J. Avian Biol. 39: 170177, 2008 doi: 10.1111/j.2008.0908-8857.04019.x # 2008 The Authors. J. Compilation # 2008 J. Avian Biol. Received 3 July 2006, accepted 4 July 2007 Hyper-dispersed cache distributions reduce pilferage: a laboratory study Lucinda H. Male and Tom V. Smulders L. H. Male and T. V. Smulders (correspondence), Newcastle University, Centre for Behaviour and Evolution, Henry Wellcome Building, Framlington Place, Newcastle upon Tyne, NE21 4HH, UK. E-mail: tom.smulders@ncl.ac.uk Many animals use hoarding as a long-term strategy to ensure a food supply at times of shortage. We suggest that long- term scatter hoarders, whose caches are vulnerable to potentially high pilferage, should hoard in ways to reduce cache loss. This could be achieved by manipulating the density and dispersal patterns of caches to reduce the foraging efficiency of pilferers. This study explores the effect of distribution patterns on cache loss in the laboratory. We recorded the discovery of food items in different dispersal patterns by two bird species: coal tits Periparus ater (a hoarder) and great tits Parus major (a non-hoarder). Hyper-dispersed distributions reduced foraging efficiency because both species used systematic local search patterns. This study shows that hyper-dispersed distributions would be advantageous to hoarding animals to reduce cache loss. Food hoarding is the handling of food to conserve it for future use (Vander Wall 1990) and many animals use it as a long-term strategy to ensure a food supply at times of shortage e.g. in the winter. However, long-term hoarding can only be adaptive if the hoarder is more likely than other individuals to recover their own caches (Andersson and Krebs 1978). This can be achieved in two ways (ideally used simultaneously): the increase of the probability of the hoarder retrieving the food, and the decrease of the probability of other animals of finding the same food. Much research has focused on how the hoarding individual can increase its own odds of retrieving the items (e.g. Sherry 1984, Brodin 2005). In this study, we focus instead on a mechanism to reduce the depletion of caches by thieves. For animals that cannot defend a single cache location, scattering the hoards around their home range is a strategy that reduces cache loss. For example, the survival of fox squirrel Sciurus niger caches tends to decrease with increas- ing density of caches (Stapanian and Smith 1978, 1984). Marsh tits Poecile palustris also hoard food at an optimal average density that reduces cache loss (Sherry et al. 1982), and willow tit Poecile montana caches in areas with higher cache density disappear more quickly than if the overall cache density is lower (Brodin 1993). But the same average density across a relatively large area can still represent different distribution patterns. For a given average density, an animal may hoard its caches in a clustered distribution (the presence of one point increasing the probability of finding another nearby), a random distribution (the points occurring independently of each other) or a hyper-dispersed distribution (the presence of one point decreasing the probability of finding another nearby; Dale 1999). Because predictability of cache locations may depend on the distribution pattern of the cached items, selection may favour food-hoarding animals’ use of one distribution pattern over others. Here, we investigate whether there is a type of distribu- tion pattern which reduces cache loss. In a recent field study, we showed that more hyper-dispersed cache distribu- tions reduce loss to thieves compared to more clustered distributions (Male and Smulders 2007a). We have ascribed this effect to the fact that many animals’ foraging strategies are adapted to foraging on clustered resources. However, rodents were likely to be the main pilferers in that field study, whereas hoarding parids’ caches are often pilfered by other birds (Brodin and Ekman 1994). Although many animals do use local search patterns tactics, such as area- restricted search (Tinbergen et al. 1967, Stapanian and Smith 1978, Traniello et al. 1991, Benedix 1993, Dejean and Benhamou 1993, Keasar et al. 1996, Withers and Harris 1996, Munyaneza and Obrycki 1998, Hill et al. 2000, Lode 2000, Leising 2001, Leising and Franks 2002), different animals may forage in slightly different ways (e.g. some rodents use olfactory cues (Reichman and Oberstein 1977), and some birds can adapt their searching strategy to the available food distribution (Dall and Cuthill 1997, Dall et al. 1997)). Therefore, it is possible that the cache distribution which reduces loss to rodent pilferers is different form the one that would reduce loss to bird pilferers. To ensure that these results are relevant to hoarding parids and their most likely pilferers, we conducted a 170