The purpose of the brain is to allow animals to make rapid, complex responses to external circumstances, which requires sensitive and effcient sensory systems. The use that can be made of a sensory modality depends entirely on how the sensory information is encoded by the primary afferents; hence, understanding the early stages of a sensory pathway can help to understand how that sensory information can be used by the brain. In this thesis we show that a simple Generalized Linear Model (GLM) describes stimulus response properties of the primary afferents of the rat whisker system to submillisecond precision, and analyse the parameters to reveal that the encoding mechanism of the system as a whole can be seen as a sparse, overcomplete basis of a low-dimensional projection of the stimulus space. We also show that although the neurons of the Ventro-Posterior Medial (VPM) nucleus of the thalamus can be described well by a GLM, more powerful models, with nonlinear multi-feature stimulus dependencies, are able to predict their responses to a much greater degree of accuracy, and that therefore the form of the information entering the somatosensory cortex and the brain beyond has already undergone some degree of information processing.