Oecologia (2001) 127:222–229 DOI 10.1007/s004420000595
Robby Stoks
Food stress and predator-induced stress shape developmental performance in a damselfly Received: 15 August 2000 / Accepted: 13 November 2000 / Published online: 2 February 2001 © Springer-Verlag 2001
Abstract I studied effects of stress factors like food shortage, non-lethal predator presence and autotomy on survival and larval performance (growth rate, development rate and developmental stability) of larvae of the damselfly Lestes sponsa. In a laboratory experiment, larvae were raised during their last two instars at two food levels (high or low) crossed with two levels of autotomy (caudal lamellae present or absent). These treatments were nested within three levels of predation risk (Aeshna cyanea absent, Chironomus-fed caged Aeshna or Lestesfed caged Aeshna). The diet of the predator had no effects. The low food level and the presence of Aeshna independently increased mortality rates of L. sponsa larvae. The low food level, presence of a caged Aeshna and autotomy all independently reduced growth rate (mass and body size at day 40) and wing size at emergence, and the first two stress factors also reduced development rate. Regardless of predator presence and autotomy, all damselfly larvae consumed the food available. This indicated that the predator-induced stress effects were not due to reduced food uptake, but probably reflected lowered assimilation efficiency and/or a higher metabolic rate. Besides a low food level, the presence of caged Aeshna predator larvae and autotomy also increased hind wing asymmetry. This result demonstrated that predatorinduced stress may reduce developmental stability in the prey. Keywords Autotomy · Fluctuating asymmetry · Life history · Predation risk · Stress
R. Stoks (✉) Evolutionary Biology Group, University of Antwerp (RUCA), Groenenborgerlaan 171, 2020 Antwerpen, Belgium Present address: R. Stoks, Laboratory of Aquatic Ecology, Katholieke Universiteit Leuven, De Bériotstraat 32, B-3000 Leuven, Belgium, e-mail: [email protected] Tel.: +32-16-323966, Fax: +32-16-324575
Introduction During ontogeny, animals may face several stress factors. Here a stress factor is defined operationally as any environmental factor that reduces population growth via a reduction in survival and/or developmental performance (modified from Calow 1989). Although most studies focus on lethal effects, nonlethal stress effects may be as or even more important to population dynamics (e.g. McPeek and Peckarsky 1998). Furthermore, although mostly studied in isolation, in nature, stress factors often operate in concert and effects may only become detectable when these factors are combined (reviewed in Folt et al. 1999; Woods et al. 1999). Most animals are confronted with biotic stress factors originating both from lower levels in the food chain (food stress) as well as from higher levels (predatorinduced stress). Often these effects are combined through direct and indirect effects in the food chain. Indeed, predators are mostly presumed to affect the prey’s developmental performance (growth rate, development rate and developmental stability) through reducing food uptake by the prey (reviewed in Lima 1998). Little is known about whether predator-associated stress can also negatively affect the developmental performance of the prey irrespective of food uptake (but see Duvall and Williams 1995; Boonstra et al. 1998). Furthermore, studies dealing with the effects of predation risk on developmental performance have primarily focused on effects caused by the immediate presence of the predator (e.g. Skelly and Werner 1990; Peckarsky et al. 1993; Ball and Baker 1996; McPeek 1998). Predators may also induce behavioural and/or morphological changes that remain present and costly even in the predator’s absence (e.g. McCollum and Van Buskirk 1996). One extreme of predator-induced phenotypes is autotomy of body parts; here, morphological effects clearly persist after the confrontation with the predator. Autotomy is also a potential stress factor, which has been shown to decrease survival and developmental performance (reviewed in Niewiarowski et al. 1997; see also Harris 1989).
