Microsoft Windows or DOS based experiments?
For hosting mission critical studies DOS can be a better bet as it is single tasking and good DOS based experiment generators (ERTS etc.) can hit the hardware directly and obtain better presentation and response timings. However, these days obtaining hardware that is DOS compliant, e.g. VESA compatible graphics cards and specific model ISA soundcards can be hard. To help ensure a similar level of accuracy you should check your own paradigms using the Black Box Toolkit.