How do we know eyespots mimic eyes?

Canadian tiger swallowtail (Papilio canadensis) caterpillar in its defensive posture.

Many animals possess large, conspicuous eye-like markings, and the longstanding claim by naturalists is that these spots mimic eyes. The implication that follows is that large eyespots intimidate insect-eating birds from their attack because these birds think they are confronted with one of their own predators. In caterpillars, this inference is often taken a step further, and it is suggested that caterpillars with eyespots are actually mimicking bird-eating or venomous snakes.

A "bird's eye" view of a Canadian tiger swallowtail caterpillar in it's defensive posture.

There are many good reasons to expect that these claims may in fact be true (see here and here), but there has been growing interest in testing the validity of this claim. This work has been spearheaded by a research group in the UK led by Dr. Martin Stevens, who suggests that eyespots may intimidate insect-eating birds simply because many animals avoid attacking prey with large, conspicuous markings. Their research group has conducted some compelling research that illustrates that this may in fact be a plausible alternative (Stevens and Ruxton 2014), despite the claims regarding eye-mimicry made by various researchers. Last year some colleagues and I developed an experiment to test the claim that the eyespots make caterpillars intimidating to birds because of the resemblance to eyes (i.e., eye mimicry), and not simply because they are conspicuous markings.

First, we had to come up with a way to design different kinds of prey items that possessed equally conspicuous markings, but where these markings differed in in how eye-like they were. Instead of doing this by changing the shape or colour of the eyespots themselves, as others before us had done (Stevens et al 2008, Stevens et al 2009), we changed the body position where they eyespots were located. Specifically we created three different kinds of artificial caterpillars:

1. No eyespots (Control)
2. Eyespots towards on end of the body (Eye-like)
3. Eyespots in the center of their body (Less eye-like)

Eyespots are not the only feature which has been implicated in the apparent mimicry of snakes by these caterpillars. Many of these caterpillars also swell their anterior body segments when confronted with a predator. This makes their anterior section look more head-like (at least to human observers). The same logic applies here, if these enlarged segments intimidate birds because it makes the caterpillar appear to be something dangerous like a snake, then it should only be effective when the end of the caterpillar is swollen. In a parallel experiment, our three caterpillar-types here were:

1. No enlarged segments (Control)
2. Enlarged segments at one end (Head-like)
3. Enlarged segments in the center of the body (Not head-like)

We presented these caterpillars to naive domestic chicks who had never seen a caterpillar or a snake in their life. Nine individual chicks were each presented with a single artificial caterpillar from of the 6 types outlined above. We measured two things: 

i) Inspection time: the time until the chick closely approached the prey to examine it.

ii) Latency to attack: the time until the chick pecked the prey item.

Results:

The results couldn't have been clearer. Eyespots and swollen body segments are only intimidating to the chicks when placed in an eye-like or head-like context. When compared to the control prey, chicks delayed their attack longer, and inspected the prey longer, when confronted with artificial caterpillars that has eyespots or a swollen "head". Note however that this was not the case if the swollen segments and the eyespots were not placed in a head-like context.

Have a look at the figure below:

Figure originally published in Skelhorn et al 2014, Behavioral Ecology.

So at least in caterpillars it appears that we have some solid empirical evidence to suggest that these two traits are likely mimetic. Also, both of these traits appear to work in isolation from each other, as we had observed previously in a field experiment. Our results here reinforces another key result from our previous work - that the birds' aversion to these traits appears to be innate, or genetically pre-programmed as Dan Janzen has long suggested.

The full version of our paper can be found here:

Skelhorn, J., Dorrington, G., Hossie, T. J., & Sherratt, T. N. 2014. The position of eyespots and thickened segments influence their protective value to caterpillars. Behavioral Ecology 25:1417–1422

Other recent work:

More recently, there has been research illustrating that eyespots on the wings of butterflies do in fact mimic eyes (De Bona et al 2015). In this paper the researchers presented birds with images of an owl's face with or without eyespots, but also presented birds with three different types of butterfly images:

1. Butterfly with real eyespots (Positive control)
2. Butterfly with modified eyespots (Less eye-like)
3. Butterfly with eyespots digitally removed (Negative control)

They found that real eyespots worked as well as real owl eyes in deterring the birds from their attack. Real eyespots were also more effective than the inverted eyespots, which were equally conspicuous but less eye-like, and images with real eyespots worked better than the no-eyespot control.