223
Both nutritional and predator-induced stress may decrease developmental performance in several ways. Traditionally, effects of food and predator stress on growth rate and development rate have been studied (reviewed in Boutin 1990; Lima 1998). Autotomy has been shown to reduce growth (e.g. Niewiarowski et al. 1997). However this pattern is not general (reviewed in Niewiarowski et al. 1997), and effects on development rate have not been studied. During the last decade, it has become evident that developmental stability may also be affected by environmental stress (reviewed by Møller 1998). Developmental stability reflects the ability of animals to achieve a predetermined phenotype that is optimal under given environmental conditions (Møller and Swaddle 1997). Fluctuating asymmetry (random deviations from the optimal symmetrical phenotype) is used as a measure of developmental stability (Møller and Swaddle 1997). Some studies have shown that food stress can increase asymmetry (reviewed in Møller and Swaddle 1997). However, results are not always consistent (e.g. Bjorksten et al. 2000). Predator-induced stress can be expected to increase asymmetry, but data on this are sparse and indirect. Witter and Lee (1995) showed that starlings raised with less cover, and hence expected to suffer from a higher perceived predation risk, developed more asymmetrical feathers. In this study, I examined how nutritional stress and two predator-induced stress factors, namely non-lethal predator presence and autotomy, affect survival and developmental performance of larvae of the damselfly Lestes sponsa. I quantify measures of developmental performance, namely growth rate and development rate (which are responsible for mass, size and age at emergence), and developmental stability (as measured by wing asymmetry), which have been shown to affect fitness in mature damselflies. Using a complete factorial design, both individual and joint effects of these biotic stress factors can be evaluated. I raised individual larvae with a certain supply of food irrespective of predator-associated treatments. This enabled me to test the hypothesis that predator-induced stress can decrease developmental performance independently of effects arising indirectly through changes in food intake. I also tested if autotomy reduces development rate, and if not only food stress, but also predator-induced stress decreases developmental stability.
Materials and methods Study species Lestid damselflies of the genus Lestes are widespread in Europe, and the study species L. sponsa is one of the most abundant species in Belgium. Within the genus Lestes, the loss of lamellae (caudal appendages) is high: up to 90% of individuals have at least one missing or regenerating lamella and almost 20% are missing all three lamellae (Stoks 1998b). Lamellae loss in this species has been shown to function as autotomy, i.e. the loss of a body part to escape predation from large dragonfly larvae, notonectids and sticklebacks (Stoks and De Block 2000). L. sponsa has a fast life-
style (Johansson 2000), which means that it achieves rapid growth and development due to active searching for prey (Pickup and Thompson 1990). Lestid damselfly larvae typically live in small temporary, or otherwise fishless, ponds (Jödicke 1997). Larvae of L. sponsa were collected from a pond near Antwerp, northern Belgium. I determined their age using a combination of head width and wing bud length (Pickup et al. 1984), and only larvae within instar F-2 (F-0 is the final instar) with three undamaged, original lamellae were used in the experiment. Regenerated lamellae can be recognized by their altered morphology (underdeveloped pattern of tracheae and setae). Each larva was housed separately in the laboratory at 18°C and at a 16:8 h light:dark constant photoperiod which approximates the average natural photoperiod during the larval life of L. sponsa (Pickup and Thompson 1984). Larvae were fed ad libitum with Daphnia sp. Directly after their moult into instar F-1, larvae were placed under the experimental conditions. I exposed the damselflies to the presence of Aeshna cyanea dragonfly larvae because they are important predators of L. sponsa (Stoks et al. 1999a, 1999b). Dragonfly larvae (instar F-2) were collected from the same pond and housed separately in the laboratory under the same conditions as the L. sponsa larvae and fed ad libitum with Chironomus larvae. Experimental procedure The experimental design included two food levels (high/low) crossed with two levels of autotomy (caudal appendages present/absent). All larvae (n=144) were placed individually in small transparent cups (diameter 5 cm, height 4.5 cm) and were given a cocktail stick to provide a site for prey capture. In the high food treatment, L. sponsa larvae received ten Daphnia each day; in the low food treatment, they received ten Daphnia every 3 days. Daphnia were sieved to standardise for size. F-2 instar L. sponsa