A marginal ice zone (MIZ) is characterised by distinct ice floes and its direct exposure to the open ocean. Sea ice is typically described as a continuous material but this description can be inappropriate in an MIZ due to the granular nature of the ice cover and the scale of processes acting on the ice floes. In this thesis, the kinematic behaviour of sea ice in an MIZ modelled as a granular material is investigated with an emphasis on the influence of ocean waves. The kinematic behaviour of a set of ice floes subject to ocean wave forcing was recorded in an experiment conducted in the Ross Sea. Kinematic data were recorded from each ice floe using a GPS receiver, tri-axial accelerometer, and compass. The data show (1) the influence of wave forcing and (2) collisions between neighbouring ice floes. It was also discovered that the GPS receivers were able to resolve the effects of ocean wave forcing despite their poor absolute accuracy. The number density and normalised structure factor (NSF) are introduced to describe the spatial structure of a set of ice floes. Four idealised distributions (in 1D and 2D) are analysed to gain insight into the way that different factors determine the shape of the NSF. It is shown that (1) a significant sinusoidal deviation causes a peak in the NSF, (2) ordered structure dominates low spatial frequencies, and (3) disorder dominates high spatial frequencies. A comparison of the contributions from these different factors is used to estimate the significance of a sinusoidal deviation in the positions of the ice floes. A granular model of an MIZ is developed using a novel set of equations of motion to examine the effect of ocean wave forcing. The equations of motion are derived for small ice floes and allows forcing by multiple waves. These equations predict a transient, wave-induced torque, which can be sustained by the application of a second force to the ice floe. Torque induced by the interaction of two forces on an ice floe may be a general feature of sea ice motion. The number density and NSF are used to analyse the distribution of ice floes in the granular model. At low solids-fractions the number density is correlated at the frequency of the wave forcing. As the solids-fraction is increased this correlation is destroyed by collisions between the ice floes and new correlations are created that are related to the packing structure of the ice floes. When the number density is weighted by the velocity of the ice floes, the correlations between floes are related to the convolution of the wave velocity field and the number density. These correlations may be incorporated into the thickness distribution of large-scale models using the maximum entropy method. The granular model was also examined as a percolating network of contacts and it was found that percolation was more likely to occur along the crest of a wave than in the direction of propagation.