Long food chains are in general chaotic Thilo Gross, Wolfgang Ebenho ¨h and Ulrike Feudel Gross, T., Ebenho ¨h, W. and Feudel, U. 2005. Long food chains are in general chaotic.  / Oikos 109: 135  /144. The question whether chaos exists in nature is much debated. In this paper we prove that chaotic parameter regions exist generically in food chains of length greater than three. While nonchaotic dynamics is also possible, the presence of chaotic parameter regions indicates that chaotic dynamics is likely. We show that the chaotic regions survive even at high exponents of closure. Our results have been obtained using a general food chain model that describes a large class of different food chains. The existence of chaos in models of such generality can be deduced from the presence of certain bifurcations of higher codimension. T. Gross, W. Ebenho ¨h and U. Feudel, Instıtut fu ¨r Chemie und Biologie des Meeres, Carl von Ossietzky Universita ¨t, DE-26111 Oldenburg, Germany (thilo.gross@ physics.org). The long term dynamics of any deterministic system can either be stationary, periodic, quasiperiodic or chaotic. Although ecological models were among the first examples of chaotic dynamics (May 1976) the question whether chaos is an ecological reality remains open (May 1987, Upadhyay et al. 1998, Rai and Schaffer 2001, Cushing et al. 2002). In nature chaos is generally difficult to detect because of the presence of observational noise (Nychka et al. 1992, Ellner and Turchin 1995). Nevertheless, chaos has been found for instance in the dynamics of perennial grasses (Tilman and Wedin 1991), flour beetles (Cushing et al. 1996) and boreal rodents (Hanski et al. 1993). Many other systems seem to be in critical states at the edge of chaos (Turchin and Ellner 2000). From the theoretical point of view population dynamics should be chaotic if chaos is in principle possible in a given system and proves to be advantageous in the evolutionary context. Regarding the effect of chaos on the evolutionary fitness of species two main lines of reasoning exist. On the one hand it is argued that the seemingly random behavior that characterizes chaos can eventually cause the extinction of species (Lande 1993). On the other hand, it has been proved that chaotic fluctuations are desirable in a spatially extended envir- onment (Allen et al. 1993, Sole ´ and Gamarra 1998, Petrovskii et al. 2004). Such fluctuations increase the chance that populations survive periods of detrimental conditions in isolated patches. Starting from these patches the surrounding area can be repopulated once the conditions improve. Following this line of reasoning chaotic dynamics can increase the chances of species survival. Consequently, it is reasonable to expect that ecological systems could evolve towards chaotic regions in parameter space if such regions exist. While chaotic attractors have been found in many models (Hastings and Powell 1991, Boer et al. 1998) they seem to be absent from others. For instance it wasshown by Ruxton and Rohani (1998) that chaotic regions exist in certain models, but disappear if the model structure is perturbed in a certain way. It has often been postulated that chaos would disappear if sufficient biological detail were taken into account (Fussmann and Heber 2002, Kondoh 2003). However, from a dynamical systems point of view one would expect that increasing the complexity of the model increases the complexity of the dynamics as well (May 1973). Consequently, the ques- tion arises whether chaos exists generically in ecological Accepted 4 October 2004 Copyright # OIKOS 2005 ISSN 0030-1299 OIKOS 109: 135 /144, 2005 OIKOS 109:1 (2005) 135