In the present thesis, we apply computational methods to the study of animal behaviour. Specifically, we are interested in the gradient navigation strategies of C. elegans, for which we show that there are many interesting questions that have not yet been answered by existing research. In order to study the behaviour of C. elegans, we first develop a range of tools to help us do so. We base a large part of our work on Markov-like models of behaviour and since these are not Markovian in the strict sense (limiting the analytical tools which can be used to study their behaviour), we first present a possible transformation from a Markov-like model with variable transition probabilities into a strictly Markovian model. We next present a framework for studying the behaviour of behavioural models which is not restricted to the work presented here but is likely to find general use in behavioural studies. Using these tools, we then analyse the chemotactic behaviour of C. elegans, showing that we can adequately explain most features of this behaviour using energy-efficiency considerations. We also show that the main behavioural strategy, so-called pirouettes is likely to be caused by an inability to sample the environment during a turn and that the animal my not be acting upon gradient information while reversing. Finally, we investigate the deterministic isotherm tracking strategy displayed by C. elegans. We develop a computational model for this behaviour which is able to reproduce all of the main features of C. elegans isotherm tracking and we propose a candidate neural circuit which might encode this strategy. Additionally, we briefly discuss the use of stochastic strategies by the animal when moving towards its preferred temperature. In summary, the work presented here therefore provides contributions to two major fields: we extend the methodology available for behavioural analysis in ethology and we contribute a number of insights and advancements to the field of C. elegans research.