Abstract: This paper deals with the problem of correlation region in the case of one site, maneuvering multi-target tracking. The statistical properties of the norm of the innovation vector, the size of the correlation region and the probability of accepting a correct return are investigated by means of Monte Carlo simulation with three different state models, different maneuvering accelerations of targets and the variance of state noise.In Model I the "current" probability density function of target maneuvering acceleration is described by a modified Rayleigh density with variable mean-value and therefrom an adaptive filtering algorithm for the mean and variance of the maneuvering acceleration is developed.In Model II the maneuvering acceleration of the target is assumed to have an approximately uniform density function in the interval （-amax, amax） where amax is the maximum maneuvering acceleration and a standard Kalman filtering algorithm is used.Model III is a two-state model （only position and velocity of targets are involved） which has been used by several authors for non-maneuvering multi-target tracking.The investigation shows that Model I has the best capability to adapt itself to target maneuvers. The mean and the mean square values of the norm of innovation vector in this model have very good relative stability when the target maneuvers over a wide range. The probability of accepting the correct return and the size of correlation region do not change drastically with the changes of the variance of state noise and of maneuvering accelerations of targets. Therefore, it is not necessary to take any other measures to maintain the probability of correct correlation and the size of correlation region to be constant. In this sense, Model I is very suitable to application in maneuvering multi-target tracking.