This thesis investigates the effects of hydrodynamic interactions within arrays of wave energy converters, provides a significant amount of much needed physical model data to the industry for validation purposes and assesses the relative suitability of various numerical models for use in isolated device and array studies. Experience gained from two large sets of physical experiments undertaken early in the project resulted in the planning, preparation, design and execution of a third set of experiments, the results of which have been reported in this thesis. The physical data obtained allowed for the estimation of the effect of array interaction on device response and performance under a wide range of incident, control and array setup conditions. To undertake these experiments novel testing protocols and procedures were developed to allow for the production of useful data where the quantity of interest was much smaller than any particular measured value, It was found that even small errors and uncertainties may have significant implications for data obtained and that the error and uncertainty inherent within a physical system is capable of misrepresenting array interactions. In outline, it has been found that investigating the effects of array interaction experimentally is difficult and requires significant attention to detail. Two different Frequency Domain Models and a Time Domain Model have been produced for modelling the effects of array interactions in regular sea states. A Spectral Domain Model and Time Domain Model were produced to estimate array interactions in irregular sea states. Results obtained from these models have been compared to those obtained from physical testing and by estimating the error in physical results the validity of a variety of industry standard numerical models has been assessed for the first time. The importance of distinguishing between model accuracy, adequacy and suitability is highlighted. It is also argued that in many cases, Time Domain Models will not provide the most accurate results achievable and that Spectral Domain Models provide a suitable alternative in these cases.