larvae were fed Daphnia of a smaller size (mean±1 SE: 1.99±0.26 mm, n=10) than F-1 and F-0 instars (2.59±0.37 mm, n=10). By extrapolating growth of L. sponsa fed different ratios in the study of Pickup and Thompson (1984, 1990), I found that ten Daphnia per day of the given size classes give development rates somewhat lower than maximal, while the low feeding regime should result in low development rates. Previous studies showed that a feeding regime of one Daphnia per day (1/3 the low ration in this study) causes moderate mortality (e.g. Johansson 1996). Lamellae were gently removed by pulling them with two fingers. Larvae of the noautotomy treatment were handled similarly, but lamellae were not removed. Each set of four treatment combinations was placed in a box (15×10×11, length×width×height), filled with aerated tap water (4 cm depth, 0.60 l). Within a box, I manipulated predator presence (three levels): (1) no predator present, (2) one free-ranging Aeshna given five Chironomus larvae every 3 days (Chironomusfed Aeshna) or (3) one free-ranging Aeshna given two L. sponsa larvae every 3 days (Lestes-fed Aeshna). Each set of four combinations of food level and autotomy was replicated 12 times for each of the three predator treatments, giving a total of 36 boxes. I offered Aeshna two diet types, because several studies have indicated that prey react more to the kairomones released by injured conspecifics than to the predator per se (e.g. Crowl and Covich 1990). Damselfly larvae respond to both visual and chemical stimuli from predators (Wisenden et al. 1997). Therefore, larvae were reared in transparent cups with the original bottom replaced with a small netting (mesh size 0.25 mm), so that the L. sponsa larvae could see and smell the predator, but no contact could be made. Response variables To test the hypotheses, I scored survival and developmental performance. The studied developmental performance parameters were selected because of their relationship with adult fitness in the study species or in other damselflies. Reduced development rates
224 increase the time larvae remain vulnerable to predators (Banks and Thompson 1987), increase the probability of mortality because of habitat desiccation (e.g. Jödicke 1997; R. Stoks, personal observation) and delay emergence which may negatively affect adult fitness (Plaistow and Siva-Jothy 1999). Growth rate and development rate together affect mass and size at emergence. Higher values of these parameters typically increase survivorship of adults of both sexes, as well as lifetime mating success in males and fecundity in females (reviewed in Sokolovska et al. 2000). Finally, males with more symmetrical wings have higher lifetime mating success in this species (R. Stoks, unpublished data; see also Harvey and Walsh 1993 for the damselfly Coenagrion puella). Larvae were checked each day for deaths and moults. After 40 days, when all surviving larvae had reached instar F-0, and when no larvae had emerged, mass and size of all larvae were measured. By measuring all larvae at the same time, these variables reflect larval growth rate. Since larvae were very fragile at the begin of the experiment (just moulted in instar F-1), I decided not to measure them at this moment, because this could have stressed them and increased stress levels across all treatments. Instead, larvae were randomly assigned to treatment combinations, making it very improbable that there were any initial mass or size differences among treatments. Larvae were gently blotted on paper towel and weighed to the nearest 0.1 mg using an electrobalance. Body length (distance from the tip of the rostrum to the end of the abdomen, excluding lamellae) was measured to the nearest 0.01 mm using dial calipers. I corrected for autotomy by subtracting the mean wet weight of three lamellae (2.37±0.25 mg, n=20) from the wet weight of larvae that had not undergone autotomy. After measurement, all larvae were replaced in their cups until emergence. For larvae that emerged, I calculated development rates separately for instar F-1 and F-0 as 1/(time spent in the instar) (units are day–1; see Pickup and Thompson 1990). After emergence, animals were killed by freezing, and their wings were cut off with scissors close to the point of insertion. Both hind wings of each animal were magnified and digitised using Optimas software. I only measured wings that were fully extended during emergence. For each wing, I recorded the distance from the junction between the costa and first antenodal crossvein towards the wing top three times. Repeated measures of the same wing are necessary to evaluate the importance of measurement error in shaping observed left-right differences (Palmer 1994; see data analysis). Measuring the left and right hind wing of the same individual enables calculation of absolute differences between sides within an individual, and I used this as a measure of wing asymmetry (cf. Palmer 1994).
