- no independent timing, fast keyboards
* DirectIN
* Black box toolkit "millisecond keyboard"
These should all work. RTBox and XXXPixx have independent clocks, good hw and vendor support and good PTB support, but the ResponsePixx probably only makes sense with a Data/View/ProPixx device. RTBox can work well in a E-Prime compatibility mode afaik. RTBox website seems to have problems atm., but "Xiangrui Li" would be the person to contact via e-mail, a little Google will get you there.
CMU/PST/Cedrus are horribly misdesigned and overprized pieces of hardware, but our CMUBox driver can do response collection and timestamping on a separate realtime background processing thread, so in practice timestamping should be roughly in the 1-3 msecs accuracy range on a properly working operating system - with a real and properly configured serial port around 1 msec, with a properly set up USB-serial converter maybe 1-3 msecs. The fORP is also handled by CMUBox, as is the Bitwhacker. Essentially CMUBox can be hacked or used to handle anything that sends single Bytes over serial port in response to a button press, or anything that uses the disaster that is the PST/CMUBox "protocol".
DirectIN and Blackbox i never used, but with custom built hardware/firmware one can in principle get 1-2 msecs timing resolution out of a USB bus and our KbQueue functions can collect and timestamp keyboard/mouse/joystick button presses also on a realtime thread, so also in the 1-3 msecs range. The main inaccuracy with keyboards/mice comes from the keyboard/mouse hardware and firmware, so normal mouse/keyboard -> BAD, special Keyboards or some special gaming mice -> Probably good.
Parallel ports are cheap and effective, but the downside is that you need to poll for activity in some tight loop, no automated precise background response collection, at least until PTB gains a proper Parallel port driver plugin, a feature i often promised but that is probably still a few months away - new always shifting ETA is sometime 2014.
-mario