• Duong, Tammy M.
  • Gomez, Ann B.
  • Sherratt, Thomas N.


Aposematism is an evolved, cross-species association between a preys’ unprofitability and the presence of conspicuous signals. Avian predators have been widely employed to understand the evolution of these warning signals However, insect predators are abundant, diverse, and highly visual foragers that have been shown to be capable of learned aversion. Therefore, it is likely that their behaviour also shapes the nature of anti-predator traits. In this study, we evaluated the rates of attack of a community (13 species) of mature adult dragonflies (Odonata) on artificial prey of varying size (2.5–31 mm lengthwise) and colour pattern (black, black/yellow striped). The relative attack rates of dragonflies on prey increased as prey size decreased, but there was no evidence that the attack rates by dragonflies were affected by prey colour pattern and no evidence for an interaction between colour pattern and size. To investigate prey selection by specific predator species under field conditions, we compared the time to attack distributions of black-painted prey presented to two common dragonflies: Leucorrhinia intactaand the larger, Libellula pulchella. We found that the two dragonfly species, as well as the two sexes, had different foraging responses. Lpulchella was more likely to attack larger prey, and females of both species more likely to attack prey than males. Collectively, our results indicate that dragonflies are highly size selective. However, while the nature of this selectivity varies among dragonfly species, there is little evidence that classic black/yellow warning signals deter attack by these aerial invertebrate predators.


We conducted two experiments: a dragonfly community prey choice experiment and a species-specific prey choice experiment. The experiments were conducted in East Field meadow (approx. 1 ha; 44°32’29” N, 76°22’17” W) near Queens University Biological Station, Ontario, Canada. All dragonflies were tested from 0900 to 1500 local time, on sunny days, with temperatures ranging from 13–30°C. Dragonflies were at particularly high density in this meadow (S1 Video) and almost certainly one of the major predators of invertebrate prey in the area. Surveys of the dragonfly community were conducted every two weeks to monitor species composition and relative abundances. These surveys were performed by first walking the perimeter of the field then spiralling inwards, towards the centre, tallying the species of individuals observed perching as we walked.

Community prey choice experiment

The community prey choice experiments were performed between 3 June—3 July and from 17 July—30 July 2015. To study the preferences of dragonflies for prey of different size and colour pattern, artificial prey were created by gluing two spherical beads of the same diameter together (loosely mimicking the shape of Hymenopteran/Dipteran prey) and painted using non-toxic acrylic paint (DecorArt Crafter’s Acrylic paint in “Bright Yellow” and “Black”) (Fig 1). Prey consisted of seven different sizes (2.5, 5, 9, 12, 16, 18 and 31 mm; lengthwise) and two colour treatments (black, and black and yellow). The black and yellow stripes were scaled such that prey of all sizes received 5 black stripes and 4 yellow stripes. However, the smallest size treatment (2.5 mm) could not be painted with stripes and was presented only in black.

Our artificial prey were presented on a pairwise 183 cm long (total length) Y-shaped rod made of bamboo (with each prong measuring 53 cm in length)–see Fig 1. Prey were hung 100 cm from the ends of each prong using dark green Power Pro microfilament braided fishing line (10 lb test; Innovative Textiles). Treatment prey were always presented pairwise with a standard bead (black 12 mm bead). To prevent potential side bias, the standard bead was alternately tethered on each prong with each presentation. This standard bead was introduced to facilitate preference standardization: the emergence times of dragonfly species did not completely overlap and with inevitable seasonal changes in environmental conditions (such as temperature and wind speed) it was appropriate to compare the number of attacks on a given prey type with the number of attacks against a constant standard model. This approach gives a relatively direct measure of the nature of preference for one phenotype over another. To present the beads, we first located a perched dragonfly and then approached it slowly while holding the rod 2–3 m above the ground. The beads were then hung approximately 0.5–1.5 m above the dragonfly and moved in a manner that mimicked natural prey movement (i.e. side to side as if flying across the field of vision). Dragonflies were given four minutes to attack prey. When a prey item was attacked (i.e. the dragonfly made direct physical contact with the bead), the type of bead (standard or treatment), size, and colour were noted. After an attack occurred, the dragonfly was caught and marked with two small dashes on the hind wing using a permanent marker, which allowed us to avoid re-testing the same dragonfly. The sex and species of all tested dragonflies were identified and recorded. The rate of re-encounter with dragonflies that had already been tested was extremely low (2 re-encounters of 1153 attempted presentations) as the density of dragonflies was extremely high. No teneral dragonflies, readily identifiable through their “glassy” wings, were tested.

In addition to our analysis of the whole community, we sought to evaluate the broad scale patterns of foraging responses of dragonfly species in relation to their body size. The foraging preferences of thirteen species of dragonflies were evaluated (Table 1). Species were classified into two (approximately equal sized) small (Leucorrhinia intactaCelithemus elisaCelithemus eponinaLeucorrhiniaglacialisSympetrum spp., Epitheca cynosuraLeucorrhinia frigida, and Pachydiplax longipennis) and large (Libellula JuliaLibellula quadrimaculataLibellula luctuosaGomphus spicatus, and Libellula pulchella) size-class categories for total body length based on [26]. We tested whether there were differences in attack frequency or size preference for given prey types that could be explained by differences in the body size class of predator by fitting a generalized linear model (see below).