+ Carnivore Coexistence Lab
CCL banner

Kill to conserve is a myth

Just as the hydra sprouted two heads where Hercules had chopped off one, some ideas multiply and regrow after decapitation. The idea that people need to kill a few animals to protect the rest (kill to conserve) seems to be such an idea. First called ‘hunt to conserve’, the idea has grown into newer notions sometimes called, ‘blood buys goodwill’ or ‘tolerance hunting’. The underlying idea has deep roots and seems to reincarnate in a new form even if it fails a test. Here we offer opposed hypotheses for why the idea of ‘kill to conserve’ re-emerges in new guises, but first, we examine the idea stripped down to is essentials, and second, we present three parallel lines of evidence from wolves that refute the newest variants sprouting from the idea.

The title above might seem to have an obvious answer, but the U.S. government believes legal killing will help against illegal killing of endangered wolves and grizzly bears. They argued that in federal court in 2006 and lost, yet they continue to propose liberalized killing for grizzly bears and wolves, as a way to improve tolerance and reduce poaching. Neither the social science we have published from 2003–2015 about people’s attitudes and inclinations to poach, nor the population ecology and survival analyses we have presented 2016-2020 seem to have swayed this deeply held myth.

Below we review the evidence, start with wolf survival, then wolf population dynamics, then human tolerance and inclinations to poach wolves.

A long-held assumption has been that some predator control (e.g., special permits for killing or hunting seasons) might increase tolerance for controversial species and thus reduce poaching; a claim first argued in a legal brief by the U.S. government in 2006 (Humane Society of US v. Kempthorne, docket DC 06-1279) and articulated by us as a scientific hypothesis in 2016, and later developed and renamed ‘tolerance hunting’.

In Wisconsin, USA a series of studies have taken up the question using mortality data or population dynamics. One early study examined reported poaching variation to hypothesize that frustration with inconsistent management may lead to increased poaching, and colleagues modeled wolf demographic parameters in relation to policy changes. However, these studies provide weak inference due to several shortcomings: reliance on correlative analyses, failure to consider cryptic poaching, plus unresolved concerns about modeling of density-dependence and its potential confounding effects of various changes in monitoring methods entangled with so-called ‘recovery periods’.

A parallel and independent analysis of the Wisconsin and Michigan wolf populations found that periods with policy that liberalized wolf-killing were followed by significant decreases in potential population growth rates independent of the number of wolves killed legally. We inferred increases in poaching during six periods of policy that liberalized wolf-killing had caused several decreases in growth rates and that the resumption of more protective policies caused several increases in growth rates. These authors suggested what we now call the ‘facilitated poaching’ hypothesis, which proposes that would-be poachers respond to the policy changes as a signal to increase their activities, possibly associated with cognitive processes relating to values (e.g., lower value of wolves in the eyes of would-be poachers), social norms (e.g., greater acceptability of poaching, or less enforcement against poaching), or perceived control (e.g., would-be poachers perceive themselves helping authorities to kill wolves). This hypothesis is supported by our research on people: four quantitative surveys of residents of Wisconsin from 2001 to 2013 and two qualitative focus groups from 2011 to 2012, which revealed increased inclinations to poach after Wisconsin wolf policies liberalized killing.

Three critiques of the ‘facilitated poaching’ hypothesis were published, and one critique of the ‘frustration’ hypothesis, so the scientific debate is lively but it remains based on indirect evidence and weak to moderate strengths of inference… until 2020 as we explain next.

The latest articles on the negative side-effects of liberalizing wolf-killing

Santiago-Ávila, F.J., Chappell, R.J., Treves, A., 2020. Liberalizing the killing of endangered wolves was associated with more disappearances of collared individuals in Wisconsin, USA. Scientific Reports 10:13881.

Brief summary of our 2020 findings:

In six periods 1979-2012 when the U.S. federal government relaxed or removed Endangered Species Act (ESA) protections for gray wolves (Canis lupus) in Wisconsin, radio-collared wolves disappeared at higher rates than during periods of strict ESA protections. These radio-collared wolf disappearances occurred at rates similar to the rate of intentional legal, lethal control. We dismiss migration and collar failure as implausible explanations and instead find further support for the idea that the proportion of cryptically poached wolves increased. We speculate that would-be poachers value wolves even less during periods of lower ESA protection or believe they are helping the government control over-population of wolves.

Until now, no study explicitly modeled the durations and periods of policy that individual wolves were exposed to legal killing. That would allow for the estimation of mortality hazard and incidence (from various causes) for individual wolves that experienced the policy over time. With individual-level estimates of hazard and incidence for marked individual wolves, we can more confidently draw inference about population-level effects on the growth rate and patterns of poaching.

