Abstract

Ants are remarkable in the way they form large, highly organized and structured colonies, consisting of up to several million individuals. This is achieved despite the small brains of individual ants and their limited information about the complex states of the colony and its environment, as well as the lack of a central guiding instance. Still, the ant colony is able to act as a whole and collectively finds solutions to the problems that it is challenged with, allowing the survival and propagation of the colony. This remarkable behavior of ant colonies is achieved by means of self-organization, which has become an important research field in biology with ants serving as model organisms. As result, researchers have successfully explained aspects of self-organization in ants by developing models of their collective behavior (Camazine et al. 2001b; Hölldobler and Wilson 2009). In this context, the results of the research group led by Jean Louis Deneubourg from Belgium are most important. Based on the finding that ants respond to trail pheromones in a probabilistic manner, they developed models to explain some fundamental properties of the social behavior of ant colonies. For instance, they explaied how ants are able to collectively find the shorter of two different paths between nest and food (shortest path experiments) or how ant colonies are able to select the higher quality food source if food sources of different quality are available (Beckers et al. 1992; Beckers et al. 1993). We refer to this class of models as the Deneubourg model and to its underlying mathematical function as the Deneubourg choice function (DCF). The DCF has been used by other models that explain different aspects of collective ant behavior, i.e. how they coordinate the division of labor between the members of the colony (Bonabeau et al. 1996), the influence of noise (Dussutour et al. 2009a), the path efficiency in artificial networks (Vittori et al. 2006), the symmetry breaking in foraging behavior (Lanan et al. 2012), the role of multiple pheromones (Dussutour et al. 2009b) and foraging in dynamic environments (Bandeira de Melo and Araújo 2011; Ramsch et al. 2012). Although the Deneubourg model can explain important aspects of the social behavior of ant colonies, it bears some shortcomings: • Computer simulations could not fully reproduce the experimental results of the shortest path experiments, despite the fact that the parameters of the DCF were freely adjusted to get an optimal fit to the experiments. • Realistic parameter values for the DCF were never deduced experimentally. • To our knowledge, the parameters of the DCF lack biological interpretation. • The model does not include a theory of perception describing the relation between pheromone perception and ant behavior. The aim of this dissertation was to overcome these shortcomings by measuring the exact dose-response relationship between pheromone concentration and ant response, deduce all relevant parameter values for the DCF, find a new model based on a theory of perception, test and compare both models and apply the new model to important biological aspects of ant colony behavior. The results presented in this dissertation may be summarized as follows: • The Deneubourg model in its original conception does not explain the experimental results if we use realistic parameter values that were deduced from experiment. • As the missing theory of perception, psychophysical theory proved to be appropriate to describe the dose-response relationship between pheromone concentration and ant response. Most importantly it could be shown that Weber’s law, a fundamental law in sensory physiology, is fulfilled. Important biological parameters like behavioral thresholds and error rates could be defined consistently and experimentally deduced. • By incorporating psychophysical theory into the model, the modified Deneubourg model was able to qualitatively and quantitatively explain the shortest path experiments. • Memory and motivation significantly altered the ants’ behavior and their response to trail pheromones, resulting in the change of characteristic psychophysical parameters. For instance, the response threshold was shifted and the error rate in trail following changed in a way that we predicted. We therefore suggest to refine one of the leading theories of collective ant behavior, the Deneubourg model, by integrating psychophysical theory into the model. Thus we may gain a more exact explanation of the experimental results, a clearer definition of important biological parameters and altogether a deeper insight into the collective behavior of ants.