Because survival to emergence is a binary variable (dead/alive), I analysed treatment effects on survival using a split-plot ANOVA with a binomial error structure and the logit link. The model was fitted using the glimmix macro of the MIXED procedure of SAS 6.12 (Littell et al. 1996). I analysed the treatment effects on the development rates of instar F-1 and F-0 only in larvae that emerged (n=87) using a repeated-measures split-plot ANOVA. Development rates were log-transformed to reduce heteroscedasticity. I averaged the three length measurements of each wing, and when both left and right hind wing were intact, I took the average length of both wings to determine the hind wing length (the single means were used for animals with only one fully extended wing). I tested for effects of treatment on wing length at emergence using a split-plot ANOVA. To analyse wing asymmetry, I first checked for directional asymmetry and measurement error by performing a two-way mixed-model ANOVA with individual as a random effect and side as a fixed effect (Palmer 1994). A significant side effect indicates directional asymmetry. Hind wing asymmetry was not directional (mixed-model two-way-ANOVA: side effect, F1, 34=0.00, P=0.99), and the between-sides variation was significantly larger than the measurement error (side×individual: F34,140=92.10, P
Robby Stoks
Food stress and predator-induced stress shape developmental performance in a damselfly Received: 15 August 2000 / Accepted: 13 November 2000 / Published online: 2 February 2001 © Springer-Verlag 2001
Abstract I studied effects of stress factors like food shortage, non-lethal predator presence and autotomy on survival and larval performance (growth rate, development rate and developmental stability) of larvae of the damselfly Lestes sponsa. In a laboratory experiment, larvae were raised during their last two instars at two food levels (high or low) crossed with two levels of autotomy (caudal lamellae present or absent). These treatments were nested within three levels of predation risk (Aeshna cyanea absent, Chironomus-fed caged Aeshna or Lestesfed caged Aeshna). The diet of the predator had no effects. The low food level and the presence of Aeshna independently increased mortality rates of L. sponsa larvae. The low food level, presence of a caged Aeshna and autotomy all independently reduced growth rate (mass and body size at day 40) and wing size at emergence, and the first two stress factors also reduced development rate. Regardless of predator presence and autotomy, all damselfly larvae consumed the food available. This indicated that the predator-induced stress effects were not due to reduced food uptake, but probably reflected lowered assimilation efficiency and/or a higher metabolic rate. Besides a low food level, the presence of caged Aeshna predator larvae and autotomy also increased hind wing asymmetry. This result demonstrated that predatorinduced stress may reduce developmental stability in the prey. Keywords Autotomy · Fluctuating asymmetry · Life history · Predation risk · Stress
R. Stoks (✉) Evolutionary Biology Group, University of Antwerp (RUCA), Groenenborgerlaan 171, 2020 Antwerpen, Belgium Present address: R. Stoks, Laboratory of Aquatic Ecology, Katholieke Universiteit Leuven, De Bériotstraat 32, B-3000 Leuven, Belgium, e-mail: [email protected] Tel.: +32-16-323966, Fax: +32-16-324575
Introduction During ontogeny, animals may face several stress factors. Here a stress factor is defined operationally as any environmental factor that reduces population growth via a reduction in survival and/or developmental performance (modified from Calow 1989). Although most studies focus on lethal effects, nonlethal stress effects may be as or even more important to population dynamics (e.g. McPeek and Peckarsky 1998). Furthermore, although mostly studied in isolation, in nature, stress factors often operate in concert and effects may only become detectable when these factors are combined (reviewed in Folt et al. 1999; Woods et al. 1999). Most animals are confronted with biotic stress factors originating both from lower levels in the food chain (food stress) as well as from higher levels (predatorinduced stress). Often these effects are combined through direct and indirect effects in the food chain. Indeed, predators are mostly presumed to affect the prey’s developmental performance (growth rate, development rate and developmental stability) through reducing food uptake by the prey (reviewed in Lima 1998). Little is known about whether predator-associated stress can also negatively affect the developmental performance of the prey irrespective of food uptake (but see Duvall and Williams 1995; Boonstra et al. 1998). Furthermore, studies dealing with the effects of predation risk on developmental performance have primarily focused on effects caused by the immediate presence of the predator (e.g. Skelly and Werner 1990; Peckarsky et al. 1993; Ball and Baker 1996; McPeek 1998). Predators may also induce behavioural and/or morphological changes that remain present and costly even in the predator’s absence (e.g. McCollum and Van Buskirk 1996). One extreme of predator-induced phenotypes is autotomy of body parts; here, morphological effects clearly persist after the confrontation with the predator. Autotomy is also a potential stress factor, which has been shown to decrease survival and developmental performance (reviewed in Niewiarowski et al. 1997; see also Harris 1989).
