Spatial information on soils generally results from local observations that are destructive and time consuming especially in the case of heterogeneous soils. Geophysical technics can be a great help for soil mapping since they are non-destructive and fast. Large areas can be surveyed with a high density of measurements. Electrical resistivity is particularly interesting for soil study because it covers a wide range of values (several decades) and depends on many characteristics of the soil. The main objective of this paper was to study soil spatial variability using an original approach to electrical data processing. Electrical data from a 3-depth survey, usually treated as three apparent resistivity maps, were considered as many electrical soundings each with three apparent resistivity values. The study of the vertical succession of these values led to define nine reference geophysical taxa. This taxonomy relies on a parameter (α) which discriminates soil layers and is defined as the interval (in Ω m) beyond which two successive apparent resistivity values on a sounding are considered different. Geophysical taxa mapping highlighted their spatial coherence, which was related to pedological characteristics such as the presence of a clay layer or the depth of the soil profile. The comparison between the spatial distribution of geophysical taxa and a pre-existing soil map showed that the delineations of taxa clusters closely matched soil unit boundaries but were much less smooth. This method allowed the assignment to each soil type of a specific apparent electrical resistivity profile which was consistent with soil profile description. The method is straightly applicable to data from other surveys, and opens the way to the development of semi-automatic soil mapping from electrical resistivity data.