A term that refers to patterned variation among spatial phenomena. The great majority of statistical inferential procedures used in archaeological predictive modeling assume independent observations, that is, that the values of some observations cannot be predicted (at a better than random chance) from the known values of other values. Since spatial phenomena generally exhibit patterned variation, or autocorrelation, the assumption of statistical independence is violated. As a result, statistical significance tends to be overestimated. In models of this kind, then, the effect of autocorrelation has to be controlled or at least taken into account.