Large carnivore science: non-experimental studies are useful, but experiments are better.

Abstract

We thank Bruskotter et al. (2017) for responding to our concerns and engaging with this important issue. We agree completely that nonexperimental studies can and do often have great value, and we recognise that in many (most) cases these types of studies may provide the only data that are available. We acknowledge the many challenges of working on large, cryptic, dangerous, and highly-mobile animals in the wild. However, the absence of more robust data and the reality of these challenges do not excuse weak inference or overstating conclusions – a practice apparent in many studies (and communication of those studies) adopting only observational or correlative methods to infer the roles of large carnivores (reviewed in Allen et al., in press). We advocated in our original article, agree with Bruskotter and colleagues, and reaffirm here, that bringing together studies based on multiple different methods is a powerful way to improve the quality of large carnivore science. But we reaffirm that not all studies are of equal value. Manipulative experiments have far greater inferential power thanobservational andcorrelative studies, whichshould accordingly be valued as ‘weaker’ than manipulative experiments (e.g. Li, 1957; Krebs, 1999; Hone,2007; Fleminget al.,2013). Theneedforsuch experiments may not be as strong where animal numbers are small and more easily observed, study area sizes are small, climates are stable, harvest does not occur, livestock are not present, land use changes are negligible, and past or present human effects are non-existent. In such cases, knowledge obtained from non-experimental studies can be informative. But where these and many other influential factors are present, manipulative experiments can be the only way to tease out the relative effects of all the potential causal factors that may explain our observations. We of course agree with Bruskotter and colleagues that the best situation is when multiple strands of evidence are considered (see also Ford and Goheen, 2015), and we freely recognise that wildlife management decision-making should be informed by more than just scientific knowledge. The challenge lies in the integration of the multiple sources of information, the appropriate weighting or value attached to each, and the way they are used to inform carnivore conservation and management attitudes, policy and practice.