The theory of seismic interferometry redicts that crosscorrelations of recorded seismic res onses at two receivers yield an estimate of the interreceiver seismic res onse. The interferometric rocess a lied to surface-reflection data involves the summation, over sources, of crosscorrelated traces, and it allows retrieval of an estimate of the interreceiver reflection res onse. In articular, the crosscorrelations of the data with surfacerelated multi les in the data roduce the retrieval of seudo hysical reflections (virtual events with the same kinematics as hysical reflections in the original data). Thus, retrieved seudo hysical reflections can rovide feedback information about the surface multi les. From this ers ective, we have develo ed a data-driven interferometric method to detect and redict the arrival times of surface-related multi les in recorded reflection data using the retrieval of virtual data as diagnosis. The identification of the surface multi les is based on the estimation of source ositions in the stationary- hase regions of the retrieved seudo hysical reflections, thus not necessarily requiring sources and receivers on the same grid. We have evaluated the method of interferometric identification with a two-layer acoustic exam le and tested it on a more com lex synthetic data set. The results determined that we are able to identify the rominent surface multi les in a large range of the reflection data. Although missing near offsets roved to cause major roblems in multi le- rediction schemes based on convolutions and inversions, missing near offsets does not im ede our method from identifying surface multi les. Such interferometric diagnosis could be used to control the effectiveness of conventional multi le-removal schemes, such as ada tive subtraction of multi les redicted by convolution of the data.