Conditional sampling techniques have been studied in order to better understand the phase relationship between the signal used to detect the events and the sampled data signal. It is shown that whenever there is a mean time delay between these two signals or whenever they are obtained from two different spatial locations, a random phase will exist between them if the signals are obtained from a random environment. This randomness is not accounted for in the usual definitions of conditional sampling, and it can cause a serious degradation of the conditionally averaged results. A corrective method is discussed and illustrated with examples from a turbulent boundary layer.