The magnetotelluric (MT) method is a geophysical technique which uses observations of the naturally occurring electromagnetic (EM) field variations to determine the electrical impedance as a function of frequency. Owing to the variable signal and often high level of noise contamination, it is usual to collect large numbers of observations and perform parameter regression in order to gain reliable estimates of this Earth response function. It is then transformed into more interpretable, albeit non-unique, resistivity versus depth models. In the course of this study, the processing, modelling and interpretation of MT observations from Northern England are considered. This includes previously unpublished data as well as a re-examination of spatially more extensive measurements which were acquired during earlier investigations. Existing MT field equipment has been adapted to allow local and remote reference observations to be collected and the results show that the amplitude of the impedance is contaminated by a lower level of noise induced bias. In addition, the resolution of the resulting model of the near structure can be enhanced by supplementary D.C. resistivity measurements. Considerable attention is focused on the development of new methods for processing and modelling MT observations. In particular, the background theory and implementation of a robust and constrained method for the estimation of the impedance is described in some detail. This method is demonstrated on real and synthetic unreferenced observations and shown to improve the consistency and the reliability of the resulting response function by reducing the level of noise contamination. One dimensional models are obtained for 42 MT response functions from Northern England. These are computed using a new 'exploratory' parametric technique which not only aims to find the best fitting model, but also to investigate its range of uniqueness. Additional, more interpretative modelling methods are also developed to improve the consistency of the models from adjacent sites. An experimental procedure which uses the constraining information provided by 2-D gravity model sections is also considered. At a subset of the sites over the Alston Block and the southern margin of the Northumberland Trough a strictly 2-D MT model is derived.