Decision-theoretic and syntactic pattern recognitions are applied to the two-dimensional seismogram for the detection of bright spots. Tree classification technique is used in the decision-theoretic apporach. Because of frequency attenuated effect in a seismogram, partitioning-method and tree classification are used to detect the candidate bright spots. In the syntactic pattern recognition, the relation between error probability and Levenshtein distance is proposed. Three kinds ot string distance computation are proposed to test the continuity of a bright spot pattern. The experiment on a relative-amplitude real stacked seismogram xs presented The classification results are quite good and the proposed method can be used to improve seismic interpretation.