Many chemically defended aposematic species are characterised by relatively low toxin levels, which enables predators to include them in their diets under certain circumstances. Knowledge of the conditions governing the survival of such prey animals – especially in the context of the co-occurrence of similar but undefended prey, which may result in mimicry-like interactions – is crucial for understanding the initial evolution of aposematism.

In a one-month outdoor experiment using fish (the common carp Cyprinus carpio) as predators, we examined the survival of moderately defended aposematic tadpole prey (the European common toad Bufo bufo) with varying absolute densities in single-species prey systems or varying relative densities in two-species prey systems containing morphologically similar but undefended prey (the European common frog Rana temporaria). The density effects were investigated in conjunction with the hunger levels of the predator, which were manipulated by means of the addition of alternative (non-tadpole) food.

The survival of the B. bufo tadpoles was promoted by increasing their absolute density in the single-species prey systems, increasing their relative density in the two-species prey systems, and providing ample alternative food for the predator. Hungry predators eliminated all R. temporaria individuals regardless of their proportion in the prey community; in treatments with ample alternative food, high relative B. bufo density supported R. temporaria survival.

The results demonstrated that moderately defended prey did benefit from high population densities (both absolute and relative), even under long-term predation pressure. However, the physiological state of the predator was a crucial factor in the survival of moderately defended prey. While the availability of alternative prey in general should promote the spread and maintenance of aposematism, the results indicated that the resemblance between the co-occurring defended and undefended prey may impose mortality costs on the defended model species, even in the absence of actual mimicry.

The experiments were conducted in concrete ponds (6 × 6 m, adjustable water depth set at 75 cm) at the Muchocin experimental station of the Poznań University of Life Sciences (52°37′14.47″ N; 15°50′36.84″ E) from 2016–18. The experimental setup consisted of mesh enclosures (100 x 100 x 100 cm, submerged up to 75 cm) with cube-shaped steel frames for internal support, placed in the ponds. The ponds, with 4 evenly distributed enclosures in each, were individually supplied with water from the neighbouring river (nutrient range values: 0.25–1.02 mg PO4^– L^–1; 0.1–0.5 mg NO3^– L^–1; total nitrogen 2.16–7.37 mg N L^–1) that passed through inlets fitted with fine-mesh screens. The walls of the enclosure (1 mm mesh size) were inserted into the bottom of the basin, giving the tadpoles and fish free access to the sandy substrate. A submerged PVC mesh cylinder (100 cm in length, 10 cm in diameter, 10 mm mesh size) was fastened to the bottom in each enclosure to create habitat complexity and to provide a refuge structure for tadpoles. The enclosures were covered with PVC mesh (10 mm mesh size) on the top to prevent predatory insects from entering. All enclosures were inoculated with zooplankton by adding 7 L of natural pond water.

The study investigated changes in the survival of chemically defended B. bufo tadpoles and in their fitness-related traits (mass at metamorphosis) along a gradient of tadpole density (both absolute and relative), crossed with two levels of availability of alternative food for the predators, with some replication across the gradient. The study consisted of 2 experiments. In both experiments, the response variables were survival of B. bufo tadpoles to metamorphosis and their mass at metamorphosis. In Experiment 1, B. bufo tadpole survival in single-species prey communities with varying absolute density was tested in the presence of a fish predator. Two independent variables were manipulated: B. bufo tadpole density (5/30/40/50/80 individuals per enclosure) and the level of alternative non-tadpole food (feed pellets) provided for the fish. Availability of alternative food was considered as a proxy for hunger level in the predators; ‘high food’ (100 g twice a week) versus ‘low food’ (2 g twice a week). In Experiment 2, B. bufo tadpole survival in two-species prey communities with R. temporaria was tested in the presence of a fish predator, with manipulation of the relative (but not absolute) density of B. bufo. The density of B. bufo tadpoles was kept constant (30 individuals/enclosure), while R. temporaria tadpole density varied (5/10/30/50/60 individuals per enclosure), as did the level of alternative, non-tadpole food provided for the fish (same as in Experiment 1). In both experiments, each enclosure contained one 1-year-old specimen of C. carpio (total length 100–130 mm) as a predator. The experimental design did not include predator-free treatments; this was based on our previous result, i.e. that the mortality of tadpoles at Gosner stage 25 or above was low in the absence of predation (Kaczmarek et al., 2018).

The tadpoles used in the experiment originated from amplexed pairs of B. bufo and R. temporaria collected from ponds in Wielkopolska province (NW Poland) and housed for approximately one week in separate pens, 1 m3, partially submerged in water. The obtained egg masses and hatched larvae remained there until B. bufo tadpoles reached Gosner stage 25. At this stage, the tadpoles were collected (tadpoles of R. temporaria were slightly larger than those of B. bufo as a result of hatching earlier in the season), randomly assigned to treatments, and stocked into experimental enclosures. All anurans used in the study were later returned to their original habitats. The fish used in the experiment originated from semi-natural ponds and had no experience with anuran tadpoles as prey. For one week prior to the experiment, the fish were kept indoors in large fibreglass tanks and provided ad libitum with commercial fish feed (pellets: 35% total protein, 9% crude lipid) that was also later used during the experiment. Individual fish were placed singly in experimental enclosures one day after the enclosures were stocked with tadpoles. Apart from the feed, the fish could also forage on the natural invertebrate prey present in the substrate and in the water column. The mean increases in total fish body length in the low- and high-food treatments during the experiment were measured in 2018, equalling 18% and 31%, respectively. The enclosures were stocked in early May of each year (range: 5–12 May), and the metamorphs were collected after 4–5 weeks. The emerging metamorphs (Gosner stage 46) were collected by dip netting and were counted and weighed to the nearest 0.01 g. The individuals that failed to complete metamorphosis during the experiment were included in the analysis of survival but not that of body mass.

The study was conducted over three consecutive years, in uniform environmental conditions. During each year of the research, Experiment 1 and Experiment 2 were conducted in different ponds. Numbers of replicates were uneven between treatments as well as seasons. Most replicates were performed in 2017 and 2018; four replicates for each experiment were carried out in 2016. Moreover, in 2017 and 2018, marsh frogs Pelophylax ridibundus oviposited in several of the ponds and, as some spawn remained undetected, the freshly hatched tadpoles were able to intrude, entering the enclosures through the mesh walls; the invaded enclosures were excluded from the data analysis. However, we believe that the unbalanced design of the experiments should not greatly affect the results; since we were investigating tadpole survival along density gradients and not across categorical levels (except for the manipulation of alternative non-tadpole food availability for fish predators), we were more interested in obtaining a broad gradient of different densities or proportions of defended/undefended tadpoles than in complete replication of all treatment combinations. The final analyses were performed on a set of 17 (Experiment 1) and 48 (Experiment 2) enclosures.