INTRODUCTION

Predators are thought to influence species composition of communities when prey avoid otherwise suitable habitats that contain predators (e.g. Fraser and Huntingford, 1986; Wilson and Lefcort, 1993). Young prey can suffer predation from different species of predators than do adult prey and this is especially true for species of amphibians and aquatic insects that have an aquatic larval stage and a terrestrial adult stage. Thus it is not surprising to find that adults of these prey species consider risk of predation to themselves as well as their offspring when selecting a suitable place to lay eggs. For example, salamanders (Hopey and Petranka, 1994), frogs (Binckley and Resetarits, 2002), mosquitoes (Blaustein et al., 2004), beetles (Resetarits, 2001), and midges (Petranka and Fakhoury, 1991) have been found to avoid ovipositing in areas that contain predators of their larvae.

Damselflies (Odonata: Suborder Zygoptera) have aquatic larvae but terrestrial adults. Fish can be important predators of larvae of some species and such species appear restricted to sites without fish (Johnson and Crowley, 1980; Pierce et al., 1985; McPeek, 1989, 1990a). This distribution is known to relate to behavior of larvae. Species with active larvae are vulnerable to fish predation and are rare or absent at sites with fish. In comparison, species with larvae that are reluctant to move when approached by predators are more vulnerable to predation by dragonfly larvae and are most common in sites with fish and few dragonflies (McPeek, 1990b). Thus, the distribution of some damselfly species is likely, at least in part, determined by direct predation on larvae of different species by different types of predators. That adults can return to natal ponds to oviposit likely reinforces the existing distribution (McPeek, 1989).

However, adults of some species disperse from their emergence sites (McPeek, 1989) and, as a result, can quickly establish a new population at sites where a new habitat is created, or where the predator community changes (Corbet, 1999). It is possible that the distribution of species across habitats also reflects the ability of dispersing adults to detect and avoid predators of their larvae. Adult females of species whose larvae cannot coexist with fish may be able to detect fish predators at a site, and choose not to oviposit there. McPeek (1989) found dispersing adult adults of some Enallagma species do not distinguish between fishless and fish-containing sites. However, he combined his results for males and females and one of his experiments was done entirely with males. Ovipositing females may be more likely to be discriminate between habitats than mate-searching males as detection of aquatic predators is likely most important during oviposition.

Our study focused on the ability of male and female damselflies to detect the presence of fish predators. We first surveyed a number of sites in eastern Ontario to determine distribution of Enallagma boreale Selys, E. ebrium (Hagen), and E. signatum (Hagen) across fish and fishless habitats. We then performed a field enclosure experiment to test if presence of fish and models of conspecific ovipositing females affect habitat selection and oviposition and two laboratory experiments to test whether females of the three species oviposited preferentially in water with or without chemical cues of fish. We experimented with ovipositing conspecifics because some species of damselflies are more likely to oviposit in areas containing conspecifics (Waage, 1987; Rehfeldt, 1990).

METHODS

Field Survey

We surveyed three fish containing and three fishless sites near the Queen's University Biological Station (44°34′N, 79°15′W) in eastern Ontario to determine if Enallagma ebrium, E. boreale, and E. signatum show similar habitat distributions as the same species seen elsewhere (McPeek, 1990a). Fishless sites were Jack's Marsh, Barb's Marsh, and Upper Dowsley Pond, fish-containing sites were Two Island Lake, Indian Lake Bight, and Lake Opinicon. All sites, except for Lake Opinicon, are described elsewhere by Forbes et al. (1995) (where Jack's Marsh is labelled SEP). Those sites were all located within 10 km of Lake Opinicon. Lake Opinicon is well studied and general descriptions of emergence sites for odonates can be found in Gerrish (1980). We surveyed sites in the first week of June, July, and August in both 1997 and 1998. During each sampling period, one of us (MM) spent 15 min catching adult damselflies with an aerial insect net at each of four locations at each site. Individuals were immediately placed in 95% ETOH for later identification and for measurements not included in this study. Larvae were surveyed the same day as adults. They were captured using three standardized dip net sweeps (Baker, 1989) at each location within each site. Penultimate and final instar larvae were immediately sorted, killed, and stored in 95% ETOH for later identification using keys in Walker (1953).

