Received 7 April 2004 Accepted 26 August 2004 Published online 7 December 2004 Do wildlife laws work? Species protection and the application of a prey choice model to poaching decisions J. Marcus Rowcliffe , Emmanuel de Merode and Guy Cowlishaw Institute of Zoology, Zoological Society of London, Regent���s Park, London NW1 4RY, UK Legislation for the protection of species is a global conservation tool. However, in many developing coun- tries lack of resources means that effectiveness relies on voluntary compliance, leading to contradictory assumptions. On one hand, laws introduced without effective enforcement mechanisms carry an implicit assumption that voluntary compliance will occur. On the other hand, it is often openly assumed that, without enforcement, there will in fact be no compliance. Which assumption holds has rarely been rigorously tested. Here we show that laws for the protection of some species of large mammal have no effect on the prey choice patterns of primarily commercial hunters in the Democratic Republic of Congo, confirming the second assumption. We established this result by using an optimal diet model to predict the pattern of prey choice in the absence of regulation. Prey choice patterns predicted by the model were accurate across a range of conditions defined by time, space and type of hunting weapon. Given that hunters will not comply voluntarily, the protection of vulnerable species can only take place through effective enforcement, for example by wildlife authorities restricting access to protected areas, or by traditional authorities restricting the sale of protected species in local markets. Keywords: bushmeat wildlife law enforcement human hunters optimum diet model 1. INTRODUCTION Species conservation is typically implemented either through the designation of protected areas or through reg- ulations to protect vulnerable or important species. How- ever, in the developing world, resource constraints greatly limit the ability of governments to implement conservation legislation (Wilkie et al. 1992 Peres & Terborgh 1995 Wilkie et al. 2001). In effect, this means that laws are often implemented with little or no enforcement to back them up, requiring an implicit assumption that resource users will voluntarily modify their behaviour in accordance with the law. In the case of protected area legislation, the avail- able evidence is largely against this assumption, with levels of illegal resource use (Leader-Williams et al. 1990 Milner-Gulland & Leader-Williams 1992 Leader- Williams & Milner-Gulland 1993 Abbot & Mace 1999) and rates of decline in exploited populations (Leader- Williams & Albon 1988) both having been shown to respond strongly to patrol effort in protected areas. There is also some evidence that the rate of offtake of protected species outside protected areas declines with increasing enforcement effort (Milner-Gulland & Clayton 2002). However, the effectiveness of species protection laws in the absence of enforcement has not previously been systemati- cally assessed. We addressed this issue in a region of central Africa where limited hunting is allowed, but where the law prohibits the hunting of some mammal species. Specifi- cally, we test the hypothesis that legal protection influences the likelihood that hunters will target unprotected species in preference to protected species. We used data on the behaviour of hunters while hunting, revealing whether any given encounter with prey led to its pursuit. This kind of information gives us a more powerful test of the above hypothesis than the presence or absence of bushmeat species in local markets or diets, for two reasons. First, if there is a small risk that ignoring protected status could lead to prosecution or harassment, capture of pro- tected species may be prevalent but concealed, making it difficult to detect away from the hunting grounds. Second, protection that is partly effective, thus reducing the rate of capture of protected species rather than eliminating it entirely, cannot be detected without knowing what rate of capture would be expected in the absence of protection. This can only be estimated when there is information on the rate at which hunters encounter these species. There is a further complication, in that failure to pursue a given species may have nothing to do with its protection status, but may instead be driven by the hunters��� own pre- ferences. We therefore need to control for any such pre- ferences if we are to be sure that a given pattern of behaviour is influenced by protection status. In other words, in order to demonstrate whether protection reduces the likelihood of attack, we need to know what the expected likelihood would be in the absence of protection. One way to assess the expected likelihood of attack in the absence of protection would be to gather baseline infor- mation before the imposition of protection laws, before carrying out the study. However, this degree of foresight is rarely possible. An alternative, and the approach taken here, is to use a model of hunter behaviour that can give us the predicted likelihood of attack across a range of species. This assumes optimal behaviour that is uninfluenced by any external force, such as the risk of penalties for catching Author for correspondence (marcus.rowcliffe@ioz.ac.uk). Proc. R. Soc. Lond. B (2004) 271, 2631���2636 2631 # 2004 The Royal Society doi:10.1098/rspb.2004.2915

