Summary. Earth-tide tilt and gravity observations from five locations in Nova Scotia, Canada are used to study both the structure of the Earth beneath Nova Scotia and the structure of the M2 ocean tide around Nova Scotia. The observed Earth tides are functions of the spatial distribution of the ocean tides and the elastic properties of the Earth. It is therefore possible, in principle, to invert tidal loading observations for both the distribution of the ocean tides, and the elastic properties of the Earth. In practice our observations from Nova Scotia are imperfectly suited to this task. They contain both observational errors and are poorly distributed. However, it is possible to separate the loading due to tides in waters adjacent to Nova Scotia from that due to the remaining ocean, thereby removing errors that result from uncertainties in the global tide distribution. This is done by considering the spatial derivative of the tilt, since the tilt due to distant loads is virtually the same at all stations within a small geographic area. Differencing pairs of stations nearly eliminates the load tilt caused by distant oceans. Using differenced tilt data the problem is reduced to the estimation of the elastic properties of the Earth beneath Nova Scotia because the tidal load in adjacent waters is reasonably well known. The crustal model inferred from a comparison of observed and theoretical difference tilts is in agreement with the results of seismic refraction experiments conducted along the Atlantic coast of Nova Scotia. Both tilt and seismic data suggest that the crust is 35 km thick, consists of two major layers and is underlain by normal mantle. Having determined an approximate crustal structure, the tilt and gravity observations (as opposed to difference observations) are used to demonstrate that the M2 tidal distribution in the northwest Atlantic is intermediate between the co-tidal charts of Tiron et al. and Zahel and is in fact closely described by Dietrich's empirical chart. A finite element model of loading on the continental margin is used to demonstrate that the lateral changes in crustal structure across the continental margin in no way influence the conclusions. The results also indicate that variations in crustal structure will modify the gravity load tide by at most 10 per cent. Consequently, the inverse problem of inferring the global tide distribution from gravity observations may be considered to be linear to a first approximation.