater insect-plant interaction. Journal of Chemical Ecology 31: McFadyen, R.E.C Biological control of weeds. Annual Review of Entomology 43: Newman, R. M Biological control of Eurasian watermilfoil by aquatic insects: basic insights from an applied problem. Archiv fur Hydrobiologie 159: Newman, R. M., D. W. Ragsdale, A. Milles and C. Oien Overwinter habitat and the relationship of overwinter to in-lake densities of the milfoil weevil, Euhrychiopsis lecontei, a Eurasian watermilfoil biological control agent. Journal of Aquatic Plant Management 39: Prokopy, R. J. and E. D. Owens Visual detection of plants by herbivorous insects. Annual Review of Entomology 28: Reeves, J. L., P. D. Lorch, M. W. Kershner and M. A. Hilovsky Biological control of Eurasian watermilfoil by Euhrychiopsis lecontei: assessing efficacy and timing of sampling. Journal of Aquatic Plant Management 46: Reeves, J. L., B. A. Foote and P. D. Lorch. 2009a. A method for growing legumes with and without root nodules for studying nodule-attacking Rivellia (Diptera: Platystomatidae). Proceedings of the Entomological Society of Washington 111:

180 170 Reeves, J. L. and P. D. Lorch Visual plant differentiation by the milfoil weevil, Euhrychiopsis lecontei Dietz (Coleoptera: Curculionidae). Journal of Insect Behavior 22: Reeves, J. L., P. D. Lorch and M. W. Kershner. 2009b. Vision is important for plant location by the phytophagous aquatic specialist Euhrychiopsis lecontei Dietz (Coleoptera: Curculionidae). Journal of Insect Behavior 22: Spanhoff, B., C. Kock, A. Meyer and E. I. Meyer Do grazing caddisfly larvae of Melampophylax mucoreus (Limnephilidae) use their antennae for olfactory food detection? Physiological Entomology 30:

181 APPENDIX I A METHOD FOR GROWING LEGUMES WITH AND WITHOUT ROOT NODULES FOR STUDYING NODULE-ATTACKING RIVELLIA (DIPTERA: PLATYSTOMATIDAE) Reprinted from Proceedings of the Entomological Society of Washington (no copyright): Reeves, J.L., B.A. Foote and P.D. Lorch A method for growing legumes with and without root nodules for studying nodule-attacking Rivellia (Diptera: Platystomatidae). Proc. Entomol. Soc. Wash. 111: Introduction Insects that live below ground or have below ground stages are often not considered in conventional ecological theory (Johnson et al. 2006). In terrestrial insects with subterranean larvae, larvae have limited mobility relative to adults. The preferenceperformance hypothesis states that herbivorous insects with larvae that have limited locomotive ability for finding new host-plants have strong selective pressure to oviposit 171

182 172 on plants that will allow for larval success (Johnson et al. 2006). Thus, in systems where terrestrial insects have subterranean larvae, plant condition below ground may drive above ground oviposition choice, though few studies have examined this idea (Johnson et al. 2006). Flies in the genus Rivellia Robineau-Desvoidy (Diptera: Platystomatidae) are an example of insects with subterranean, nodule-feeding larvae. They are all specialists on legumes in the subfamily Faboideae (Foote et al. 1987). Females oviposit in the soil of their host-plants and newly hatched larvae enter and consume the root nodules of their host-plant (for an overview of the development of and role played by root nodules in legume growth see Lindemann and Glover 2003). These larvae have limited mobility and must find nodules below ground after hatching on the soil surface. This creates strong selection for females to oviposit on host-plants in appropriate condition (i.e, plants with healthy nodules). For this reason, flies in the genus Rivellia are good candidates for testing whether below ground plant condition affects above ground ovipositional behavior. Koethe and Van Duyn (1984) showed that the larvae of R. quadrifasciata (Macquart) had an 89% survival rate on nodulating soybean plants compared to 9% survival on nodule-free plants, so a selective advantage clearly exists for oviposition on nodulating plants. Oviposition tests using species of Rivellia will require the ability to see how gravid females respond to plants with and without root nodules. In this note, we describe a successful method for growing two legumes, both with and without root nodules, that may be attacked by species of Rivellia. This growing method will be useful