Experiment in Semi‐natural Enclosures: Effects of Fish and Conspecifics on Habitat Selection

We tested effects of fish and conspecifics on habitat selection by allowing damselflies to choose among four combinations of ‘‘habitats’’ (fish presence or absence crossed with conspecific presence or absence). Experiments took place in a mesh-walled tent (base = 3.2 m2, height = 2.2 m, mesh size = 2.0 mm2) containing four plastic wading pools (diameter =  105 cm, depth = 20.5 cm), each representing one of the four habitat combinations. Pools were filled with well water to a depth of 15 cm over a black plastic liner chosen to mimic the color of pond sediments. Each pool contained four ‘‘oviposition stations.’’ Stations consisted of three wooden dowel perches (length =  15 cm, diameter = 2 mm) wrapped in strips of green construction paper and planted in a block of Oasis artificial planting material (5 cm × 8 cm × 8 cm) on the bottom of the pools. A circle of green construction paper (diameter = 8 cm) was placed between the perches below the water line to provide additional substrate for oviposition. A fourth wooden dowel perch was attached horizontally to the side of the pool, in close proximity to each station.

The four treatment combinations (Control, Fish, Conspecifics, Fish + Conspecifics) were randomly assigned to pools before each trial. For treatments using conspecifics, a ‘‘model female’’ (a recently killed female of the same species pinned to a perch as if ovipositing) was added to each station. For treatments with fish, two fish of approximate fork length 10 cm were lifted from a tank containing both pumpkinseed sunfish (Lepomis gibbosus (Linnaeus)) and bluegill sunfish (L. macrochirus (Rafinesque)) and added to the pool. All fish were collected from Cow Island Marsh in Lake Opinicon, near the Queen's University Biology Station. Fish were fed earthworms and maintained in a flow-through tank supplied with water from the lake. Groups of approximately 10 fish were kept for 2–4 weeks before being released and replaced by a new group. New fish were selected from the stock populations for each trial.

We performed 14 trials with Enallagma ebrium between June 26 and July 26, 1997. We also studied E. boreale and E. signatum but sample sizes were too low for analysis. Animals were collected from the field adjacent to Two Island Lake (approximately 6 km from Queen's University Biological Station). On the day prior to a trial, 12 tandem pairs and 8 single males were captured between 3:30 and 4:30 p.m., placed in a cage and transported to the station. Upon arrival, all males were released into the tent; females remained in the cage overnight and were released in the tent at approximately 9:30 a.m. to coincide with the beginning of the flight period. We also added fish and model females to the appropriate pools at this time.

We made observations between 10:30 a.m. and 3:30 p.m. During this time, we recorded number of ‘‘visits’’ made to each pool by single males, single females, and tandem pairs. A visit consisted of a damselfly, or pair of damselflies, landing on a perch or on the side of a pool. We also recorded the number and length of ovipositions performed by single females and tandem pairs. An oviposition consisted of a female curving her abdomen to touch the substrate, and pumping the abdominal muscles to release and insert eggs into the oviposition substrate. All ovipositing behavior performed on a single visit to a pool was recorded as a single oviposition.

Statistical analysis was performed on log (x+1) transformed data. We used a two-way, repeated-measures MANOVA on visit data. Categorical variables were fish presence or absence crossed with female model presence or absence; response variables were visits by males, females, and tandem pairs. Each trial was used as a single independent replicate and we pooled the data across oviposition sites within each pool and across all individuals and used each of the four pools as the repeated measure. We used two separate two-way, repeated-measures ANOVAs for oviposition frequency and oviposition duration with the same categorical variables and repeated measures as for the MANOVA. Significance of the oviposition data was determined with a sequential Bonferroni test. We used randomization analysis because data were non-normal (Shapiro and Wilks’ W statistic, p<0.05). Data sets were reordered between treatments using an SAS random number generator and reordered data were subjected to the MANOVA testing as outlined above. We repeated this process 5000 times and compared the distribution of resulting Wilks’ lambdas to those provided by the original data. Original results were considered significant if they fell within the lower 5% of the randomized results distribution (Manly, 1991).