certain species, and can therefore potentially provide base- line predictions against which the observed pattern of choice can be tested. Optimal prey choice models have been developed and tested, in many cases successfully, for a wide range of animal foragers (Sih & Christensen 2001). They have also been successfully applied to human hunters (Hawkes et al. 1982 Kuchikura 1988 Smith 1991 Alvard 1993), but thus far only in subsistence hunter���gatherer societies the relevance of prey choice models to human hunters operating in a primarily commercial environment has not previously been tested. Here we test the reliability of a prey choice model when applied to commercial hunters in the Democratic Republic of Congo (DRC), and use the model���s predictions as a baseline against which to test the hypothesis that species protection status influences the prey choice patterns of hunters. Data were collected during a period of civil con- flict and during the preceding period of peace, from hun- ters operating illegally within a protected area (Garamba National Park) and those operating legally in a hunting reserve, and from hunters using either shotguns or rifles. We predicted that during the period of conflict, breakdown of civil society would lead to a relaxation of any controls on hunter behaviour that may previously have existed. With regard to location, hunters operating in the protected area were ignoring regulations simply by going there, so we pre- dicted that their prey choice patterns would be less likely to take account of species protection rules than those of hun- ters in the hunting reserve. Similarly, since hunters using automatic rifles are acting illegally simply by possessing these weapons, we would expect their prey choices to be less influenced by species protection status than that of shotgun hunters. In each example, we would expect the case with less compliance to show more closely optimal prey choice than the case with more compliance. These predictions thus provide us with a means to refine our test of the hypothesis set out above. 2. MATERIAL AND METHODS The methods used to collect data on hunting were designed according to the specific hunting practices observed at Kiliwa and Mamba, two villages in the Azande Hunting Reserve, northeast DRC (de Merode 1998). Mamba is located less than 10 km from Garamba National Park, and hunters from this village regularly hunted inside the Park. Kiliwa is about 40 km from the National Park, and hunters from here operated entirely within the hunting reserve. These communities are made up primarily of Azande sub- sistence agriculturalists, with smaller numbers of Logo and Mondo migrants from the east. The Zande economy is pre- dominantly agricultural, but, as in most central African rural cul- tures, their modes of subsistence extend into the surrounding wild lands through hunting, fishing and gathering (de Schlippe �� 1954 Evans-Pritchard 1971). Information was obtained during 1996 and 1997 from 27 hun- ters, over 189 hunts, totalling 568 hunting hours, during which there were 1111 encounters with potential prey, covering 33 spe- cies (31 mammals and 2 birds). Care was taken to monitor an equal number of hunts from each of the hunters in order to avoid a bias towards the results of any particular individual. This was achieved by selecting hunters from a complete list of those under- taking various types of hunting that was established as part of a vil- lage survey prior to the hunting study (de Merode et al. 2004). For each hunt, total time hunting was recorded as the period during which the hunter was actively seeking animals to shoot. The start of the hunt was marked by the hunter preparing the gun, while the end was marked by the hunter stating when the hunt was over. Stalking time was recorded as the time from when an animal was first detected to the point at which either the hunt was aban- doned or the carcass was retrieved. All encounters with potential prey species were recorded, together with the outcome of the encounter, specifically whether they were stalked, and if so, whe- ther the stalk ended in one or more animals being killed. Finally, the weights of carcasses were recorded. In the case of shotgun hunts, the above data were monitored directly. Permission to join the hunts was solicited from the hun- ters, who were consulted several days beforehand in order to agree on a date and time for the researcher to join a hunt. However, it was not possible to participate in hunts involving automatic rifles because the unauthorized possession of automatic weapons in Congo is a criminal offence. As a consequence, data on rifle hunts were collected by interviewing the hunters immediately on their return from a hunt, recording the hunters��� own recollections of encounters, actions and timings, and weighing the carcasses brought home. For predicting expected prey choice patterns, we used a classic optimal diet model (Pulliam 1974 Charnov 1976). The basis of this model is that prey types (in this case species) are added to the chosen set in descending order of profitability, F, until the profita- bility of the next type is less than the current cumulative gain rate, G, including the ith type and all those more profitable: Fi��1 Gi: Prey type i �� 1 and all less profitable types are then excluded from the chosen set. This can be intuitively understood by seeing that the overall gain rate would be reduced by adding this next prey type. Profitability is defined as the total value gained per unit handling time: Fi �� qiVi=Hi, where V is the value of an individual animal, q is the number of individuals killed per prey group, and H is the handling time per prey group. The gain rate is defined as the cumulative value obtained, divided by the total cumulative time spent searching for and handling this catch: Gi �� X i j��1 Vj qj Ej = 1 �� Xi j��1 Hj Ej , where E is encounter rate. Hunters were frequently accompanied by assistants who, in the event of a large animal being killed, processed the carcass and transported the meat, in return for a share of the meat. Further- more, the monetary value of meat per unit weight did not differ between species (de Merode 1998). Species value, V, could there- fore be quantified as the average carcass weight, reduced by the average amount of meat given in payment. For 12 of the 33 spe- cies recorded, no carcass or payment weights were obtained. In these cases, value was given by average body mass, as reported elsewhere (Silva & Downing 1995), minus the expected average payment based on the observed relationship between payment and body mass. The number of individuals killed per group, q, was calculated as the total number of individuals of a given species that were killed divided by the number of encounters with that species in which at least one kill was made. Handling time was defined as the time required to make a kill (including time wasted in unsuccessful attempts). Time costs for 2632 J. M. Rowcliffe and others Species protection and hunter prey choice Proc. R. Soc. Lond. B (2004)