A direct comparison of the effectiveness of two anti-predator strategies under field conditions


  • Editor: Jane Waterman


Aposematism and crypsis are two widespread defensive strategies that have evolved in organisms to reduce attacks by predators. However, although both have been studied extensively, predation rates on unpalatable conspicuous prey have seldom been directly compared to those on palatable cryptic prey, and never in the field. In this study, we use established methods to compare the effectiveness of both defensive traits, by presenting artificial prey targets on trees where they were subject to attack by wild avian predators in a natural field setting. When partially consumed prey and those that had been completely removed were both treated as attacked by predators, there were no differences in attack rates between targets with the two defensive strategies. However, aposematic prey were completely consumed less often than cryptic prey, and partially consumed more often. This suggests that predators engage in taste rejection of unpalatable prey and/or feed on conspicuous prey more cautiously (‘go-slow’ predation). We also observed significant differences in predation among experimental sites, in spite of their similarity and relatively close proximity, and among trials, which suggests that prey may experience highly variable predation in the wild. If aposematic prey are capable of surviving attacks by predators, then this represents a potential defensive benefit of aposematism over crypsis.


Many insects experience high rates of predation in the wild, and because of this, species have evolved a range of defensive strategies to avoid detection and/or deter predators when encountered (Poulton, 1890; Cott, 1940). One way that insects avoid detection is by adopting colour patterns that resemble their backgrounds (Endler, 1984). Another (potentially complementary, see Fraser et al., 2007) strategy is disruptive coloration (Cott, 1940; Cuthill et al., 2005). Disruptively patterned individuals employ contrasting markings to break up their outlines, for instance, by bisecting their bodies with dark lines or breaking up their edges with irregular blotches, thereby hindering recognition (Merilaita & Lind, 2005; Stevens & Merilaita, 2009). Note that the two above camouflage mechanisms are not mutually exclusive, and both may be present in a single-prey individual (Endler, 1984; Merilaita & Lind, 2005).

Nevertheless, not all insects have evolved camouflage as a response to predation. Many insects, including many species of Lepidoptera (Nishida, 2002; Mappes, Marples & Endler, 2005), are aposematic. Aposematism is a defensive strategy in which characteristics that render prey unprofitable to attack (for instance, stings or toxins) are coupled with conspicuous colour patterns (Poulton, 1890). Predators that attack aposematic individuals soon learn to avoid similar-looking prey due to unpleasant or painful secondary defences such as defensive chemicals (Mappes et al., 2005). However, developing chemical defences can be costly (Nishida, 2002; Mappes et al., 2005), and high levels of conspicuousness can potentially lead to aposematic prey experiencing higher attack rates than cryptic prey, especially at low population densities and in the presence of naïve predators (Lindstrom et al., 2001; Ruxton, Speed & Broom, 2009; Marples & Mappes, 2011).

Avian predators are often considered the model receivers when quantifying predation on cryptic and aposematic prey because they are common predators of insects and because they are primarily visual predators, which respond to the colour-based cues involved in both defensive strategies (Endler, 1978, 1981; Cuthill et al., 2005). Many studies have separately quantified the effectiveness of either crypsis or aposematism in reducing predation by wild avian predators (Speed et al., 2000; Cuthill et al., 2005; Stevens et al., 2006; Skelhorn & Rowe, 2009, 2010), and there is some evidence from captive predation studies that aposematic prey experience lower predation rates than cryptic prey (Alatalo & Mappes, 1996), even when both prey types are chemically defended (Sillen-Tullberg, 1985; Halpin, Skelhorn & Rowe, 2008). However, although some studies have directly compared the fates of cryptic and conspicuous prey (Thomas et al., 2004; Saporito et al., 2007; Stevens, Stubbins & Hardman, 2008), to our knowledge, the survival rates of palatable cryptic and unpalatable conspicuous (aposematic) prey have never been directly compared using wild predators under field conditions.