Laboratory Experiment 1: Effects of Fish-Cured Water on Oviposition When No Choice of Oviposition Site was Offered

We tested if chemical cues of fish presence affected oviposition by collecting females and exposing them to either fish-cured or control water in the laboratory. Because of differences in phenology, E. boreale females were collected June 4 – June 9, 1997, from Jack's Marsh, E. ebrium were collected June 19 – July 16, 1997, at Two Island Lake, and E. signatum were collected July 29 – August 9, 1997, from Lake Opinicon.

We collected only mature damselflies in tandem to increase the probability that females were receptive. We collected between 10 and 20 tandem pairs at the start of the daily flight period on each day. We placed animals in a cage (30 cm × 30 cm × 40 cm) and transported them to the Queen's University Biological Station where we left them in a sun lit area during the normal flight period for a minimum of 3 h. This allowed damselflies to finish mating but prevented them from ovipositing.

We then released individual females in randomly assigned, covered 75 mL plastic cups. We used 82 E. ebrium, 62 E. boreale, and 110 E. signatum with equal numbers of females within each species in each treatment. Each 75 mL cup contained a wooden dowel perch and a small piece of filter paper resting on a platform of plastic mesh. The filter paper provided a substrate for oviposition and was immersed in 50 mL of either well water (Control) or in 50 mL of water from a flow-through tank housing fish. Pumpkinseed sunfish and bluegill sunfish were again collected from Cow Island Marsh in Lake Opinicon. All fish had a fork length of ca. 10 cm and were fed earthworms; fish were maintained in a flow-through tank supplied with water from the lake. Again, groups of approximately 10 fish were kept for 2–4 weeks before being released and replaced by a new group.

We checked cups the following morning and recorded whether oviposition had occurred and counted number of eggs laid. On the basis of McPeek's (1990a) results and our own experience with Enallagma species co-existing with fish (Baker, 1989), we predicted that E. boreale would oviposit less on fish-cured substrates, that such avoidance be less expressed by E. ebrium, and that E. signatum may oviposit more on fish-cured substrates.

We first used 2 × 2 contingency tests to compare, within each species, the proportion of females that did or did not lay eggs in control versus fish treatments. We then excluded data for non-egg-laying females and used a two-way ANOVA to test whether number of eggs laid depended on the species being considered, the fish treatment or the interaction between the two.

Laboratory Experiment 2: Effects of Fish‐Cured Water on Oviposition When a Choice of Oviposition Sites was Offered

In the previous experiment, females were not presented with a choice of substrates. We designed a second experiment to test whether fish chemical cues affect oviposition by females allowed to choose between Fish and Control Treatments. In 1998, E. boreale adults were collected May 22 – June 17; E. ebrium were collected June 21 – July 21, and E. signatum were collected July 27 – August 8 at the same sites and using the same techniques described above.

Once mating was completed, single females were placed individually in plastic tubs (28 cm × 19 cm × 12 cm) covered with window screening. Each tub contained two plastic dishes (12 cm diameter, 4 cm deep), with a plastic mesh platform surrounding and extending between the dishes; the platform prevented damselflies from becoming trapped between dishes and sides of the tub. Each dish contained a wooden dowel perch and three small pieces of filter paper, immersed in water, and resting on a plastic mesh platform. Each tub had both Control and Fish Treatments present (as described above); treatment location was alternated between left and right side of replicate tubs. We recorded data as in the previous experiment.

A total of 78 E. boreale, 98 E. ebrium, and 48 E. signatum females were tested. We used McNemar tests (Zar, 1996) to compare, within each species, the proportion of egg-laying females in control versus fish treatments. We used paired t‐tests to examine whether number of eggs laid by egg-laying females differed between the two types of substrates; separate analyses were used for each species.

