Quantifying insect predation with predator exclusion cages: the role of prey antipredator behavior as a source of bias Ignacio Castellanos 1 *, Pedro Barbosa 2 , Iriana Zuria 1 & Astrid Caldas 3 1 Centro de Investigaciones Biologicas, Universidad Autonoma del Estado de Hidalgo, Km 4.5 carr. Pachuca-Tulancingo s/n, Mineral de la Reforma, Hidalgo, CP 42184, Mexico, 2 Department of Entomology, University of Maryland, College Park, MD 20742, USA, and 3 Smithsonian Institution, NHB MRC 105, PO Box 37012, Washington, DC 20013-7012, USA Accepted: 24 August 2015 Key words: defensive behavior, predation risk, herbivore, methodology, Orgyia leucostigma, Lepidoptera, Lymantriidae Abstract There are limitations imposed by current methodologies to detect and quantify insect predation. However, there has been relatively little effort to experimentally document the sources of biases asso- ciated with the various methodologies. In this study, we examined how predation estimates in the field using predator exclusion cages may be biased when one fails to account for antipredator behav- ioral responses. To do this, we did the usual comparison of the number of insects missing from plants where predators were allowed access to the number missing from plants where predators were excluded, but also determined how many of the missing insects reacted to predators by dropping from plants and how many were actually preyed upon. Our results provide evidence that estimates of insect mortality in the field are significantly reduced if prey antipredator behavior is taken into account. As it is commonly assumed that prey missing in the field are predated, documenting the incidence of predator-mediated ‘disappearance’ and capturing insect prey before they escape can provide with a relevant estimate of bias. Introduction The detection and quantification of insect predation is frequently difficult due to the small size of prey and invertebrate predators, the concealed habits of prey, and because the interaction between a predator and its prey can be relatively brief. In addition, there often are few if any prey remains that can be used as evi- dence of a predation event. Several methods have been suggested for the estimation of the impact of predators (Luck et al., 1988; Sunderland, 1988; Kidd & Jervis, 2005). One of the most frequently used approaches for studying the impact of predators on insect prey has been predator exclusion (Luck et al., 1988; Medina & Barbosa, 2002; Kidd & Jervis, 2005; Furlong & Zalucki, 2010). Exclusion techniques compare the number of missing insects from plots where predators are allowed access to the number missing from plots where preda- tors are excluded. The difference between the two prey survival values is then reported as an estimate of prey mortality caused by a predator (Luck et al., 1988; Kidd & Jervis, 2005). Several studies have addressed potential disadvantages that can result when insect predation is estimated with predator exclusion techniques (Kiritani & Dempster, 1973; Luck et al., 1988; Kidd & Jervis, 2005). There has been relatively little effort to experimentally document the sources of bias associated with the use of predator exclusion techniques. For example, the use of cages to exclude predators may alter prey behavior, and thus influence estimates of the impact of predation (Kiritani & Dempster, 1973; Luck et al., 1988; Kidd & Jervis, 2005). Prey mortality may be overestimated because the number of confined prey in caged plants (in the preda- tor exclusion treatment) may be abnormally high com- pared to prey densities in open-caged plants [i.e., the predator access treatment, where insects are able to emigrate (Chambers et al., 1983; Sarvary et al., 2007)]. *Correspondence: Ignacio Castellanos, Centro de Investigaciones Biol ogicas, Universidad Aut onoma del Estado de Hidalgo, Km 4.5 carr. Pachuca-Tulancingo s/n, Mineral de la Reforma, Hidalgo, CP 42184, Mexico. E-mails: ignacioe@uaeh.edu.mx, ignacioe.castellanos@gmail.com 360 © 2015 The Netherlands Entomological Society Entomologia Experimentalis et Applicata 157: 360–364, 2015 DOI: 10.1111/eea.12375