Because of its importance to many Earth science analyses, it is worth assessing whether gravity modelling can be simplified depending on the intended purpose and required precision. While it is obvious that large-scale gravity studies should account for the sphericity of the Earth, each case should be examined on its own merits. Demonstrations are useful for providing estimates of the errors in much simpler 2D modelling. The example of the Mid-Atlantic Ridge serves to compare "large" 2D and spherical 3D models. My model extends horizontally +/- 2,000 km (+/- 18 degrees) from the model profile across and along the straight ridge axis (along a great circle) and to a depth of 82 km across the axis. 3D modelling would generally be considered obligatory, but this is not clearly necessary from this study. The density structure is highly idealised, the asthenospheric uplift or lithosphere thinning is simplified. The Bouguer anomaly is fitted by least-squares for the density contrast, and the 2D-3D difference of the results is taken as the error. A lithosphere-asthenosphere density contrast of 86.56 kg/m(3) was computed for the 2D model, and 84.14 kg/m(3) for the spherical model. The difference is small, in the order of 3%, well within all the other uncertainties. My study shows that despite the significant sphericity of the structure, 2D models are well suited for such ridge studies, or generally for models with a laterally extended layered structure, and that spherical modelling can be applied discriminately.