A detailed theoretical analysis is presented of the dephasing signal measured in four-beam picosecond time-resolved stimulated-Raman-gain experiments. The analytical expressions derived are successful in simulating observations on the CS2 model system under various experimental conditions. It is established that the tail of the curve for sufficiently long delay times is uniquely determined by the vibrational Raman term that yields the vibrational dephasing time T2. It is also found that the rotational dephasing time cannot be fundamentally determined from the rotational Raman peak centered around zero delay. Furthermore, it is shown that the coherence properties of the stimulating laser fields, as distinguished from their pure pulse-envelope properties, are crucially important in determining the time resolution of the experiments and for explaining several experimental phenomena observed under mismatched excitation-pulse conditions. In this respect, the domain model for mode-locked pulses has proved to be useful. It is found that the time resolution of the dephasing measurements is determined by the combined pulse-coherence time of the driving fields and not by the cross-correlation width of the pulse envelopes.