223
Both nutritional and predator-induced stress may decrease developmental performance in several ways. Traditionally, effects of food and predator stress on growth rate and development rate have been studied (reviewed in Boutin 1990; Lima 1998). Autotomy has been shown to reduce growth (e.g. Niewiarowski et al. 1997). However this pattern is not general (reviewed in Niewiarowski et al. 1997), and effects on development rate have not been studied. During the last decade, it has become evident that developmental stability may also be affected by environmental stress (reviewed by Møller 1998). Developmental stability reflects the ability of animals to achieve a predetermined phenotype that is optimal under given environmental conditions (Møller and Swaddle 1997). Fluctuating asymmetry (random deviations from the optimal symmetrical phenotype) is used as a measure of developmental stability (Møller and Swaddle 1997). Some studies have shown that food stress can increase asymmetry (reviewed in Møller and Swaddle 1997). However, results are not always consistent (e.g. Bjorksten et al. 2000). Predator-induced stress can be expected to increase asymmetry, but data on this are sparse and indirect. Witter and Lee (1995) showed that starlings raised with less cover, and hence expected to suffer from a higher perceived predation risk, developed more asymmetrical feathers. In this study, I examined how nutritional stress and two predator-induced stress factors, namely non-lethal predator presence and autotomy, affect survival and developmental performance of larvae of the damselfly Lestes sponsa. I quantify measures of developmental performance, namely growth rate and development rate (which are responsible for mass, size and age at emergence), and developmental stability (as measured by wing asymmetry), which have been shown to affect fitness in mature damselflies. Using a complete factorial design, both individual and joint effects of these biotic stress factors can be evaluated. I raised individual larvae with a certain supply of food irrespective of predator-associated treatments. This enabled me to test the hypothesis that predator-induced stress can decrease developmental performance independently of effects arising indirectly through changes in food intake. I also tested if autotomy reduces development rate, and if not only food stress, but also predator-induced stress decreases developmental stability.
Materials and methods Study species Lestid damselflies of the genus Lestes are widespread in Europe, and the study species L. sponsa is one of the most abundant species in Belgium. Within the genus Lestes, the loss of lamellae (caudal appendages) is high: up to 90% of individuals have at least one missing or regenerating lamella and almost 20% are missing all three lamellae (Stoks 1998b). Lamellae loss in this species has been shown to function as autotomy, i.e. the loss of a body part to escape predation from large dragonfly larvae, notonectids and sticklebacks (Stoks and De Block 2000). L. sponsa has a fast life-
style (Johansson 2000), which means that it achieves rapid growth and development due to active searching for prey (Pickup and Thompson 1990). Lestid damselfly larvae typically live in small temporary, or otherwise fishless, ponds (Jödicke 1997). Larvae of L. sponsa were collected from a pond near Antwerp, northern Belgium. I determined their age using a combination of head width and wing bud length (Pickup et al. 1984), and only larvae within instar F-2 (F-0 is the final instar) with three undamaged, original lamellae were used in the experiment. Regenerated lamellae can be recognized by their altered morphology (underdeveloped pattern of tracheae and setae). Each larva was housed separately in the laboratory at 18°C and at a 16:8 h light:dark constant photoperiod which approximates the average natural photoperiod during the larval life of L. sponsa (Pickup and Thompson 1984). Larvae were fed ad libitum with Daphnia sp. Directly after their moult into instar F-1, larvae were placed under the experimental conditions. I exposed the damselflies to the presence of Aeshna cyanea dragonfly larvae because they are important predators of L. sponsa (Stoks et al. 1999a, 1999b). Dragonfly larvae (instar F-2) were collected from the same pond and housed separately in the laboratory under the same conditions as the L. sponsa larvae and fed ad libitum with Chironomus larvae. Experimental procedure The experimental design included two food levels (high/low) crossed with two levels of autotomy (caudal appendages present/absent). All larvae (n=144) were placed individually in small transparent cups (diameter 5 cm, height 4.5 cm) and were given a cocktail stick to provide a site for prey capture. In the high food treatment, L. sponsa larvae received ten Daphnia each day; in the low food treatment, they received ten Daphnia every 3 days. Daphnia were sieved to standardise for size. F-2 instar L. sponsa larvae were fed Daphnia of a smaller size (mean±1 SE: 1.99±0.26 mm, n=10) than F-1 and F-0 instars (2.59±0.37 mm, n=10). By extrapolating growth of L. sponsa fed different ratios in the study of Pickup and Thompson (1984, 1990), I found that ten Daphnia per day of the given size classes give development rates somewhat lower than maximal, while the low feeding regime should result in low development rates. Previous studies showed that a feeding regime of one Daphnia per day (1/3 the low ration in this study) causes moderate mortality (e.g. Johansson 1996). Lamellae were gently removed by pulling them with two fingers. Larvae of the noautotomy treatment were handled