In this study, we modified methods from Cuthill et al. (2005) and Stevens et al. (2006), which used artificial cryptic prey placed on tree trunks to measure predation by wild avian predators, to include aposematic prey (see, e.g. Speed et al., 2000 and Skelhorn & Rowe, 2010). While crypsis and aposematism both vary continuously in terms of their effectiveness in deterring predation (Turner, Kearney & Exton, 1984), the question of relative effectiveness (albeit at arbitrarily low and high values of defence) has important implications for the life histories of organisms that co-evolve with these defences. For example, Blanco & Sherman (2005) found that chemically protected species from a range of taxa had overall higher longevities than unprotected species, and proposed that these observations could be explained by chemically protected species evolving under lower overall extrinsic mortality than unprotected species (see also Hossie et al., 2013). We were interested in testing this assumption, and our expectation was that aposematic prey would experience reduced predation compared to cryptic prey, particularly at high levels of chemical defence.

Materials and methods

Fieldwork was conducted during July and August 2010 in four sites in Gatineau Park, near Gatineau, Quebec, Canada, which were separated by at least 1.7 km (Supporting Information Fig. S1). Prey were made from pastry dough (360 g flour, 210 g lard, 30 g water), which was stapled to tree trunks underneath a triangle of ‘Rite in the Rain®’ waterproof paper (www.riteintherain.com) to simulate wings. In each site, five types of artificial prey were presented. There were two palatable cryptic prey types (with either uniform grey wings or wings with a cryptic colour pattern), two aposematic prey types (with conspicuous wings and different levels of unpalatability), and a white palatable control (Supporting Information Fig. S2). The white palatable control was included to provide a prey target that did not benefit from either crypsis or aposematism as it was both palatable and conspicuous, but lacking typical warning coloration.

To create the high- and low-crypsis targets, reflectance measurements were taken from samples of sugar maple bark (Acer saccharum) using an Ocean Optics S2000 spectrometer (Ocean Optics Inc., Dunedin, FL, USA). Two colours were chosen, which approximated relatively low and high reflectance values within the sample measurements (see Supporting Information Fig. S3 for a comparison of reflectance values between the two colours and sugar maple bark). The wings of the ‘high-crypsis’ targets were produced by printing triangles of ‘Rite in the Rain®’ paper with the two colours in a pattern based on sugar maple bark, which was created by manipulating photographs of sugar maple bark to produce monochrome images, and substituting the two cryptic prey colours. The wings of the ‘low-crypsis’ targets were uniformly printed with the lighter colour. The high-crypsis targets were expected to be more cryptic than the low-crypsis targets because they better matched the background, and were also potentially disruptive because of the presence of edge-intersecting patches (Stevens & Cuthill, 2006). The pastry in both the high and low-crypsis targets was dyed with 1 mL of black Wilton® gel icing colour (http://www.wilton.com/) per 500 g pastry. The wings of the white palatable controls had no colour pattern printed on them, and the pastry (white in colour) was not dyed.

The remaining two prey types were modified to have either a low (0.6 g quinine hydrochloride, 1.2 g ground mustardseed, 0.012 g Bitrex per 500 g pastry) or high (1.5 g quinine hydrochloride, 3 g ground mustardseed, 0.3 g Bitrex per 500 g pastry) level of unpalatability. Quinine hydrochloride has been shown to be aversive to wild avian predators when combined with pastry (Speed et al., 2000), and is chemically similar to quinine compounds found in species of aposematic insects, arachnids and other arthropods (Eisner, Eisner & Siegler, 2005). Quinine compounds are not toxic to birds, but are bitter tasting and elicit an emetic response at high doses (Alcock, 1970). Bitrex is a bitter-tasting chemical that has been shown to elicit an aversive response in birds (Skelhorn & Rowe, 2009, 2010), but is not toxic or emetic even at very large doses (Schafer, Bowles & Hurlbut, 1983), so its only role was to provide an unpleasant or aversive taste to predators. The low and high unpalatability treatments were given conspicuous wings coloured either red or yellow depending on the site, to control for possible pre-existing predator colour preferences. In sites 1 and 2, the prey with a low level of unpalatability were given yellow wings while highly unpalatable prey were given red wings; these colours were reversed in sites 3 and 4. Both types of unpalatable pastry were dyed with 1 mL of orange Wilton® gel icing colour per 500 g pastry.