Nor did these studies model the effect of legal killing on wolf disappearances (those animals ‘lost-to-follow-up’; LTF). LTF animals could not have been killed by legal means or by conspicuous causes, otherwise their carcasses would have been recovered. Thus, LTF could conceal a component of cryptic poaching, in addition to those collared individuals that moved out of radio-telemetry range or those who died from natural causes but whose radio-transmitters suffered mechanical failure beforehand.

Stay tuned for upcoming results on Mexican wolves and red wolves.

The evidence from population ecology


Chapron, G. and Treves, A. 2016a and b, 2017a and b.

We first showed that Michigan and Wisconsin wolf population growth slowed whenever the government liberalized wolf-killing and the slow-down was proportional to the length of time that culling was liberalized, regardless of how many wolves were killed. In 2016, we published a correction

that showed the probability that culling slowed population growth was 92%. In 2017, Pepin et al. started a lively debate when they tried to counter our hypothesis but did not succeed in our opinion. That debate improved our model which strengthened its findings, also in the pages of Proceedings of the Royal Society B. Pepin provided no evidence for a biological mechanism for density-dependent population growth as we showed in a 2017 rebuttal.

Then Stien 2017, and Olson et al. 2017 tried again. Those critiques only made our evidence stronger.

Olson et al. in particular weakened their own position by omitting methods and citing studies that were irrelevant. Furthermore, Chapron and Treves were pleased to hear of independent findings for Mexican wolves presented by David Parsons in 2014. Soon to come out in 2019, new findings by Santiago-Ávila are sure to revive the debate!

The evidence from human tolerance and inclination to poach

Browne-Nuñez et al. 2015 were the first to test this with a quasi-experimental design by conducting focus groups before and after a change in policy allowed legal killing of wolves in Wisconsin. Although the focus groups were qualitative, an anonymous survey handed out before each focus group allowed quantification of intentions to poach also. The qualitative portion of the focus groups indicated an equivalent or increased call for all methods of killing wolves after the policy legalizing killing was implemented. The quantitative portion of that study suggested increased inclinations to poach wolves following that policy.

Then Hogberg et al. 2015 showed that the same individuals resampled between 2009 and 2013 after the major change in policy was delisting wolves and allowing both legal, lethal control and a public hunting, trapping and hounding season was associated with a decline in tolerance for wolves among men residing in wolf range. That sample had been and continued to be dominated by those with experiences of hunting wildlife.

Looking back to Treves et al. 2013, — our first longitudinal study of attitudes in the same individuals over time as policies changed and other events transpired — we find that tolerance for wolves declined and inclinations to poach wolves increased from 2001/2004 (our first two panels of respondents) and 2009 when we resampled members of both panels. The main change in those periods was continued increase it he wolf population and several periods of liberalized lethal control of wolves. That paper showed no significant effect of individual experience with wolves on attitudes or inclination to poach.


2014. Treves, A. Bruskotter, J.T. Tolerance for predatory wildlife. Science 344: 476-477.


Take-home messages: Tolerance for predators did not reflect individual economic losses but rather social identity and peer group complaints. Intolerance for wolves and inclinations to poach wolves rose when the government culled wolves. Tolerance for wolves did not increase when wolf hunting and trapping season was implemented.


2015. Hogberg, J., Treves, A., Shaw, B., Naughton-Treves, L. Changes in attitudes toward wolves before and after an inaugural public hunting and trapping season: early evidence from Wisconsin’s wolf range. Environmental Conservation, doi 10.1017/S037689291500017X.


2015. Browne-Nuñez, C., Treves, A., Macfarland, D., Voyles, Z., Turng, C. Tolerance of wolves in Wisconsin: A mixed-methods examination of policy effects on attitudes and behavioral inclinations. Biological Conservation 189: 59-71.


Lethal control does not seem to protect domestic animals from predation by large carnivores

Despite the importance of preserving nature and protecting human safety and other interests, and despite the confluence of interests that share common ground that such interventions should be non-lethal if we wish to maximize good outcomes for all, the scientific study of methods to prevent animal threats to human interests faces numerous challenges.

Recent reviews emphasize the scarcity of strong inference about the effectiveness of diverse methods for preventing predation on livestock (van Eeden et al. 2018; Treves et al. 2016, 2019), or for methods to prevent bear (Ursus spp.) damage to those and other human properties (Khorozyan et al. 2020),. Even for the well-studied systems of wild carnivores threatening human property including livestock, we lack systematic evidence for the durations of many preventive effects (Khorozyan et al. 2019). Even for the longer-studied subfield of protecting wild game, scientists struggle to account for and manage the variability of the effectiveness of killing wild predators (Reynolds & Tapper 196, Laundré & Papouchis 2020; Clark & Hebblewhite in review).

