S U M M A R Y2-D grid skeletonization, modelled after seismic skeletonization, can be applied to any gridded data, including images commonly used in the geosciences. It affords a map of distinctly separable discontinuities together with a suite of attributes associated with each discontinuity. Such a map, with an accompanying catalogue of attributes, offers a broad scope for data analysis such as directional or any attribute-oriented filtering. The versatility of our method lies in the choice of the primitive feature set used for pattern recognition, and in the two-pass application of the detection process. Application of the method to different data sets, including images of drainage basins, potential field data, reflection seismic data and invertebrates provide examples of how attribute catalogues are analysed to extract additional data-dependent information such as directional, event length, pulse width and non-linearity characteristics.