De Bona, S., Valkonen, J. K., Lo´pez-Sepulcre, A. & Mappes, J. 2015. Predator mimicry, not conspicuousness, explains the efficacy of butterfly eyespots. Proceedings B: Biological Sciences 282: 20150202

Body size affects the evolution of eyespots in caterpillars

Eyespots are an incredibly widespread strategy used by animals to protect themselves from predators. There are examples from all over the world, and from many different animal groups. Although they are well-known from the wings of butterflies and moths they are perhaps even more common in caterpillars. In fact, there are examples of caterpillars with eyespots on every continent except Antarctica, and from many different families of butterfly and moth.

Examples of caterpillars with eyespots from various places across the globe

Conventional wisdom has long suggested that eyespots help deter attacking predator by resembling the eyes of the predator's own enemies (see this recent paper on butterfly eyespots!). In caterpillars, the claim is often taken one step further: caterpillars with eyespots are generally thought to mimic dangerous snakes that might attack a hungry insect-eating bird. My research has set out to test this claim and to better understand how mimicry in this system operates.

After 4 years of research on the topic we have convincing evidence that 1) eyespots can protect caterpillars from insect eating birds under natural field conditions (see posts here and here), 2) eyespots are not simply conspicuous signals, but probably do in fact mimic eyes (see paper here), and 3) that many of these caterpillars engage in behavioural mimicry to enhance protection (here and here). With this baseline knowledge in hand my co-authors and I set out to ask bigger questions. Why do some caterpillars have eyespots, but not others? What drove the evolution of eyespots in some caterpillars, while other species evolved different defenses like camouflage or aposematism?

Generating Hypotheses:

After familiarizing myself with the various species of caterpillars with eyespots (you can do this for yourself here!), it became clear that many of these species are relatively large. In addition, eyespots tend not to occur small early instars of these caterpillars, but only show up in the larger late instars. For example, the Canadian Tiger Swallowtail caterpillar shifts its strategy from resembling a bird dropping when small, to having eyespots when they get bigger.

Canadian Tiger Swallowtail (Papilio canadensis) progressing through instars. Photos: T. Hossie.

Using this set of observations we came up with a few reasons why this might be the case:

1. Being large makes it harder to hide, so big caterpillars need to find another way to escape predation. Mimicking a dangerous snake offers a solution.
2. Large prey are a bigger meal for insect-eating birds, meaning that these predators have a vested interest in seeking out large caterpillars and eating them upon discovery. If this is the case, large prey need better protection from these highly motivated predators and have to do something extreme to dissuade their attackers - like mimic one of the bird's own predators!
3. Effective mimicry might depend on body size. Maybe only large caterpillars are big enough to convincingly resemble a legitimate threat.

Validating Our Initial Observation:

Now it was time to do some science! First we had to confirm our original inkling that large-bodied caterpillars are in fact more likely to evolve eyespots. To do this we assessed the presence/absence of eyespots from photos and recorded maximum body size in the final caterpillar instar. We did this for as many species as possible for one well studied group of moths (Sphingidae). Next we tested if larger species were more likely to have eyespots. Since many of these species are closely related and share similar traits (e.g. size, or presence/absence of eyespots) not because they each evolved them separately, but instead because they descended from a common ancestor we used a "phylogenetic comparative analysis". This enabled us to statistically "remove" this effect of shared ancestry and see if the relationship between body size and eyespots remains.

Our full list of species from our analysis and their shared evolutionary history (left). I have highlighted a sub-group (right) to illustrate the importance of controlling for shared ancestry when you are using species-level data to look for a relationship between two traits. Note that the species highlighted on the right are more likely to be both large and have eyespots than you expect by chance alone because they descend from a species which was large and had eyespots.

Here I have highlighted a different potion of the phylogeny. Clearly there is less of an effect of common ancestry on the relationship between body size and eyespots, and it appears more likely that eyespots appear evolved multiple times independently. Interestingly, although Hemeroplanes triptolemus and Madoryx plutonius are closely related they probably evolved eyespots independently because these markings occur on different parts of therr bodies.