RESULTS

Field Survey

Adults of Enallagma ebrium were found at all six sites and larvae of E. ebrium were found at two of the fishless sites and two of the sites with fish (Table I). No Enallagma boreale larvae were found at any site and no adults were found at sites with fish; however, adults were found at all three fishless sites. Larval E. signatum were found only at one fish site: Lake Opinicon (also the collection site for E. signatum and fish used in laboratory studies). Our field surveys included dates when E. signatum adults should have been active but we did not find any in surveys at Lake Opinicon. In contrast, we had no trouble collecting sufficient numbers of adults of this species for our experiments. This may reflect the timing of daily activity of this species (to maximize likelihood of getting enough adults for experimentation, collections of this species for experiments were done later in the day than for surveys). On the basis of similar surveys over several years, we are quite certain Enallagma signatum are absent from fishless sites.

Table I. Total Number of Adults/Larvae of the Three Species Found at the Six Sites During the Six Sampling Periods
Full size table

Effects of Fish and Conspecifics on Habitat Selection

The two-way repeated measures MANOVA indicated no effect of model females on visits (Wilks’ lambda = 0.77, p = 0.39) but fish presence affected number of visits (Wilks’ lambda = 0.30, p = 0.003) (Fig. 1). Both males and females avoided pools with fish (males: F = 8.90, p = 0.003; females: F = 6.55, p = 0.017). Tandems made fewer visits to pools containing fish, but this difference was not significant (F = 0.14, p = 0.70; total visits by tandems were fewer than those by males or females and thus the apparent difference in response to fish presence may simply reflect reduced statistical power). There was no significant interaction effect of model females and fish presence on visits (Wilks’ lambda = 0.85, p = 0.61).

Fig. 1.
figure 1

Mean (±SE) visits by Enallagma ebrium males, females, and tandem pairs to pools with neither fish nor models (Control), models only, fish only, or both.

Full size image

The two-way, repeated measures ANOVAs on oviposition frequency and oviposition duration found no significant effects of model females (frequency: F = 1.24, p = 0.28; duration: F = 0.003, p = 0.95), fish (frequency: F = 3.60, p = 0.08; duration: F = 3.29, p = 0.09), or any interaction between the two (frequency: F = 0.21, p = 0.65; duration: F = 1.50, p = 0.24) (Table I). However, both the number and duration of oviposition events tended to be higher when there were no fish in the pools (Fig. 2).

Fig. 2.
figure 2

Mean (±SE) frequency of ovipositions (upper part) and duration of oviposition (lower part) by E. ebrium females under the same treatments listed in Fig. 1.

Full size image

Laboratory Experiment 1: Effects of Fish‐Cured Water on Oviposition When No Choice of Oviposition Site was Offered

Seventy-nine percent of E. boreale females laid eggs but whether they laid eggs or not was not affected by fish‐cured versus control water (χ2 = 0.09, df = 1, p = 0.76, Table II); similar results were found for E. ebrium (78% laid eggs, χ2 = 1.1, df = 1, p = 0.29). For E. signatum, 64% of the females laid eggs in the fishy‐cured water (χ2 = 4.81, df = 1, p = 0.03); however, after using a Bonferroni correction to control for the multiple tests (three species), there was no significant effect. We next excluded females that failed to lay eggs and compared whether log 10 (numbers of eggs laid) depended on species, treatment, or an interaction between the two. Species accounted for significant variation in numbers of eggs laid (F(2, 182) = 14.3, p < 0.001), after controlling for the insignificant effect of control versus fish treatment (F(1, 182) = 0.025, p = 0.87) and after removing the insignificant interaction term (F(1, 182) = 2.15, p = 0.14) as suggested by Wilkinson (1989). Whereas E. boreale and E. ebrium females averaged similar numbers of eggs after combining treatments (169.8 versus 204.2, respectively), E. signatum females laid far fewer eggs (47.8 eggs, pooled standard error was 85 eggs for all three species, Table II).