similarly, but lamellae were not removed. Each set of four treatment combinations was placed in a box (15×10×11, length×width×height), filled with aerated tap water (4 cm depth, 0.60 l). Within a box, I manipulated predator presence (three levels): (1) no predator present, (2) one free-ranging Aeshna given five Chironomus larvae every 3 days (Chironomusfed Aeshna) or (3) one free-ranging Aeshna given two L. sponsa larvae every 3 days (Lestes-fed Aeshna). Each set of four combinations of food level and autotomy was replicated 12 times for each of the three predator treatments, giving a total of 36 boxes. I offered Aeshna two diet types, because several studies have indicated that prey react more to the kairomones released by injured conspecifics than to the predator per se (e.g. Crowl and Covich 1990). Damselfly larvae respond to both visual and chemical stimuli from predators (Wisenden et al. 1997). Therefore, larvae were reared in transparent cups with the original bottom replaced with a small netting (mesh size 0.25 mm), so that the L. sponsa larvae could see and smell the predator, but no contact could be made. Response variables To test the hypotheses, I scored survival and developmental performance. The studied developmental performance parameters were selected because of their relationship with adult fitness in the study species or in other damselflies. Reduced development rates
224 increase the time larvae remain vulnerable to predators (Banks and Thompson 1987), increase the probability of mortality because of habitat desiccation (e.g. Jödicke 1997; R. Stoks, personal observation) and delay emergence which may negatively affect adult fitness (Plaistow and Siva-Jothy 1999). Growth rate and development rate together affect mass and size at emergence. Higher values of these parameters typically increase survivorship of adults of both sexes, as well as lifetime mating success in males and fecundity in females (reviewed in Sokolovska et al. 2000). Finally, males with more symmetrical wings have higher lifetime mating success in this species (R. Stoks, unpublished data; see also Harvey and Walsh 1993 for the damselfly Coenagrion puella). Larvae were checked each day for deaths and moults. After 40 days, when all surviving larvae had reached instar F-0, and when no larvae had emerged, mass and size of all larvae were measured. By measuring all larvae at the same time, these variables reflect larval growth rate. Since larvae were very fragile at the begin of the experiment (just moulted in instar F-1), I decided not to measure them at this moment, because this could have stressed them and increased stress levels across all treatments. Instead, larvae were randomly assigned to treatment combinations, making it very improbable that there were any initial mass or size differences among treatments. Larvae were gently blotted on paper towel and weighed to the nearest 0.1 mg using an electrobalance. Body length (distance from the tip of the rostrum to the end of the abdomen, excluding lamellae) was measured to the nearest 0.01 mm using dial calipers. I corrected for autotomy by subtracting the mean wet weight of three lamellae (2.37±0.25 mg, n=20) from the wet weight of larvae that had not undergone autotomy. After measurement, all larvae were replaced in their cups until emergence. For larvae that emerged, I calculated development rates separately for instar F-1 and F-0 as 1/(time spent in the instar) (units are day–1; see Pickup and Thompson 1990). After emergence, animals were killed by freezing, and their wings were cut off with scissors close to the point of insertion. Both hind wings of each animal were magnified and digitised using Optimas software. I only measured wings that were fully extended during emergence. For each wing, I recorded the distance from the junction between the costa and first antenodal crossvein towards the wing top three times. Repeated measures of the same wing are necessary to evaluate the importance of measurement error in shaping observed left-right differences (Palmer 1994; see data analysis). Measuring the left and right hind wing of the same individual enables calculation of absolute differences between sides within an individual, and I used this as a measure of wing asymmetry (cf. Palmer 1994).
Because survival to emergence is a binary variable (dead/alive), I analysed treatment effects on survival using a split-plot ANOVA with a binomial error structure and the logit link. The model was fitted using the glimmix macro of the MIXED procedure of SAS 6.12 (Littell et al. 1996). I analysed the treatment effects on the development rates of instar F-1 and F-0 only in larvae that emerged (n=87) using a repeated-measures split-plot ANOVA. Development rates were log-transformed to reduce heteroscedasticity. I averaged the three length measurements of each wing, and when both left and right hind wing were intact, I took the average length of both wings to determine the hind wing length (the single means were used for animals with only one fully extended wing). I tested for effects of treatment on wing length at emergence using a split-plot ANOVA. To analyse wing asymmetry, I first checked for directional asymmetry and measurement error by performing a two-way mixed-model ANOVA with individual as a random effect and side as a fixed effect (Palmer 1994). A significant side effect indicates directional asymmetry. Hind wing asymmetry was not directional (mixed-model two-way-ANOVA: side effect, F1, 34=0.00, P=0.99), and the between-sides variation was significantly larger than the measurement error (side×individual: F34,140=92.10, P