Our analysis confirmed our original observation that eyespots tend to occur on caterpillar from species that attain larger body sizes (even after controlling for shared ancestry), suggesting an evolutionary relationship between these traits.

It was now time to conduct experiments to determine the mechanism. Broadly speaking, two processes could be driving this relationship: i) relatively low survival of small caterpillars with eyespots (e.g., compared to small caterpillars with no eyespots), and ii) relatively high survival of large caterpillars with eyespots (e.g. compared to large no-eyespot caterpillars). Of course both processes could also be happening concurrently. We set out to look for these patterns using both a field experiment and a lab experiment. Both experiments used "pastry" caterpillars (fake caterpillars made of flour and lard, then dyed with food coloring), as we have done previously (here, here, and here).

Identifying the Mechanism: Field Experiment

In this experiment we created four different prey types arranged following a 2x2 design:

• Small-bodied caterpillars without eyespots 
• Small-bodied caterpillars with eyespots 
• Large-bodied caterpillars without eyespots 
• Large-bodied caterpillars with eyespots 

For the prey with eyespots we hand-pained eyespots onto the prey. We pinned one of each caterpillar type to the branches on 96 trees in the forest and let wild birds attack them over a period of 90h. We checked up on these caterpillars 3 times a day and recorded any prey that had peck marks or were missing - these were considered "killed". Over the course of the summer, this process was repeated 6 times in different areas for a total of 576 prey deployed and tracked!

One of our artificial caterpillars (made of flour and lard) deployed in the field. 

We found that small prey with no eyespots had the highest survival, probably because they were hardest to find. Interestingly, adding eyespots to these prey substantially reduced survival. We think that adding eyespots made these small prey easier to find, but didn't make them intimidating to the attacking birds. Large prey with or without eyespots had about the same survival, though both had lower survival than small prey without eyespots. This seems to indicate that adding eyespots didn't make large caterpillars any easier to find, perhaps because large-bodied prey are already highly conspicuous. We were a bit surprised that eyespots didn't reduce survival in large prey (as we have seen in previous studies here and here), but some additional analyses showed that eyespots benefit large prey in some microhabitats more than others (a result we plan to follow up with future research). Either way this experiment convincingly shows that the effect of eyespots on survival depends on the caterpillars body size - eyespots are strongly detrimental to small prey and are either neutral or protective in large-bodied prey.

Identifying the Mechanism: Lab Experiment

One possible weakness of our field experiment is that we scaled eyespot size with body size, meaning that large-eyespot prey had bigger eyespots than small-eyespot prey in absolute terms. We chose to do this because the in real caterpillars with eyespots these markings scale proportionally as they grow. However, the design of our field experiment leaves a possible confound - it could be that eyespot size is what's really driving the difference in how effective these marking are, not the relationship between body size and eyespots as we had proposed*. If this was the case, a caterpillar would have to be relativity large to possess eyespots big enough to deter birds, and small caterpillars would be constrained to having markings too small to be effective. Maybe the attacking birds cue into eyespot size as an indicator of whether the animal they are examining is one of their predators or a prey item.

Given this possibility we had to design an experiment where we could disentangle the effects of eyespot size from the relationship between body size and presence of eyespots. We also wanted to confirm the results of our field experiment using a distinct system and environment which would allow us to make a more detailed assessment of how predators react to our experimental caterpillars. Our lab experiment consisted of 6 different prey types:

• Small-bodied caterpillars without eyespots 
• Small-bodied caterpillars with small eyespots
• Small-bodied caterpillars with large eyespots
• Large-bodied caterpillars without eyespots 
• Large-bodied caterpillars with small eyespots
• Large-bodied caterpillars with large eyespots 

Our model predators for this experiment were naive domestic chicks. These are baby chickens that have never seen a caterpillar or snake (or other predators) before in their life. Each chick was placed into an arena where a single pastry caterpillar from one of the 6 treatments pinned to a tree branch. We recorded the time it took the chick to go up and inspect the prey ("time to first inspection"), and the time from first inspection to attack ("inspection time"). The total time from the beginning of the trial to attack (i.e., the sum of time to first inspection and inspection time) is the overall latency to attack, which is a measure of survival. We also recorded "approach-retreat" behaviour, when chicks quickly retreated from the prey upon inspection.