Trials were conducted for 5 weeks. Each week, one transect was laid in each of the four sites. Each transect contained 12 replicates of the five prey types, for a total of 60 prey items per transect, or 240 per week over all four sites. Individual prey targets were stapled to tree trunks at a height of 2 m, with the paper wings covering the pastry bodies. Only deciduous trees with a diameter greater than 10 cm were used, and trees with prey targets were a minimum of 3 m apart. Transects were left out for 4 days, and prey targets were surveyed at 24, 48, 72 and 96 h for signs of predation by avian predators. Two types of predation events were recorded: the time until the pastry bodies were damaged or partly removed from the prey targets, and the time until complete removal of the pastry bodies. Targets with damaged or partly removed pastry were left on the tree, and subsequent complete removal of the pastry bodies was also recorded. Data was censored if there was evidence of attacks by invertebrates such as ants or slugs, which were detectable through the presence of numerous small bite marks and slime trails, respectively. Targets censored in this way were considered to have survived only until they were damaged by invertebrates, but were not counted as having been attacked by predators. Non-avian predators (chipmunks and squirrels) were also present in the study sites, but we observed beak marks in the pastry bodies of the targets, and small holes and tears in the paper wings, which suggest that avian predators were responsible for much of the observed predation.

Predation was analyzed using a Cox proportional hazards regression (Cox, 1972), which has been used in similar predation studies with censored data and non-uniform predation risk (Cuthill et al., 2005; Cuthill, Hiby & Lloyd, 2006; Stevens et al., 2006). Analyses were conducted using the survival library (Therneau, 2013) in R (R Development Core Team, 2008). Preliminary analyses indicated that hazard rates differed significantly between the four sites used in the study, as well as between each trial (see Fig. 1 and Supporting Information Fig. S4). Given this variability, and because we had no a priori hypotheses regarding the effect of trial or site on predation, the analyses were stratified, which allowed hazard rates to be fitted separately for each trial and site. Defensive strategy was included as a factor in the fitted model, but tree type was not, as the majority of trees used (1053 out of 1200) were sugar maple, and preliminary analyses showed that there was no significant effect of tree type on hazard rates. Overall significance was measured using the Wald test, and pairwise contrasts were used to compare specific treatments.

Figure 1.

Cumulative survival probability for each prey type, separated by experimental site (rows) and predation measure (columns). There were significant differences between sites for all three predation measures.


Predation was assumed to have occurred if either part or all of the pastry was removed from the target. Predation rates over the total 96-h collection period ranged from 33% to 92% (mean ± se: 77 ± 3.7%), and there was no significant effect of defensive treatment on overall predation rates (fit of stratified Cox model: Wald = 6.01, d.f. = 4, P = 0.1985). To differentiate between exploratory attacks and complete consumption by predators, the above two measures were also analyzed separately.

When predation was assumed to have occurred only if the pastry bodies were entirely removed from the targets, there were significant differences in mortality between defensive treatments (Wald = 17.08, d.f. = 4, P = 0.0019). The pastry bodies were entirely removed from highly unpalatable targets at a significantly lower rate than high-crypsis (Wald = 7.99, d.f. = 1, P = 0.005), low-crypsis (Wald = 10.55, d.f. = 1, P = 0.001) and white (Wald = 12.44, d.f. = 1, P < 0.001) targets, but not significantly less than targets with low unpalatability (Wald = 1.97, d.f. = 1, P = 0.161). There were no other significant differences between defensive treatments (all P > 0.7043).

When predation was assumed to have occurred only in the unusual instance of the pastry bodies being partly removed from the targets (and targets with pastry entirely missing were censored by considering them as surviving up until that point, but not attacked by predators), there were again significant differences in mortality between defensive treatments (Wald = 21.38, d.f. = 4, P < 0.001). The pastry bodies were partly removed from highly unpalatable targets significantly more than low-crypsis (Wald = 14.3, d.f. = 1, P < 0.001) and white (Wald = 8.84, d.f. = 1, P = 0.0029) targets, but not significantly more than high-crypsis targets (Wald = 2.18, d.f. = 1, P = 0.1403) or targets with low unpalatability (Wald = 0.01, d.f. = 1, P = 0.95). There were no other significant differences between defensive treatments (all P > 0.1679).


