Seismic data contains reflections along with diffractions. Diffractions in seismic data are related to different features like faults, fractures, pinchouts and rough edges of salt bodies. It follows that suppression of reflections and enhancement of diffractions can be very useful for seismic interpretation of various subsurface features that cause diffractions. During the last years, different techniques have been proposed to separate reflections from diffractions. In this study both simple synthetic test data and complex synthetic data (Sigsbee2a) have been used to investigate the modified version of the Common Reflection Surface (CRS) technique. The simple synthetic data were sorted into different domains to investigate the difference in signature of reflections and diffractions. A two-step procedure based on CMP-sorting followed by zero-offset (eventually constant-offset) sorting seemed to give a good diffraction separation. Optimal parameters were determined based on Semblance as a coherency measure. This separation scheme was applied to a complex controlled data set (Sigsbee2a) and diffractions were successfully separated from reflections.