A schematic of our lab experiment set up. Chicks become stressed if fully isolated for long periods of time, so we included "buddy" chicks in a separate compartment (isolated but visible) so the experimental chick would feed and behave normally. 

Here we found that small caterpillars without eyespots took a long time to find, but small caterpillars with  small or big eyespots were found and eaten right away. In general, chicks were hesitant to approach large caterpillars, and once approached they were inspected longer prior to attack than any of the small prey. Interestingly, after the first inspection large caterpillars with small eyespots were inspected for about the same amount of time as caterpillars without eyespots, but large caterpillars with large eyespots were inspected much longer than any other treatment. In addition the approach-retreat behaviour was almost exclusively observed in chicks that were presented with large caterpillars with large eyespots.

In general these results tell us that in order for eyespots to effectively deter birds from attacking caterpillars both body size and eyespots probably need to be large. Eyespots (big or small) do not appear to protect small caterpillars which strongly suggests that the effect of eyespots on survival depends on both body size, and not simply on eyespot size. In large-bodied caterpillars small eyespots do not enhance survival, only large eyespots do. This seems to suggest that for caterpillars to deter their attackers they require large body size and large eyespots.

Conclusions & Summary:

In small caterpillars, eyespots reveal their location to their predators but fail to intimidate those attackers regardless of eyespot size. On the other hand, large caterpillars are already so conspicuous that adding eyespots does little to make them easier to find. In addition, large caterpillars can benefit from large eyespots which deter attacking birds (at least in some microhabitats). The lab experiment also shows that even chicks which had never seen a snake before were intimidated from attacking large caterpillars with large eyespots. The aversion to these caterpillars therefore appears to be innate, that is birds may be genetically pre-programmed to fear things that look like the eye or face of a dangerous predator (as suggested by Dan Janzen).

Caveats & Future Work:

Another important point to make is that eyespots are not the only solution to the problem of being a large and conspicuous caterpillar. For example the Frangipani hornworm (Pseudosphinx tetrio) caterpillars are large and poisonous, and may even be coral snake mimics!

Frangipani hornworm (Pseudosphinx tetrio) caterpillar. A large caterpillar which can attain body sizes up to 11.3 cm. Instead of having eyespots this species appears to be aposematic, and may even mimic coral snakes.
Photo by: KoS. Licensed under CC BY-SA 3.0 via Wikimedia Commons

Eyespots are also common in other families where caterpillars tend to be much smaller (e.g., Hesperiidae). These small caterpillars often live and feed inside leaf shelters for most of their life prior to metamorphosis. Leaf shelters might make it more difficult for attacking birds to estimate the caterpillar's body size (i.e., by obscuring much of its body) and force birds to flee from the possible threat instead of gamble with their life. In our research described above we found that eyespots were generally detrimental to small prey because they made them easier to find, but perhaps by inhabiting leaf shelters these small prey are able to minimize their conspicuousness (even with eyespots) while simultaneously maximizing the “startle” effect of the markings when they pop out of a leaf shelter besieged by an insect-eating bird. I like to think of this as a "Jack-in-the-box" effect, and this is currently being examined.

A 5 cm-long final instar caterpillar Ridens panche (Hesperiidae) peeking out of its leaf shelter.
Photo from Fig. 2 of Janzen et al 2010.

Our manuscript is now available in Early View. To read about this research in more detail click here to get the full-text version of the scientific article. If you are unable to get a copy, please feel free to contact me directly for a PDF reprint.


Citation:

Hossie, T.J., Skelhorn, J., Breinholt, J.W., Kawahara, A.W. and T.N. Sherratt. (2015) Body size affects the evolution of eyespots in caterpillars. Proceedings of the National Academy of Sciences Early View

*The eyespot size confound was brought to our attention by some thoughtful reviewers. In fact the prey types in our original lab experiment were identical to our field experiment. The reviewers felt that this confound was a serious weakness and suggested that it needed to be addressed prior to publication. Luckily we were able to revise the design and run the new experiment relatively quickly. This new experiment confirmed and extend the results of our original lab experiment and I think makes the paper substantially stronger. This is a great example of rigorous and constructive peer-review helping to improve the quality of published research.