At subduction plate boundaries, strains accumulate within the overriding plate in response to motion of the down-going plate. To a first approximation, margin-normal interplate slip and shortening in the overriding plate sum to the total plate convergence. Therefore margin-normal interplate slip can be estimated by measuring shortening in the overriding plate and total plate convergence. Here we quantify margin-normal interplate slip by measuring shortening of the upper plate along a profile which extends up to 480 km across, and approximately normal to, the central Hikurangi Margin, New Zealand. Long-term deformation is estimated for periods of millions of years using data from seismic reflection lines and cross sections, while contemporary strain rates are derived from Global Positioning System (GPS) geodetic velocities. These data permit comparison of plate convergence rates determined by seafloor spreading data with the rates of contraction over short (~7 years) and long (~2.5 and 5 Myr) periods of time. On this part of the plate boundary, total margin-normal shortening in the overriding plate over the last ~5 Myr was ~9--28 km and accounts for ~6--19% of the total convergence (145--160 km) at rates of ~2--6 mm/yr. The remaining ~80% or more of plate convergence has been accommodated by slip on the subduction thrust. GPS contraction along approximately the same profile indicates ~14--16 mm/yr of contemporary shortening across the terrestrial part of the upper plate, which is approximately modeled by ~80% interplate coupling on the subduction interface down to depths of ~30 km. The contrasting proportions of inferred interplate coupling for contemporary and geological data sets are consistent with stick-slip behavior on the plate interface and with slip being principally achieved during intermittent large earthquakes. The general shapes of short- and long-term shortening profiles are similar and suggest that the highest strains occur at, or immediately east of, the Mohaka Fault. These strain profiles are consistent with a model where high strains in the upper plate are concentrated near the down-dip end of the contemporary locked zone on the plate interface. This model requires that the down-dip end of the interseismic locked zone has, on average, been approximately stable over a period of at least 2.5--5 Myr.