The aim of this research was to determine whether tunnelling machine noise could be used to predict ground conditions ahead of an advancing tunnel face and also, if other types of machine noise could be used to detect subterranean cavities using the techniques of spectral analysis. The initial study was a feasibility trial of a tunnelling machine at a site in Warrington, Cheshire. The data was recorded using a single -channel and then a four-channel F.M. system, allowing simultaneous recording of the output from three geophones. Subsequent spectral analysis of this data showed that the frequency spectrum of energy imparted to the ground by the tunnelling machine was not stationary, thus, precluding the use of quantitative analysis to determine ground conditions. The next field experiment, at a site near Cocking Village, West Sussex, was a study of the effects of a large cavity on the characteristic spectra of a continuous-wave source operated at the surface of the ground. The source used was a petrol-dri ven soilcompaction tool, operated near a disused railway tunnel which acted as the cavity. Spectral analysis showed anomalous attenuation caused by the disturbed ground around the tunnel. Cavity resonance was not observed and correlation and phase differencing techniques failed to provide information about seismic velocities in the near-surface.