The factors influencing initiation of detonation in gap tests for liquid explosives are investigated experimentally. A calibrated donor charge (nitromethane) and PMMA attenuator disk arrangement are used to transmit shocks of known strength (2-10 GPa) into a test explosive of nitromethane sensitized with 5% diethylenetriamine. The test explosive is contained in capsules of different wall materials (PVC, Teflon, aluminum), and the dimensions of the charges vary from 25 mm to 100 mm in diameter. For the small-scale charges, the presence of the confining wall of the test capsule is seen to have a pronounced effect on the detonation initiation. Certain wall materials (PVC, Teflon) exhibit a multi-valued critical gap thickness, meaning that a weaker shock may result in initiation while a stronger shock does not. The effect of the wall materials could not be correlated with their acoustic or shock impedance, and the only way to eliminate these effects was to make the diameter of the test charge larger than the donor charge. When the size of the donor charge was increased, the critical pressure required for initiation decreased. These results could be correlated to "ideal" shock initiation experiments that use flyer plates as shock sources assuming that lateral rarefactions quench detonation initiation if they reach the central axis of the charge before the onset of detonation is complete.