Post provided by Daniela C. Rößler 

© Daniela C. Rößler

Understanding interactions between predators and prey is of interest to a variety of research fields. These interactions not only hold valuable information about ecological dynamics and food webs but are also crucial in understanding the evolution of predatory and anti-predator traits such as vision, visual signals and behavior. Thus, the “who attacks what and why” is key to approach broad evolutionary and ecological questions.

Studying predator-prey interactions, however, is extremely challenging. Even with the use of camera traps, we only rarely observe predation events in nature. In order to improve the collection of that information, about 30 years ago, researchers started using dummy prey (also known as “clay models”) to collect information on attacks in the field. Artificial prey models can be shaped to closely resemble various types of prey including both invertebrate and vertebrate species. The method has since been replicated hundreds of times, being an ethical, powerful, cheap and easy way to collect data on predator-prey interactions. Models, mostly made from non-toxic clay, can easily be manipulated to test a range of questions, for example different visual signal functions in prey. They are placed in the field in large numbers, sometimes in the thousands, and apparent attacks such as bite marks left in the clay are visually inspected and allocated to predator types. This is when the difficult part begins.

Unidentifiable Attacks and Bias

Having used clay models in various studies ourselves, we realized that the interpretation of attack marks is far from easy and oftentimes includes a lot of guess work. A method so powerful and easy suddenly revealed its weak spot – the uncertainty. The uncertainty in identification of marks means this method has a high potential for interpretation bias and a lack of ability to standardize across studies. While what we see in publications are pictures of perfect bite marks (which arguably can clearly be allocated to predator groups) the reality is that the majority of attacks are far less clear. Sometimes what is left in the field are mere crumbles of clay. For these instances it is then either a very vague guess informed e.g. by the estimated force/size necessary to inflict the observed deformation, or these models are not counted in the results because they cannot be identified. In either case, we felt a strong need to find a way towards standardization.

Towards an Unbiased and Standardized Method

Model salamanders are handcrafted artwork. For us, ensuring that clay models resemble the studied species as closely as possible is extremely important when addressing predator-prey interactions. Making these models can be highly meditative but it can also be tough on your tendons; Plasticine clay is pressed into silicone molds and subsequently hand-painted using non-toxic acrylic ink.

With sequencing technology becoming more accessible and more sensitive, and potentially having watched one too many true crime series, we decided to try and “go all forensic” on our clay models. In fact, it started with a paper in a forensics journal describing the successful isolation of salivary DNA of a perpetrator out of bite mark on a body submerged in water. This very much supported our idea, that getting DNA out of the attack marks on clay models should indeed be possible. We planned and conducted a study using over 800 clay models of European fire salamanders (Salamandra salamandra) and deployed them in the field over several weeks with regular checks. We cut out pieces from attack marks and isolated DNA from these. Despite the overall amount of DNA expectedly not being high, we were thrilled to see that the DNA isolation step already showed some positive results. We did initially however co-amplify slug DNA, so we adjusted subsequent primers to suppress slug DNA amplification. The true excitement came after the amplicon sequencing results, when we were presented with a whole range of attackers that were aligned with expected attackers such as red fox and wild boar. Despite the development of a robust protocol, we explicitly wanted to compare the results between a standard visual identification and the sequencing approach for our samples. We thereby uncovered a high rate of misidentifications, which underlines the need of a more standardized identification process.

Prospects of the DNA-clay Method

Our study, being a pilot to test the feasibility of identification via trace DNA, only used presumed mammalian attack marks and mammal-specific primers. Clay models however are attacked by a number of other animals including invertebrates but especially by birds. Thus, the next step will be to design primers that work for more taxa to make the DNA method more comprehensive. In future use of this approach samples should then be analyzed with multiple primers, without prior (subjective) allocation to predator groups.

Using the common visual assessment of attacks allows (if at all) identification of a predator group. Genetic assessment, as presented in our protocol, allows identification down to the species level.

Seeing that bird DNA was already amplified with our primers to some degree, we have no doubt that with specific primers, the protocol will be applicable to birds as well as other taxa in the future. This will be crucial particularly in studies taking place in locations with high biodiversity such as the tropics. Identifying attacks to the species level in such biodiverse habitats would really take clay model studies to a new level. To ensure the feasibility, there will have to be studies examining how long DNA remains intact on the clay under different temperature and humidity conditions.

Lastly, the method allows us to think about possible applications beyond the traditional clay model use. Combining clay models with DNA sequencing could contribute to invasive species monitoring as well as to inventories. The ease with which artificial prey can be manipulated not only in respect to shape and coloration, but also olfaction, can be used to specifically target focal species, providing reliable presence data.


Overall, using our DNA-clay method provides accurate species‐level classification. Beyond that, we could show that the number of attacks that could not be identified visually decreased by almost 50%. For a method that often struggles with overall low attack rates, maximizing the results is a fantastic improvement. However, no method is perfect, and we do think that attention should be paid to ensure ecological validity of future studies. This entails not only aligning the DNA results with the observed attack marks to double check for plausibility, but also to double check that the sequencing results actually mirror natural predators that exist in the area.

To read the full study, check out the Methods in Ecology and Evolution article, ‘An amplicon sequencing protocol for attacker identification from DNA traces left on artificial prey’.