Spatial sutocorrelation is one of the basic characteristics of regionalized variable, and statistics based on spatial autocorrelation can be used to test if the distribution of a variable in a two-dimentional space follows a certain pattern or at random. Values of Moran's I of trace element contents in soil of Tianjin plain area were calculated. The spatial sutocorreiation characteristics, the directional characteristics of spatial autocorrelation, the relationship between spatial autocorrelation and lag of sapmling and the relationship between spatial autocorrelation in different directions and lag of sampling were investigated.Under the sampling density of 300 km² per sample, significant by positive spatial sutocorrelations were detected for most trace elements studied. The most significant spatial autocorrelation was found in horizon A where the original distribution of the elements had been modified remarkably by natural forces and human activities. The spatial structure characteristics of Hg, Cd, Pb, Zn, and As contents in surface soil were further altered by pollution and agricultural practice.