This paper deals with the preliminary experiment of the automatical classification and mapping of soils by means of remote sensing method which was part of the research work of the author in Mexico. Experimental results showed that the automatical classifieation method not only raised the working efficiency by nearly twofolds, but also increased the cartographical precision up to the level required by a soil map on the scale of 1:250,000. But different soil types vary in mapping pattern error. According to the errors occurred from minimum to maximum, the soils are arranged as follows: (a) pellic vertisol. deep phase, (b) orthic phaeozem, deep phase and cnomire luvisol, deeply lithic phase, (c) pellic vertisol, deeply lithic phase and lithic phase, (d) orthic phaeozem, deeply lithic and lithic phase;natric vertisol, deep phase;natric phaeozem, deep phase. Of course, some mapping patterns of the soils with great errors are below the level of cartographical precision required.