To our knowledge, this is the first time that predation rates on cryptic and aposematic prey have been directly compared in a field setting using wild predators. When prey were considered killed if either part or all of the pastry body was removed, the total mortality rate at the end of the 4-day survey period was high (77%); however, we found no significant difference in survivorship between defensive treatments over the course of our experiment. This was an unexpected result, especially considering that we also found no significant difference in overall survivorship between the different defensive treatments (cryptic and aposematic) and the white palatable control, or between the different cryptic treatments. The white palatable control was assumed to be conspicuous but not warning coloured; however, it is possible that the white coloration was aversive to predators (see Lyytinen et al., 1999). Likewise, we did not observe a significant effect of high-crypsis targets on predation by wild birds compared to low-crypsis targets, even though the presence of edge-intersecting patches made them putatively disruptive (Stevens & Cuthill, 2006), and differences in predation between similar disruptive and monochrome prey have been demonstrated in previous studies (Cuthill et al., 2005; Stevens et al., 2006). One possible reason for this is that the colours on our high-crypsis targets may have had insufficient contrast (Schaefer & Stobbe, 2006), and therefore failed to achieve a disruptive effect.

It is important to note that we observed significant variation in hazard rates between our experimental sites and between trials (Fig. 1 and Supporting Information Fig. S4, respectively), which could have masked differences in the effectiveness of our defensive treatments. Even though the experimental sites contained similar types of trees, and were separated by a maximum of 5 km, there were clear differences in the overall amount of predation and the types of prey that were attacked most often (although many of the differences within each site were non-significant). There could be several reasons for these differences, including predator experience (Skelhorn & Rowe, 2007), availability of alternative prey or even light quality, which can affect target conspicuousness (Endler, 1993). We did not collect quantitative data on potential predator populations in our study sites, but based on personal observations, the most common species included black-capped chickadees (Poecile atricapillus), white-breasted nuthatches (Sitta carolinensis), northern cardinals (Cardinalis cardinalis), American crows (Corvus brachyrhynchos), American robins (Turdus migratorius) and yellow-bellied sapsuckers (Sphyrapicus varius). Whatever the cause, the observed variation between sites and trials clearly reflects the diverse selection pressures that are likely to be experienced by prey in the wild, at least in the short term.

In contrast to the combined analysis, we found a significant effect of defensive treatment on predation rates when prey targets with pastry completely removed and partly removed were considered separately. Previous experiments with wild avian predators have generally considered predation to have occurred if the edible portion of the prey target was either partly or entirely missing (Cuthill et al., 2005, 2006; Schaefer & Stobbe, 2006; Stevens et al., 2006; Rowland et al., 2008). In doing so, no distinction is made between exploratory attacks and complete consumption by predators (but see Hossie & Sherratt, 2012), which is of considerable interest when comparing defensive strategies such as crypsis and aposematism. Indeed, when we analyzed our predation measures separately (i.e. predation was defined as the entire pastry being removed vs. part of the pastry being removed), they produced very different results. Specifically, when attacks were considered only if they resulted in the complete removal of the pastry, highly unpalatable prey experienced significantly less predation than high-crypsis, low-crypsis and white prey, but not significantly less than prey with low unpalatability. Conversely, when only partial removal of the pastry was considered, highly unpalatable prey experienced significantly more predation than low-crypsis and white prey, but not significantly more than high-crypsis prey or prey with low unpalatability. These results suggest that predators may have been sampling highly unpalatable prey at higher rates than cryptic prey and controls, but consuming them at a lower rate. It should be noted that we were not able to distinguish between single and multiple attacks by predators, and indeed, it is possible that completely ‘consumed’ prey targets were simply attacked multiple times by different predators. However, the fact that the observed differences were in the anticipated direction (unpalatable targets consumed less often than palatable ones) suggests that the distinction between partially and completely eaten prey is a valid one.