183 173 for examining the ovipositional response of these terrestrial insects with subterranean larvae that attack legume root nodules. The legume growing methods presented here should be useful to anyone studying host-plant choice in legume pest (and non-pest) systems with subterranean larvae. These methods should work for any nodule-producing legume species and thus may be applicable to many systems beyond Rivellia, such as the weevil Sitona lineatus L. (Coleoptera: Curculionidae) that attacks peas and beans (George 1962). Methods Two legume species were used for this experiment: Desmodium paniculatum and D. canadense, as at least one Rivellia species (R. steyskali Namba and probably R. quadrifasciata) utilizes Demodium spp. (Foote et al. 1987). Seeds for both plant species were obtained from the Prairie Moon Nursery (Winona, MN, USA) and germinated in Fafard Superfine Germinating Mix. All germinating mix and soil was moistened and autoclaved for 20 minutes prior to use to eliminate all Rhizobia, the bacteria responsible for triggering nodule growth in legumes (Lindemann and Glover 2003). About two weeks after germination, the seedlings were transferred to 10.5cm x 10.5cm pots filled with autoclaved Fafard 1-P soil. Forty plants of each species were transplanted singly into pots for a total of 80 plants. Half of these pots for each species were inoculated with Rhizobia by liberally applying Burpee Booster for Peas and Beans (which contains many different Rhizobium spp.) to the soil, creating nodule and non-nodule treatments. Because root nodules help fix nitrogen from the air for use by the plant, we fertilized our

184 174 non-nodule treatment plants with a general (N-P-K) fertilizer to ensure enough nitrogen was present for proper plant development. We waited five days before fertilizing any of the seedlings to allow them to get established in their respective pots. After a few weeks of only fertilizing the nodule-free plants, fertilizer was applied to the Rhizobiatreated plants, as they were not growing vigorously and this kind of fertilizer does not negatively affect root nodules in legumes (Lindemann and Glover 2003). All plants were grown in the Kent State University research greenhouse under natural light conditions between May and August and were watered daily. The Rhizobia-inoculated and Rhizobiafree plants were kept separated on a long bench in the greenhouse. About seven weeks after transplanting the seedlings into individual pots, all plants were killed and their roots examined for nodules. Results and Discussion This method of growing plants to obtain individuals with and without root nodules was successful. Of the 40 plants (20 from each species) in the Rhizobiainnoculated treatment, all 40 had nodules. Of the 40 plants in the Rhizobia-free treatment, only two had nodules (both D. canadense). The two plants with unexpected nodules likely grew the nodules via soil (Rhizobia) contamination from one of the Rhizobiainoculated pots. We presume the contamination occurred when seedlings were knocked over within their pots during watering and propped up by the experimenter without washing hands after contacting plants or soil, or by soil insects moving between pots. These results suggest that with reasonable care and better plant separation, this method of

185 175 growing plants may be a reliable way to get plants for study of female oviposition choice in R. steyskali or R. quadrifasciata. This growing method should also work for other legume species and thus for other species of Rivellia or other nodule-attacking insects. Durst and Bosworth (1986) noted that nodules with a pink center are active and healthy, so when the plants were sacrificed, a plant was randomly selected and a sample of its nodules cut open. The insides of the nodules were as described by Durst and Bosworth (1986) as being healthy and active, so this method produces plants with healthy nodules. When the plants were sacrificed, the Rhizobia-free individuals were slightly larger, possibly due to the use of chemical fertilizer early in development. The roots were also finer and denser in the Rhizobia-free plants. However, based on qualitative observations, the Rhizobia-inoculated plants were generally greener and healthier looking. To see how our greenhouse-grown root nodules compared to naturally grown nodules in the field, 28 nodules representing four D. paniculatum plants were collected in the field (all plants collected in Portage County, OH, USA) and compared to 28 nodules representing four greenhouse-raised D. paniculatum plants. Diameters of the widest axis on the mostly spherical nodules were measured and compared between the two populations. The nodules were not different between greenhouse and field plants (Means ± 1SE: greenhouse nodules = 1.67 ± 0.13 mm, field nodules = 1.67 ± 0.11 mm; t-test: t = 0.07; df = 54; P = 0.94). Though nodule size was similar between greenhouse and fieldgrown plants, it was observed that large tap roots were present in all field plants, while they were greatly reduced or absent in the greenhouse-raised plants. This could be