Table II. Proportion of Females (% Females) of Each Species that Oviposited in Experiment 1 (No Choice of Substrate) and Experiment 2 (Choice of Substrate) and the Mean Number of Eggs (±SE) Laid by Ovipositing Females of Each Species
Full size table

Taken together, we found interspecific variation in oviposition by the three species but no evidence oviposition was affected by chemical cues of fish. If fish cues are important in egg laying for these species, these results indicate they are not detectable by chemical cues alone.

Laboratory Experiment 2: Effects of Fish‐Cured Water on Oviposition When a Choice of Oviposition Sites was Offered

Surprisingly, female E. boreale and E. ebrium were far less likely to lay eggs in this experiment than they were in the previous experiment; only 33% of E. boreale females and 37% of E. ebrium laid eggs. In contrast, 70% of the female E. signatum laid eggs, a value similar to that in the first experiment. Many females in all three species laid eggs on only one type of substrate; specifically, 22 E. boreale females laid eggs on one type of substrate only compared to four that laid eggs on both substrates. Numbers of females ovipositing on single versus both substrates were 32 and 5 for E. ebrium and 18 and 15 for E. signatum. McNemar tests indicated no significant differences between treatments in presence or absence of eggs (E. boreale: χ2 = 1.14, df = 1, p = 0.29; E. ebrium: χ2 = 2.53, df = 1, p = 0.11; E. signatum: χ2 = 0.06, df = 1, p = 0.81). Egg numbers were not significantly different between control and fish-cured substrates for any of the species (E. boreale: t = 1.15, df = 25, p = 0.87; E. ebrium: t = −1.49, df = 35, p = 0.15; E. signatum: t = 0.91, df = 32, p = 0.81, Table II). This experiment is consistent with the results of the first experiment; moreover all three species showed similar patterns of egg laying with respect to substrates. More E. signatum chose to lay eggs, and even though there was no preference for control or fish-cured substrate, most laid all of their eggs on one substrate. In other words, once females chose to lay their eggs, they deposited most or almost all of them on one substrate.

DISCUSSION

Field Distribution and Dispersal

Distribution of the three species across fish and fishless habitats were similar to those found by McPeek (1990a) in south western Michigan; E. boreale was restricted to fishless sites and E. signatum was restricted to sites with fish. Our results differed from those of McPeek in that he found adults of E. ebrium at both types of habitat but never found larvae in sites without fish whereas we found relatively large numbers of E. ebrium at both types of sites and found larvae of this species in sites without fish. Similarly, we have found E. ebrium in fishless lakes in southern Ontario. For example, Baker (1989) found larvae of E. ebrium in 15 of the 17 sites surveyed and larvae of Lestes (not identified to species) were found at 9 sites, all of which contained E. ebrium. Larvae of Lestes are more common at sites without fish (Macan, 1977; Stoks and McPeek, 2003) but the presence of Lestes larvae alone is not particularly strong evidence for the absence of fish as some species can co-exist with fish. For example, Stoks and McPeek (2003) found L. vigilax only in sites with fish. However L. vigilax were not likely present at any of the sites in Baker's (1989) study as he collected and identified adult Lestes at all sites and never found any L. vigilax. More important, E. boreale larvae (a species widely recognized as not co-existing with fish) were found in three of the sites that contained E. ebrium. It may be that the presence of larvae of E. ebrium, E. boreale and Lestes represents a ‘‘mistake’’ in oviposition site selection by adults of either group and that the larvae are unlikely to survive. However, four of the ponds in Baker's (1989) survey had final instar larvae of both E. ebrium and Lestes, and one had final instar larvae of both E. ebrium and E. boreale, indicating the sites were suitable for complete development of E. ebrium despite the apparent lack of fish predation and an expected higher risk of predation from dragonflies.