The lack of significance between prey with high and low levels of unpalatability may indicate that low levels of unpalatability have an intermediate effect on predation. Indeed, it may be profitable for aposematic prey to invest in lower levels of unpalatability in light of the metabolic costs of chemical defences (Nishida, 2002; Mappes et al., 2005). However, the lack of significance in predation rates between low unpalatability prey and cryptic prey suggests that there is a benefit to being more unpalatable, particularly as predators may strategically consume aposematic prey based on factors such as hunger and toxin load (Sherratt, Speed & Ruxton, 2004; Skelhorn & Rowe, 2007). The lack of significance between the two types of unpalatable prey could also have been caused by predators moving between sites, because in two of the sites, the colour treatments were reversed and this may have confused predator learning. However, we consider this unlikely because both conspicuous prey types possessed some level of chemical defence, and the colour treatments were never reversed within a single site.

The differences between complete and partial consumption of cryptic and aposematic baits are readily explained by ‘go-slow’ predation (Guilford, 1994), a strategy in which predators cautiously sample aposematic prey and reject those that are unpalatable without necessarily killing them. This allows predators to avoid the cost of consuming chemically defended prey, while still being able to sample novel or rare conspicuous species; at the same time, aposematic individuals may avoid the disproportionately high mortality rates that are often a consequence of conspicuousness. Go-slow predation may therefore represent a potential defensive advantage of aposematism over crypsis, especially because aposematic insects can survive sampling and rejection by both captive and wild avian predators (Wiklund & Jarvi, 1982; Sillen-Tullberg, 1985; Pinheiro, 1996). Go-slow predation can also help to explain the evolution and spread of novel aposematic species, which has been traditionally considered problematic because of the presence of anti-apostatic (positive frequency dependent) selection (Endler, 1988; Skelhorn & Ruxton, 2007), by providing a benefit for honest signalling (Holen & Svennugsen, 2012). To date, several experiments with captive avian predators have demonstrated the presence of go-slow predation in response to novel aposematic prey (Sillen-Tullberg, 1985; Gamberale-Stille & Guilford, 2004; Skelhorn & Rowe, 2006a,b; Halpin et al., 2008), but as far as we are aware, it has not yet been documented in wild predators.

It is important to note that our results could also have been caused by simple taste-rejection behaviour. Taste rejection differs from go-slow predation in that predators reject prey based solely based on palatability, and do not exhibit cautious attack behaviour when chemically defended prey are conspicuous. This means that taste-rejection behaviour could be exploited by both cryptic unpalatable and aposematic prey. In retrospect, it would have been interesting to include a cryptic unpalatable control in our defensive treatments, to better distinguish between go-slow predation and taste rejection by predators. However, while experiments with captive avian predators have shown that taste rejection occurs with both cryptic and conspicuous chemically defended prey, predators were more likely to learn cautious sampling or outright avoidance when chemical defence was paired with conspicuous coloration (Sillen-Tullberg, 1985; Halpin et al., 2008).

Our results have demonstrated a potential defensive advantage for aposematic prey that is consistent with go-slow predation. Although aposematic and cryptic prey are attacked at similar rates, aposematic prey are consumed less often, indicating that they may be more often rejected by predators after sampling. This could represent an important benefit of aposematism as a defensive strategy, and may have played a role in the evolution of aposematism in the face of significant metabolic and signalling costs.


We thank Dr Innes Cuthill and two anonymous reviewers for their constructive critiques, as well as T. Hossie for providing helpful comments on the paper, and E. Korshikov and J. Kong for their work conducting field experiments. We also thank the Ottawa National Capital Commission (NCC) for permission to work on their land. This experiment was approved by the Carleton University Animal Care Committee and conducted in accordance with research guidelines set out by the Canadian Council on Animal Care. This research was funded through a Natural Science and Engineering Research Council of Canada Discovery grant awarded to T.N.S.