186 176 because the small size of the pots prevented the greenhouse-raised plants from producing a large taproot, or because the taproots in the field represented growth from previous years since Desmodium are perennial plants. Growing legumes with and without nodules can help answer some fundamental questions about host/oviposition choice in legume-associated insects such as Rivellia. Future work using these methods could include testing whether species of Rivellia will oviposit only on nodule-producing plants, or only on plants with large, active nodules. There are different Rivellia species that attack different legumes, one of which (R. quadrifasciata) is a pest on cultivated soybeans in the southern U.S. (Koethe and Van Duyn 1984). Gaining an understanding of basic oviposition choice by nodule-attacking insects such as Rivellia will not only be useful in testing hypotheses like the preferenceperformance hypothesis relative to subterranean insects, but may lead to better methods of controlling pests like R. quadrifasciata. Koethe and Van Duyn (1984) found that R. quadrifasciata larvae are more successful on nodulating soybean plants than nonnodulating plants, however they did not examine oviposition behavior relative to the two plant types. Also, the nodulating and non-nodulating plants used by Koethe and Van Duyn (1984) were different strains of soybeans. Differences between strains (other than nodule production) may have affected larval survival, so repeating their larval success experiment using the plant growing methods reported here with the same plant strains may prove useful. The ability to produce nodulating and non-nodulating plants of the same strain should open up the possibility for studies of oviposition behavior in species of Rivellia and other species with nodule-feeding larvae.

187 177 Acknowledgments We thank Chris Rizzo for his advice and assistance during this experiment. References Durst, P. and S. Boworth Inoculation of forage and grain legumes. Agronomy Facts 11, Department of Crop and Soil Sciences Cooperative Extension, Pennsylvania State University. Foote, B. A., B. D. Bowker and B. A. McMichael Host plants for North American species of Rivellia (Diptera: Platystomatidae). Entomological News 98: George, K. S Root nodule damage by larvae of Sitona lineatus and its effect on yield of green peas. Plant Pathology 11: Johnson, S. N., A. N. E. Birch, P. J. Gregory and P. J. Murray The mother knows best principle: should soil insects be included in the preference-performance debate? Ecological Entomology 31: Koethe, R. W. and J. W. Van Duyn Aspects of larva/host relationships of the soybean nodule fly, Rivellia quadrifasciata (Diptera: Platystomatidae). Environmental Entomology 13: Lindemann, W. C. and C. R. Glover Nitrogen fixation by legumes. Guide A-129, Cooperative Extension Service, College of Agriculture and Home Economics, New Mexico State University.

188 APPENDIX II REPRINT PERMISSIONS FOR PUBLISHED CHAPTERS Chapter 3 Reeves, J. L., P. D. Lorch, M. W. Kershner and M. A. Hilovsky Biological control of Eurasian watermilfoil by Euhrychiopsis lecontei: assessing efficacy and timing of sampling. Journal of Aquatic Plant Management 46: Reprinted with permission from Michael D. Netherland, Editor, Journal of Aquatic Plant Management: 178

189 179 Chapter 4 Reeves, J. L., P. D. Lorch and M. W. Kershner Vision is important for plant location by the phytophagous aquatic specialist Euhrychiopsis lecontei Dietz (Coleoptera: Curculionidae). Journal of Insect Behavior 22: Reprinted with permission from Springer:

190 180 Chapter 5 Reeves, J. L. and P. D. Lorch Visual plant differentiation by the milfoil weevil, Euhrychiopsis lecontei Dietz (Coleoptera: Curculionidae). Journal of Insect Behavior 22: Reprinted with permission from Springer: