I have used GS in developing reaction time tests. The maximum accuracy with the engine is one frame-cycle. This can be as little as 5 msec but is computer and program dependent. Another serious concern for mseconds of accuracy is the time it takes for the BIOS or operating system to poll the input devices and the hardware response times of those devices. This is the reason that I try to have data be a difference between two measures.

I am very curious what technology you will use that can be near an MRI scanner.

Thanks