Many animal phyla contain clades in which most or all species are venom-injecting predators. An example, in the arthropods, is the class Chilopoda, containing the approximately 3500 species of centipedes. Very little ecological or behavioural work yielding quantitative data has been conducted on centipede predation. Here, we describe a study of this kind. Our experiments employed one centipede species - a large tropical one, Scolopendra subspinipes mutilans - and two species of prey - a cricket, Gryllus assimilis, and a locust, Schistocerca gregaria. We conducted two experiments. The first was aimed at investigating the extent to which the centipedes attacked prey in particular tagmata as opposed to at random over the whole body surface. The results showed that the centipedes were highly selective, preferring to attack the head or thorax rather than the abdomen; indeed, they often reoriented the prey in order to achieve this. A possible explanation of this behaviour is to maximize the speed with which the neurotoxins in the venom reach either the brain or the thoracic ganglia that control limb movement. The second experiment was aimed at investigating the effect of venom-extraction on the attack rate, and specifically at testing if the magnitude of any such effect differed between the two types of prey, which differ considerably in size. The results showed a major effect of venom extraction in relation to both types of prey, but with the time taken to return to a 'normal' attack rate being longer in the case of the larger prey-type, namely the locust. We discuss these results in relation to the 'venom optimization hypothesis' and, more generally, to the principle of minimizing the production/use of venom, which is an energetically expensive resource. (C) 2012 Elsevier Ltd. All rights reserved.