McPeek (1989) notes that, owing to mortality costs associated with dispersal, and given adults cannot discriminate between fish and fishless sites and thus risk ovipositing in an inappropriate site, natural selection should favor philopatry. This suggestion is supported by four of the five species he studied being strongly philopatric. He also notes that E. ebrium disperses much more than the other species and this agrees with our work near the Queens University Biological station; i.e., E. ebrium can be very common as adults at sites with relatively few or no larvae. McPeek suggests the tendency for E. ebrium to disperse is related to the cycling of winter kill fish lakes between fish and fishless states, i.e. he notes selection may favor dispersal as some dispersing animals may find sites where E. ebrium populations are reduced and because dispersing animals may survive to reproduce when a winter kill lakes converts to fishless conditions. McPeek's argument is suitable for his study site but the fact that we find E. ebrium larvae in fishless sites suggests a simpler explanation for why that species disperses more than its congeners in eastern Ontario. Unlike its congeners, E. ebrium can live in both fish and fishless sites in Ontario and thus, unlike its congeners, it is obviously less likely for females of that species to make a mistake and oviposit in an inappropriate site. The ability of E. ebrium to live in sites without fish in southern and eastern Ontario but not in such sites in southern Michigan may relate to differences in density and species composition of Anisoptera between the two geographic areas.

Fish Detection in Enallagma ebrium Adults

Our results suggest it is unlikely that adult Enallagma damselflies detect or interpret chemical cues of fish presence when making decisions about where to oviposit. In our laboratory experiments, there were no significant differences found in the number of eggs laid or oviposition frequency between the two treatments, either when each treatment was presented alone, or when both treatments were presented simultaneously.

One concern with our second laboratory experiment is that, because the control and fish-cured substrates were in close proximity, any putative kairomones from the fish-cured substrate may have diffused throughout the container. This diffusion would make it more difficult for females to assess which substrate is most appropriate based on air-borne cues. However, because the tubs were covered with mesh and thus open to the surrounding air, it is extremely likely that the concentration of any kairomones in the water of the dishes containing fish-cured substrates would have been higher than that in dishes with no fish-cured substrate. Thus our results provide no evidence that females can detect or respond to presence of fish on the basis of expected differences in kairomone concentration in the water. Our results are consistent with the recent review by Corbet (1999) who notes adults are known to select oviposition sites on the basis of visual, tactile, and even thermal cues but there is no clear evidence that any species makes use of olfactory cues.

In contrast to our laboratory results on chemical cues of fish, our field experiment indicated that both male and female E. ebrium made fewer visits to pools with fish although they did not respond to the presence of conspecifics ovipositing females. This clearly suggests adults are capable of detecting and avoiding fish predators; we suspect they simply see fish and move away from them. Duration and frequency of oviposition were not significantly affected by fish presence but the trend was for longer durations and more ovipositions in the absence of fish.

We did not expect this disparity between visiting and ovipositing behavior with regard to fish presence; we expected response to aquatic predators would be strongest during oviposition. The disparity may stem from there being fewer ovipositions than visits made at each pool on any given day, resulting in decreased statistical power for the analysis of oviposition. We ran power tests on both frequency and duration data for each factor (fish and model females) and for the interaction between the two. In each case, power was less than 0.5; thus power was generally low for oviposition data and the chance of making a Type II error was always higher than 50%.

However, it may be that oviposition is a fixed action pattern and this may explain why oviposition duration was not affected by fish presence. More specifically, a female may not detect the presence of a fish when first visiting a site and, once she begins to oviposit, she may no longer be able respond to fish presence and will continue to oviposit, even after a fish moves into her immediate area. It is also possible the oviposition substrate acts as an effective barrier to fish detection or subsequent predation attempts.

Another possibility is that adults may choose to avoid fish when visiting a site due to a perceived predation risk to themselves (ovipositing Zygoptera are susceptible to predation by fish and invertebrates such as Notonecta; Zeiss et al., 1999) and not because of any concern over their larvae). Note that E. ebrium is a species that is found at sites both with and without fish and therefore, since the larvae are capable of surviving in both situations, it is possible that adults do not consider fish presence when ovipositing. However, while female E. ebrium may not heed the presence of fish when selecting sites for their larvae, the fact that they do detect and avoid fish suggests other species, i.e. those whose larvae cannot co-exist with fish, may also be able to detect fish presence and use the information to select areas safe for their larvae. Studies similar to ours on species such as E. boreale would be useful in this context.