Kohonen's Self-Organizing Map method can be used in several different ways for lithologic interpretation. One of the most obvious ways is to use a typical portion of a 2-D or 3-D data set to train the network; then use the reference weights as classifiers for the rest of the data set. If we have a lithological column, we can experiment with a number of different and logical attribute combinations to see which clusters are most similar to the lithological column.