The scarcity of evidence is compounded by difficulties in completing the gold-standard of randomized controlled experiments in the field or inattention to the ubiquity of scientific biases. For example, our 2016 review set aside more than a dozen peer-reviewed attempts because they had flaws in design that rendered them unreliable (e.g., Wagner & Conover 1999; Mitchell et al. 2004; for review see Treves et al. 2016). In our view, that Wagner & Conover study is the most likely candidate for retraction because the methods are incomplete or contradicted by later federal court testimony by the lead author (Western Watersheds Project et al. v USDA Wildlife Services. 2018. U.S. District Court Idaho. 1:17-cv-00206-BLW Document 22-3). Such reporting biases are not unique to that study, it is simply the most egregious example we have found (Treves et al. 2016).

Furthermore, the scientific community is divided on what constitutes good evidence for wild predator control (Miller et al. 2016, Treves et al. 2016, 2019, van Eeden et al. 2018a,b, Eklund et al. 2017; Moreira-Arce et al. 2018; Lennox et al. 2019; Khorozyan et al. 2019,2020; Santiago-Ávila et al. 2018, 2019; Bradley et al. 2015; Wielgus & Peebles 2014, Poudyal et al. 2016; Kompaniyets & Evans 2017).

Many of these scientific debates have become embroiled in partisan bickering over values, which divert time and resources away from the fundamental challenge that strong scientific inference is hard to come by.

>p? Against this background of scarce and variable experimental evidence we find other challenges to drawing strong inference about effective methods of preventing animal threats to human interests.

We largely lack tests of the same methods under different biophysical conditions (Eklund et al. 2017). Although randomized, controlled trials have been published for four non-lethal methods for the protection of livestock on working farms from wild predators (livestock guarding dogs LGD, fladry, Foxlights®, and eye-spots: Davidson-Nelson et al. 2010; Gehring et al. 2010; Ohrens et al. 2019; Radford et al. in press) and a few lethal methods tested on feral predators (e.g., Greentree et al. 2000; Allen & Sparkes 2001), none of the methods has heretofore been subjected to more than a single such gold-standard experiment without bias. Instead, for replication one must turn to captive conditions (fladry: Young et al. 2019), experiments without livestock (fladry and siren and light deterrents: Shivik et al. 2003), or lower standards of inference, such as correlations and silver-standard before-and-after comparisons without randomized controls (e.g., Coppinger 1988; Musiani et al. 2003; Stone et al. 2017). In all the reviews since 2016, we have never found a replication of a gold-standard, randomized, controlled experiment without bias and conducted on domestic animals and non-captive, non-feral wild predators..

Replication efforts are crucial in many scientific disciplines because of a much-studied reproducibility crisis whereby seemingly reliable findings fail to reproduce when independent teams using the same methods fail to find as strong effects or any effects at all. When replication experiments are attempted under slightly different conditions, we also gain insight into the limits of inference from the first experiments and further insight into variability in conditions that weakens or strengthens scientific results. We are presently writing up one such effort at technical replication for Foxlights®.

Nevertheless, we can conclude about the state of the science for preventing predation on livestock. At present, the evidence is better for non-lethal methods and they seem on average more effective at protecting livestock (Treves et al. 2016; van Eeden et al. 2018, Khorozyan et al. 2020). Furthermore lethal methods pose a risk of counter-productive increases in livestock loss, detected in two studies in Europe and several in the USA on wolves (Fernandez-Gil et al. 2016; ImBert et al. 2016; Santiago-Ávila et al. 2018), and studies of recreational hunting of cougars in Washington state (Peebles et al. 2013) and beyond (Laundré & Papouchis 2020).

The picture for bears is considerably less clear. Garshelis et al. 2020 report public hunting The picture for bears is considerably less clear. Garshelis et al. 2020 report public hunting helped to reduce conflicts with people whereas Obbard et al. 2014 reported the opposite and in a worldwide review Khorozyan et al. 2020 reported inconsistent effects that were weaker than the most effective non-lethal methods. Considering the poor record of experimental design in government agency interventions that novel predator control (Treves et al. 2016; Clark & Hebblewhite in review), we recommend policy-makers place a moratorium on lethal predator control until the gold-standard experiments are done as has been recommended since 2005 (Treves & Naughton-Treves 2005).

Many of the works cited above can be found for free below or elsewhere in this website. For others, email atreves at wisc.edu for full citations or copies of the articles.


Treves, A., Krofel M, Ohrens O, and van Eeden 2019 Predator Control Needs a Standard of Unbiased Randomized Experiments with Cross-Over Design. Frontiers in Ecology and Evolution 7:402-413. doi: 10.3389/fevo.2019.00462

Ohrens, O., Bonacic, C., Treves, A. 2019. Non-lethal defense of livestock against predators: Flashing lights deter puma attacks in Chile. Frontiers in Ecology and the Environment 17(1):1-7. Gold standard (also platinum and silver standards!) experiments are explained at our new web page about

standards of evidence in animal research.


Treves, A., Krofel, M., McManus, J. (equal co-authors).2016. Predator control should not be a shot in the dark. Frontiers in Ecology and the Environment14: 380-388.


short video explaining the findings

In a nutshell: