Three data acquisition schemes for two-particle coincidence experiments with a continuous source are discussed. The single-start/single-stop technique, implemented with a time-to-pulse-height converter, results in a complicated spectrum and breaks down severely at high count rates. The single-start/multiple-stop setup, based on a time-to-digital converter and the first choice in today's similar coincidence experiments, performs significantly better at high count rates, but its performance is still hampered if the time-of-flight range is large, and the false coincidence background is variable if the event frequency and the collection efficiency of the starts are both high. A straightforward, multistart/multistop setup is proposed for coincidence experiments. By collecting all detector data, it ensures the highest signal-to-noise ratio, constant background, and fast data acquisition and can now be easily constructed with commercially available time-to-digital converters. Analytical and numerically evaluated formulas are derived to characterize the performance of each setup in a variety of environments. Computer simulated spectra are presented to illustrate the analytically predicted features of the various raw time-of-flight distributions obtained with each technique.